This article has been accepted for publication in IEEE Intelligent Systems but has not yet been

fully edited.
Some content may change prior to final publication.
1

Demand Response Management in Power

Systems Using a Particle Swarm Optimization
Approach
Pedro Faria, Zita Vale, Joo Soares, Judite Ferreira
AbstractDemand response (DR) is not a new concept but it is gaining a growing focus of attention in nowadays electric power systems
operation and planning, with several advantages for the reliable power system functioning and for electricity prices. In this paper, price-based
DR is applied to electricity consumers through the management of electricity prices. This management is based on demand elasticity and
consumers are expected to react enabling to accomplish the required load reduction. The methodology is implemented in a developed DR
simulator DemSi - that uses PSCAD for technical validation of solutions and Particle Swarm Optimization (PSO) for solution optimization.
The performance of PSO is evaluated in terms of running time and obtained solutions in comparison with the Non-Linear Programming
(NLP) solutions obtained in GAMS. Case studies involving 32 and 320 consumers are used to illustrate the proposed methodology and to
discuss its performance.
Index Terms Demand response, distribution network, electricity markets, particle swarm optimization, power systems, simulation.

1 INTRODUCTION
Electricity markets have arisen as result of the power sector
restructuration and power systems deregulation. The players
participating in the competitive electricity markets must define
strategies and take decisions using all the available information
and business opportunities to accomplish their goals [1]-[3].
Demand Response (DR) has proved to be a good opportunity
for loads to participate in this environment, gaining competitive
advantage, and represents significant benefits for the whole
electricity market performance. DR programs may produce an
increase in power consumption efficiency through active
consumer participation, making evident the value that each
consumer attributes to his individualized additional demands.
Recent efforts are aiming at improving wholesale markets
with more intensive use of DR. This includes, for example, the
acceptance of demand bids/offers for ancillary services; the
specification by the DR resources of the frequency, duration and
the amount of their participation in consumption reduction; and
the existence of aggregators that bid into the market on behalf of
customers [4].
Making full use of all the advantages due to active consumers
participation requires an infrastructure able to accommodate all
centralized and distributed energy resources, including intensive
use of renewable and distributed generation, storage, electric
vehicles, and demand response, seeing consumers as active
players, in the context of a competitive business environment.
This structure corresponds to the practical implementation of the

x Pedro Faria , Z ita V ale, Joo Soares and Judite Ferreira are with GECA D
Knowledge Engineering and Decision Support Research Group /
Polytechnic of Porto, Portugal. E-mail: pnf@isep.ipp.pt, zav@isep.ipp.pt,
japs@isep.ipp.pt, mju@isep.ipp.pt .

Digital Object Indentifier 10.1109/MIS.2011.35

concept of smart grids [5] for which significant efforts are

presently taking place, increasing DR relevance. In practice, DR
programs implementation is in an initial stage using technologies
basically close to smart metering.
This paper presents DemSi, a demand response simulator that
has been developed by the authors to simulate the use of DR
programs. DemSi uses PSCAD(1) for network simulation and
provides users with optimized DR actions management.
The available DR opportunities should be used in the best way
to attain the involved agents goals. This leads to highly time
consuming complex optimization problems, requiring huge
computational means. Traditional optimization methodologies are
usually not able to cope with this kind of problems for realistic
cases. Artificial intelligence techniques have been used to address
several problems in the scope of power systems and electricity
markets [6]-[7]. Particle Swarm Optimization (PSO) [8]-[9] has
been successfully applied to power systems [10] and is proposed
in this work to address DR management.
The DR problem considered in this paper is triggered by a
need to reduce the energy supply to a set of consumers which are
managed by an aggregator, aiming at minimizing the global value
paid by the consumers. Consumers price elasticity, which relates
demand reduction with price increase, is used to determine the
optimal individual load reduction and price increase, using the
proposed PSO based m ethod ology.
The paper includes case studies involving 32 and 320
consumers for which the performance of the PSO approach is
compared with the Non-Linear Programming (NLP) solutions
obtained in GAMS(2).
After this introduction section, section 2 explains the demand
response concepts and programs and their implementation by
market players in electricity markets. Section 3 presents the
1

http s:/ / pscad .com / hom e/

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
2

proposed methodology that has been implemented in the simulator

presented in section 4. Section 5 presents a case study. Finally, the
most important conclusions of the work are presented in Section
6.

