www.ietdl.

org
Published in IET Generation, Transmission & Distribution
Received on 29th August 2008
Revised on 9th June 2009
doi: 10.1049/iet-gtd.2008.0424

ISSN 1751-8687

Automatic tuning method for the design

of supplementary damping controllers for
exible alternating current transmission
system devices
R.V. de Oliveira1 R. Kuiava2 R.A. Ramos2 N.G. Bretas2
1

Electromechanical Engineering Department, Federal Technological University of Parana, CEP 85503-390, Pato Branco,
PR- Brazil
2
Engineering School of Sao Carlos, University of Sao Paulo, CEP 13566-590, Sao Carlos, SP- Brazil
E-mail: [email protected]

Abstract: The design of supplementary damping controllers to mitigate the effects of electromechanical oscillations
in power systems is a highly complex and time-consuming process, which requires a signicant amount of
knowledge from the part of the designer. In this study, the authors propose an automatic technique that takes
the burden of tuning the controller parameters away from the power engineer and places it on the computer.
Unlike other approaches that do the same based on robust control theories or evolutionary computing
techniques, our proposed procedure uses an optimisation algorithm that works over a formulation of the
classical tuning problem in terms of bilinear matrix inequalities. Using this formulation, it is possible to apply
linear matrix inequality solvers to nd a solution to the tuning problem via an iterative process, with the
advantage that these solvers are widely available and have well-known convergence properties. The proposed
algorithm is applied to tune the parameters of supplementary controllers for thyristor controlled series
capacitors placed in the New England/New York benchmark test system, aiming at the improvement of the
damping factor of inter-area modes, under several different operating conditions. The results of the linear
analysis are validated by non-linear simulation and demonstrate the effectiveness of the proposed procedure.

Introduction

Inter-area oscillations are a common phenomenon observed

in power systems world-wide, where groups of synchronous
generators are interconnected over long transmission lines.
Usually, these lines create a weak electric coupling among
the generator groups and must sustain high levels of power
ow during normal operation. Consequently, these
oscillations can exhibit poor damping in the absence of an
adequate stabilising control [1].
In recent years, due to the development of power
electronics, exible alternating current transmission system
(FACTS) devices have been successfully used to improve
steady-state and dynamic system performance, and became
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

an interesting cost-effective alternative compared to the

system expansion [2, 3]. There are several types of FACTS
devices that