2 DEMAND RESPONSE CONCEPTS AND PROGRAMS

Demand Response (DR) includes all intentional electricity
consumption pattern modifications by end-use customers that are
intended to alter the timing, level of instantaneous demand, or
total electricity consumption [11] in response to changes in the
price of electricity over time. Further, DR can be also defined as
the incentive payments designed to induce lower electricity use at
times of high wholesale market prices or when system reliability
is jeopardized [12].
An important advantage of demand response implementation
is the postponement of investments in generation resources and
transmission/distribution lines. This is highly important when the
generation is near its maximum capacity with exponentially
increasing generation costs. In these conditions, a small reduction
in load will cause a big reduction in generation costs and,
therefore, a significant reduction in the price of electricity.
Usually the actions that result from demand side behavior or
those that are intended to manage consumer behavior are referred
as DR, load management and Demand Side Management (DSM).
Traditionally these are seen as measures taken to encourage
consumers to reduce their electricity consumption during times of
especially high demand [13] and usually done through utility load
management programs aiming essentially at obtaining peak
reduction. Competitive electricity markets enable a wide set of
new opportunities for more strategic consumers behavior and
new models of demand response.
Several studies have proved that loads are not rigid, exhibiting
elasticity that can be used for mutual benefits of power systems
and consumers. Changes in electricity price over time and
incentive payments give place to increased demand flexibility as
end-use customers intentionally modify their electricity
consumption patterns as a response to exterior stimulus. DR can
be contracted over longer or shorter periods either as result of its
inclusion in capacity markets or directly through bilateral
contracts.
DR, including Real time pricing, can be used as a means to
optimize distribution network operation, to reduce incident
consequences, and to reduce wind curtailment [14].
DR programs can be used both to increase and to decrease
load demand. The use of DR to obtain a reduction in load
consumption in peak, congested, and/or incident periods is
favorable to distribution network operators relieving the network
components. In this way, important gains in reliability and service
quality, cost minimization, and even avoidance or deferral of
network investments can be achieved. In case of incidents, the
contracted load curtailment is expected to minimize the monetary
global value of the non-supplied load. The increasing of wind
based electricity generation and the wind intermittent nature often
2

lead to periods with excess of generated energy. This imposes

relevant losses in wind curtailment making wind farms less
efficient and the corresponding investment payback period higher.
The use of DR programs to increase load consumption in these
periods helps to overcome this problem.
Price elasticity is a measure used in economics to evaluate the
responsiveness of the demanded quantity of a good or service to a
change in its price, or the percentage change in quantity demanded
on response to a one percent change in price [15]. In what
concerns electric loads, price elasticity is a normalized measure of
the intensity of how usage of electricity changes when its price
changes by one percent. In the opposite way, demand elasticity is
a measure of how price changes when usage of electricity
changes.
Demand price elasticity can be evaluated using (1),
where Quantity is the quantity of the usage of the good or
service and Price is the price of this good or service [16].
'Quantity and 'Price refer to the quantity of usage and price
variations respectively, between the periods before and after the
implementation of DR programs.

'Quantity

'Price

Quantity

(1)

Price

Demand response programs can be divided in two wide

groups: price-based demand response and incentive-based demand
response [17].
Price-based demand response is related with the changes in
energy consumption by customers in response to variation in the
prices they pay. This group includes Time-Of-Use (TOU), Real
Time Pricing (RTP), and Critical-Peak Pricing (CPP) rates. For
different hours or time periods, if the price varies significantly,
customers can respond with changes in energy use. Response to
price-based demand response programs is entirely voluntary.
TOU is a rate that includes different prices for usage during
different periods, usually defined for periods of 24 hours. This
rate reflects the average cost of generating and delivering power
during each period.
RTP is a rate in which the price of electricity is defined for
shorter periods of time, usually one hour, reflecting changes in
wholesale price of electricity. Customers usually have the
information of prices on a day-head or hour-ahead basis.
CPP is a hybrid of the TOU and RTP programs. The base
program is TOU and a much higher peak pricing is used under
specified conditions (e.g. when system reliability is compromised
or supply costs are very high).
Incentive-based demand response includes programs that give
customers incentives which may be fixed or time varying and that
are complementary to their electricity rate. These can be
established by utilities, load-serving entities, or a regional grid
operator. Some of these programs penalize customers that fail the
contractual response when a priori specified events are declared.
This group includes programs such as Direct Load Control (DLC),
Interruptible/Curtailable Service (ICS), Demand Bidding/Buyback

http :/ / w w w .gams.com /

Digital Object Indentifier 10.1109/MIS.2011.35

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
PEDRO FARIA ET AL.: DEMAND RESPONSE MANAGEMENT IN POWER SYSTEMS USING A PARTICLE SWARM OPTIMIZATION APPROACH

(DBB), Emergency Demand Response (EDR), Capacity Market

(CM), and Ancillary Services Market (ASM) Programs.
The expected responsibilities and functions of each player in a
deregulated electricity market are exposed in [18]. The authors
state that larger participation is required for DR to be viable in the
scope of electricity markets, and for this, it is required a more
intensive collaboration between regulators and also between these,
market participants, and market and system operators.
[19] proposes a new complex-bid market-clearing mechanism
that considers price-sensitive bids made by consumers. The effect
of the increasing participation of demand-side on the various
categories of market participants is quantified. It is concluded that
the increase of demand shifting causes the reduction of marketclearing prices, benefiting all bidders even if they do not
participate in the shifting activities.
Direct Load Control (DLC) is a demand response model for
which the utility is able to control customer equipment. This
control has shown positive results avoiding the use of additional
generation. Implemented models have been applied to air
conditioner and water heaters. However, some problems related
with the good functioning of the switches installed by the utility
have been reported. In future, with further implementation of the
smart-grid concept, with two way communication ability, it is
expected that utilities have a better control over the target
equipment. This is important since incentive payments are usually
monthly paid to the customers as fixed rates [20]. Although this
concept is not in a state that can be presently considered as useful,
utilities are planning to offer programs where price signals are
given to customers through smart pricing programs. These signals
are expected to give incentives to customers so that they make
their own investments in DR equipment [20].
The present paper proposes a methodology that considers the
demand response to the electricity price variation imposed by the
Distribution Network Operator (DNO) in the presence of a
reduction need due to any reason such as a lack of generation, or
high electricity market prices. The proposed method is based on
the use of real time pricing. This price-responsive approach could
be adequate to fight the difficulties in the monthly fixed
remuneration programs, like DLC, and when the concept of smart
grid is an initial stage of implementation.

3 PROBLEM FORMULATION AND RESOLUTION

The DR problem that we are considering in this paper is
triggered by a need to reduce the energy supply to a set of
consumers by a specified amount. This event is managed by these
consumers aggregator, aiming at minimizing the global value
paid by the consumers. To achieve its goal, the aggregator
considers the individual consumers price elasticity which relates
demand reduction with price increase. Individual load reduction
and price increase for each consumer are determined as the
solution that minimizes the consumers global cost. This section
presents the optimization problem mathematical formulation and
the d evelopm ent of the PSO solver.

Digital Object Indentifier 10.1109/MIS.2011.35

3.1 Problem Formulation

The problem to be solved consists in the optimal minimization
of the global costs from the point of view of electricity consumers,
regarding loads managed by a load aggregator. Problem
characteristics lead to a non-linear model.
In practice, when a reduction in electricity consumption is
needed, the aggregator, based on the knowledge about consumers,
raises the price of electricity expecting that consumers reduce the
usage of electricity. The objective fu nction can be expressed as
in (2) and is su bjected to several constraints.

 ELoad ( c )  ELoadRed ( c )  u

1
Price
Price



EnergyInitial ( c )
EnergyVar ( c )

nc

Min Cost

u
c

(2)

The objective function (2) aims the minimization of the costs

with the electricity consumption (i.e. the total amount consumers
pay for their electricity consumption), when a reduction in the
overall demand is required. These costs can be calculated based
on the final load demand (initial load demand minus demand
reduction value) and on the final price (initial price plus the price
increment used to obtain the required consumption reduction).
Limitations are imposed to power (3) and price (4) values
variations of each customer, according to the extent in which they
can and/or want to participate in the demand response program
and to their price elasticity. Power system operation requires the
balance between load and generation to be guaranteed at all times
(5).
The consideration of load response is formulated based on
elasticity (6) values. Since the elasticity is considered as a fixed
and constant value for each load, the optimal relation between
load and price variation is determined in the optimization. The
present study considers the obligation of having the same price
variation for the loads of the same type as expressed in (7).

PLoadRed ( c ) d MaxPLoadRed ( c )

(3)

PriceEnergyVar ( c ) d MaxPriceEnergyVar ( c )

(4)

PMain  PReserve

nc

nc

Load ( c )

c 1

Elasticity( c )

 PLoadRed ( c )

(5)

c 1

PLoadRed ( c ) u PriceEnergyInitial ( c )

PriceEnergyVar ( c )
w here

PLoad ( c ) u PriceEnergyVar ( c )

PriceEnergyVar (T ), c T

(7)

Cost

Total load consu m ption cost

Price
EnergyInitial (c)

Initial
electricity
consu m er c

0885-9000/$26.00 2011 IEEE

(6)

price

for

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
4

Price
EnergyVar(c)

Elasticity
(c)
E
LoadRed (c)
E
Load (c)

nc

Variation in consu m er c electricity

price
Price elasticity for consu m er c
Energy consu m ption red u ction of
consu m er c
Initial energy consu m ption of
consu m er c
N u m ber of consu m ers

Maxim u m perm itted variation in

MaxPrice
EnergyVar(c) energy price for consu m er c

MaxP
LoadRed (c)

P
Load (c)
P
LoadRed (c)

Maxim u m perm itted variation in

pow er for consu m er c
Initial pow er consu m ption of
consu m er c
Pow er consu m ption red u ction of
consu m er c
Pow er received from the m ain grid

P
Main
P
Reserve

Reserve pow er

Consu m er type

3.2 Particle Swarm Optimization

The optimization of the formulated non-linear problem
consists on the minimization of a multimodal function with many
local minima and a global optimum. This is considered a NP-Hard
problem because the computational complexity is high even in
simple cases.
In the last decades Artificial Intelligence (AI) techniques have
deployed a set of effective and efficient methods to mitigate the
difficulties of solving complex problems, namely in terms of
computational time. These algorithms explore a given search
space and return the best solution found. The Particle Swarm
Optimization (PSO) [8] belongs to the category of Swarm
Intelligence methods [9] and was used in this work to solve the
demand response problem due to being effective in difficult
optimization tasks namely non-linear problems [21]. The used
algorithm can be described as follows:

4 SIMULATOR
DemSi is a demand response simulator that has been
developed by the authors to simulate the use of DR programs.
PSCAD is used as the basis platform for the network simulation
allowing to have detailed models of electrical equipment and to
consider transient phenomena. This is very relevant to analyze the
technical viability of the DR proposed solutions, both for steady
state and transients although network response to the changes in
the loads is not being presented in this paper.
DemSi considers the players involved in the DR actions and
results can be analyzed from the point of view of each specific
player. This includes four types of players: electricity consumers,
consumer aggregators, electricity retailers (suppliers) and
Distribution Network Operator (DNO). In this paper the case
study is analyzed from the point of view of a consumers
aggregator.
Consumers can be characterized on an individual or in an
aggregated basis. Based on their profiles, some clients can
establish flexible supply contracts with their suppliers. The
information concerning the quantity of load that can be cut or
reduced and the corresponding compensations for each client are
considered by DemSi.
The loads are classified in 5 main types in function of their
peak power consumption, destination of energy, and load diagram.
These types are:
x Domestic (DM);
x Small Commerce (SC);
x Medium Commerce (MC);
x Large Commerce (LC);
x Industrial (IN).
Figure 1 shows the general architecture of DemSi.

1. START
2. Initialization of parameters (maximum velocities, minimum
velocities, position limits, maximum iterations)
3. Random generation of initial values (swarm)
4. REPEAT
5. Reproduction: Each particle generates 1 new descendent
(movement, new position)
6. Evaluation: Each particle has its fitness value, according to its
current position in search space
7. Store the best solution of swarm
8. UNTIL termination criteria (Number of generations)
9. END PSO
The results and the performance of this technique were then
compared with those obtained with conventional techniques using
the professional optimization tool GAMS.

Digital Object Indentifier 10.1109/MIS.2011.35

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
PEDRO FARIA ET AL.: DEMAND RESPONSE MANAGEMENT IN POWER SYSTEMS USING A PARTICLE SWARM OPTIMIZATION APPROACH

START

Network data

Supply information
Network simulation
(PSCAD)

Consumer
knowledge base
Simulation
timeline events

END
Yes

Simulation
timeline end
?

No
Demand response
program management

Figure 1 DemSi general architecture

To fully attain our goals, PSCAD is linked with

MATLAB(3) and GAMS. These links allow using
programmed modules able to model the relevant players behavior
and all the relationships among them, namely the contracts
between each client and each supplier. The solution of the
formulated optimization problem is found using MATLAB
and/or GAMS. Using diverse approaches for solving the
optimization problems, it is possible to derive the best approach
for each type of situation.
Once the simulation is started, a simulation timeline is used to
feed DemSi with time tagged events. All the physical phenomena
related with the power system are simulated by PSCAD . The
management of DR programs is undertaken by a module
developed in MATLAB. The simulation end is determined by the
end of the simulation timeline.
Every time the simulator is initiated an initial state (e.g. value
of loads, state of breakers, etc.) is considered as the departing
simulation point. Once the simulation is launched, the supply
information and the consumer knowledge base have the required
information that allows optimizing demand response program use
over time, allowing simulation to go on. The demand response
programs management module optimizes the use of demand
response opportunities for each situation.

5 CASE STUDIES
This section illustrates the use of the proposed methodology in
the developed demand response simulator DemSi. The case study
considers a distribution network with 32 buses from [22] which
the authors evolved to 2040 in terms of load characterization [23].
All the results presented in this paper are obtained for two
scenarios with 32 and with 320 consumers.
5.1 Case characterization
In the first scenario, with 32 consumers, the results are
obtained for a period for which the load demand is presented in
Table I. For the second scenario, there are 10 consumers in each
bus, corresponding to a total of 320 consumers. In this scenario,
the 10 loads connected to each bus have the total power presented
in Table I and are of the same load type. Table I also presents the
type of each consumer.
In both scenarios, it was considered that all the loads of the
same type have the same price variation during the application of
demand response program. Each scenario and reduction need are
solved by two approaches - with the developed PSO module and
with Non-Linear Programming (NLP) implemented in GAMS;
results are compared in terms of time of execution and solutions
values.
The values of elasticity are 0.14, 0.12, 0.20, 0.28, and 0.38,
respectively for DM, SC, MC, LC, and IN consumers types. The
corresponding values for electricity price, which correspond to the
values of flat-rate tariff of retailer are 0.18, 0.19, 0.20, 0.16, and
0.12, in /kWh.
As a restriction of the proposed formulation, a price cap and a
power cap are considered; these cap values can be parameterized
for each case study. For this case study the price cap is equal to
150% of the value of energy price and the power cap is 15% of
the power consumption value for every customer.
The demand response program use is triggered by a load
reduction required by the supplier. A set of seven reduction values
are considered for each scenario. For each reduction requirement,
the energy price for each consumer type and the load reduction for
each consumer are obtained as a result of the optimization
problem.

http :/ / w w w .mathw orks.com /

Digital Object Indentifier 10.1109/MIS.2011.35

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
6

Table I Consumers data

Bus

Power (kW)

Consumer Type

Bus

Power (kW)

Consumer Type

Bus

Power (kW)

Consumer Type

169.1

MC

12

91.3

DM

23

674.8

IN

148.9

SC

13

181.3

MC

24

669.3

IN

147.1

SC

14

91.1

DM

25

93.8

DM

145.5

SC

15

91.1

DM

26

93.2

DM

94.2

DM

16

91.9

DM

27

92.2

DM

311.1

LC

17

135.5

SC

28

183.0

MC

308.7

LC

18

152.4

MC

29

295.3

MC

89.3

DM

19

151.7

MC

30

225.4

MC

90.6

DM

20

151.6

MC

31

315.1

LC

10

67.0

DM

21

151.5

MC

32

89.8

DM

11

91.1

DM

22

147.3

SC

Total

5831.3

--

5.2 PSO Application Details

The problem described in this paper has 5 variables that are the
price variation upper limits for the 5 load types. In PSO these
variables are easily coded, i.e. each particle has a dimension space
of 5. Table II shows the results of the performance and parameters
sensitivity analysis of the used method for 1000 runs.
Table II PSO parameters sensitivity analysis
Configuration

Max. vel
Min. vel
N o. of iterations
Solutions w ithout
violations
Mean fitness
Worst fitness
Best fitness
Mean tim e (s)

0.01
- 0.1
200
94 %

0.01
- 0.01
200
93 %

0.01
- 0.1
100
82 %

0.01
- 0.01
100
82%

1422.20
1482.00
1379.80
0.0716

1427.00
1534.00
1377.50
0.0715

1426.60
1.527.10
1366.70
0.0387

1433.10
1654.70
1378.50
0.0382

According to the obtained results, configuration A has been

adopted for this case study. The max position of each particles
dimension is the price variation upper limit whereas the min
position is zero. This case study uses 60 particles and 200
iterations. The maximum and the minimum velocity are 0.01 and 0.1 respectively.

Control penalties were used to address the energy balance

constraint. The penalties were added to the fitness function in
order to control the system energy balance (equality equation).
5.3 Results
The results can be analyzed in function of the execution time
and the obtained solutions. It was verified that the execution time
does not depend on the consumption reduction need. PSO takes
about 0.07 seconds and 0.15 seconds, for 32 and 320 consumers,
respectively. NLP takes about 0.32 seconds for both cases.
Although the execution time difference between NLP and
PSO may seem irrelevant, this advantage of PSO approach can
reveal important for some studies for which a large number of
simulations is required in a relatively short time. This may be the
case of decision making concerning DR events to be declared,
when these impose minimum anticipation time for the consumer
notification prior to the DR event.
Figure 2 presents the values of the Objective Function (OF)
for each approach and reduction need and for each scenario. These
values, in , correspond to the global costs for the loads after
implementation of the demand response program.
Comparing PSO and NLP, one can conclude that, for lower
reduction needs, the results are almost the same. For higher
reduction needs, PSO returns slightly worse results. These
conclusions are not influenced by the dimension of the problem
(number of consumers).
For a more detailed analysis, Figure 3 shows the energy price
variation for the 32 loads scenario and for each approach. In

Digital Object Indentifier 10.1109/MIS.2011.35

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
PEDRO FARIA ET AL.: DEMAND RESPONSE MANAGEMENT IN POWER SYSTEMS USING A PARTICLE SWARM OPTIMIZATION APPROACH

Figure 3(a), results for the major reduction need are grouped by
consumer type according to the nomenclature presented in section
4. For this reduction need, all loads are required to participate.
The slight differences between the two approaches solutions are
due to the fact that PSO is a stochastic method. As finding the
global optimum cannot be guaranteed, the obtained solution tends
to be a local optimum with an objective function value close to the
global optimum. Although the value obtained for the objective
function is close to the one obtained by the NLP approach, the
solution itself can present some differences, namely in what
concerns the consumers involved in the load reduction for each
specific DR event. In Figure 3(b), for each reduction need, the
maximum variation on energy price is compared with the
maximum allowed variation. In all results, the variations are
largely bellow the maximum permitted since the formulation

considers both price and power cap and the power cap became
prevalent, limiting higher response from loads.
Finally, Figure 4 concerns the participation of loads in the
demand response program, in the two scenarios. In general,
comparing the results for the two scenarios, one can say that the
behavior is similar. In terms of the number of loads affected, the
PSO solutions tend to correspond to a more distributed
participation of the loads. For the loads that reached the maximum
power variation, NLP makes a rational management of loads,
scheduling loads like using an order of merit: the next load is used
when the load under consideration has no more capacity. On the
contrary, PSO spreads the variation among the loads, so the
maximum variation is not reached except for higher reduction
needs when load variations are forced to the limit.

Figure 2 Values of objective function for each approach and reduction needs, for each scenario ((a) - 32 loads; (b) - 320 loads)

Figure 3 Detailed results for energy price variation, for 32 loads scenario: (a) price variation in the major reduction need for each consumer,
grouped by consumer type; (b) major energy price variation in comparison with the maximum permitted.

Digital Object Indentifier 10.1109/MIS.2011.35

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
8

Figure 4 Number of loads that have participated in DR program and the ones that reached maximum load variation, in function of reduction
need, for 32 loads scenario (a) and 320 loads scenario (b).

6 CONCLUSIONS

REFERENCES

This paper proposes a Particle Swarm Optimization (PSO)

based methodology to address price-based demand response. The
model aims at minimizing the costs of electricity for the
consumers when a consumption reduction is required by the
energy supplier.
The proposed methodology has been implemented in the
scope of a demand response simulator (DemSi) developed by the
authors, in which PSCAD is used to undertake realistic power
system simulation. The results obtained with the proposed PSO
approach are compared with those obtained using a classic Non
Linear Programming approach, leading to the conclusion that the
PSO approach presents relevant advantages, with running times
significantly lower.
The paper presents the results of two case studies, with 32 and
with 320 consumers of different types:
Domestic, Small
Commerce, Medium Commerce, Large Commerce, Industrial.
The elasticity, which characterizes each consumer, has a great
influence in the response of the loads to DR programs.
In terms of the number of loads affected by DR, PSO solution
corresponds to a participation of more loads for each imposed
consumption reduction. PSO tends not to achieve the maximum
load reduction for each consumer, spreading load reduction by
more load types. Due to these reasons, the PSO leads to more
interesting solutions, allowing achieving the envisaged
consumption needs involving a large number of participants, with
only a slightly increase of the total cost.

[1]

ACKNOWLEDGEMENTS

[11]

The authors would like to acknowledge FCT, FEDER, POCTI,

POSI, POCI, POSC, POTDC and COMPETE for their support to
R&D Projects and GECAD Unit.

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]
[10]

[12]

[13]

Digital Object Indentifier 10.1109/MIS.2011.35

Z. Vale, T. Pinto, I. Praa, H. Morais, "MASCEM - Electricity markets

simulation with strategically acting players", IEEE Intelligent Systems, vol. 26,
n. 2, Special Issue on AI in Power Systems and Energy Markets, 2011.
Y. Guan, M. Kezunovic, Grid Monitoring and Market Risk Management,
IEEE Intelligent Systems, vol. 26, n. 2, Special Issue on AI in Power Systems
and Energy Markets, 2011.
J. Ferreira, S. Ramos, Z. Vale, J. Soares, A Data-Mining-Based Methodology
for Transmission Expansion Planning, IEEE Intelligent Systems, vol. 26, n. 2,
Special Issue on AI in Power Systems and Energy Markets, 2011.
F. Rahimi, and A. Ipakchi, "Overview of Demand Response under the Smart
Grid and Market paradigms", Innovative Smart Grid Technologies (ISGT),
2010, Jan. 2010.
Z. Vale, H. Morais, and H. Khodr, Intelligent Multi-Player Smart Grid
Management Considering Distributed Energy Resources and Demand
Response, IEEE PES GM 2010, Minneapolis, July 2010.
K.Y. Lee, and M.A. El-Sharkawi, Modern heuristic optimization techniques:
theory and applications to power systems, Piscataway, N.J., IEEE Press,
Hoboken, N.J. , Wiley-Interscience, 2008.
C. Ramos, C. Liu, Intelligent Systems in Power Systems and Energy
Markets, IEEE Intelligent Systems, vol. 26, n. 2, Special Issue on AI in Power
Systems and Energy Markets, 2011.
J. Kennedy, and R.C. Eberhart, Particle Swarm Optimization, IEEE
International Conference on Neural Networks Proceedings, pp. 1942-1948, vol.
4, Dec. 1995.
E. Bonabeau, M. Dorigo, and G. Theraulaz, Swarm intelligence: From natural to
artificial systems, Oxford University Press, 1999.
M.R. AlRashidi, and M.E. El-Hawary, "A Survey of Particle Swarm
Optimization Applications in Electric Power Systems", IEEE Transactions on
Evolutionary Computation, vol. 13, pp. 913-918, Aug. 2009.
D.S. Kirschen, "Demand-side view of electricity markets", IEEE Transactions
on Power Systems, vol. 18 Issue 2, pp. 520-527, May 2003.
P.R. Thimmapuram, J. Kim, A. Botterud, and Y. Nam, "Modeling and
simulation of price elasticity of demand using an agent-based model,"
Innovative Smart Grid Technologies (ISGT), 2010, Jan. 2010.
L. Stuntz, Keeping the Lights On in a New World, tech. report, US DOE Electricity
Advisory Committee, Jan. 2009.

0885-9000/$26.00 2011 IEEE

This article has been accepted for publication in IEEE Intelligent Systems but has not yet been fully edited.
Some content may change prior to final publication.
PEDRO FARIA ET AL.: DEMAND RESPONSE MANAGEMENT IN POWER SYSTEMS USING A PARTICLE SWARM OPTIMIZATION APPROACH

[14] Z. Vale, C. Ramos, H. Morais, P. Faria, and M. Silva, The role of demand
response in future power systems, Transmission and Distribution Conf. &
Exposition: Asia and Pacific (T&D), Seoul, Korea, Oct. 2009.
[15] R.A. Arnold, Economics, Cengage Learning, South-Western College Pub, 9th
ed., 2008.
[16] M.H. Albadi, and E.F. El-Saadany, A summary of demand response in
electricity markets, Electric Power Systems Research, vol. 78 Issue 11, pp.
1989-1996, Nov. 2008.
[17] J. Bushnell, B. Hobbs, F. Wolak, When It Comes to Demand Response, Is
FERC Its Own Worst Enemy?, The Electricity Journal, vol. 22 Issue 8, pp. 9-18,
Oct. 2009.
[18] L. A. Greening, Demand response resources: Who is responsible for
implementation in a deregulated market?, Energy, vol. 35 Issue 4, Demand
Response Resources: the US and International Experience, pp. 1518-1525, April
2010.
[19] K. Hamilton, N. Gulhar, "Taking Demand Response to the Next Level", Power
and Energy Magazine, IEEE , vol.8, no.3, pp.60-65, May-June 2010.
[20] S. Chua-Liang, D. Kirschen, "Quantifying the Effect of Demand Response on
Electricity Markets" IEEE Transactions on Power Systems, vol.24, no.3,
pp.1199-1207, Aug. 2009.
[21] A. Engelbrecht, Computational Intelligence: An Introduction, John Wiley & Sons,
Ltd, 2007.
[22] M. Baran, and F. Wu, Network reconfiguration in distribution systems for
loss reduction and load balancing, IEEE Transactions on Power Delivery,
vol.4, no.2, pp.1401-1407, Apr. 1989.
[23] P. Faria, Z. Vale and J. Ferreira, DemSi A Demand Response Simulator in
the context of intensive use of Distributed Generation, Proc. 2010 IEEE
International Conference on Systems, Man, and Cybernetics (SMC 2010),
Istanbul, Oct. 2010.

AUTHORS
Pedro Faria is a graduate student in the Knowledge Engineering and Decision
Support Research Group (GECAD) of the School of Engineering at the
Polytechnic of Porto. His research interests include electricity markets, distributed
generation, and demand response. Faria has a BSc in power systems from the
Polytechnic of Porto. Contact him at pnf@isep.ipp.pt.
Zita Vale is the director of the Knowledge Engineering and Decision Support
Research Group (GECAD) and a tenured Coordinator Professor at the School of
Engineering of the Polytechnic of Porto. Her main research interests concern AI
applications to power system operation and control, electricity markets, and
distributed generation. Vale has a PhD degree in Electrical and Computer
Engineering from University of Porto, Portugal. Contact her at zav@isep.ipp.pt.
Joo Soares is a graduate student in the Knowledge Engineering and Decision
Support Research Center (GECAD) of the School of Engineering at the
Polytechnic of Porto. His research interests include heuristic optimization in power
and energy systems. Soares has a BSc in informatics from the Polytechnic of
Porto. Contact him at japs@isep.ipp.pt.
Judite Ferreira is a professor and researcher in the Knowledge Engineering and
Decision Support Research Center (GECAD) of the School of Engineering at the
Polytechnic of Porto, Portugal. Her main research interests include electricity
markets operation. Ferreira has a PhD in Electrical Engineering from the
University of Trs-os-Montes e Alto Douro. Contact her at mju@isep.ipp.pt.

Digital Object Indentifier 10.1109/MIS.2011.35

0885-9000/$26.00 2011 IEEE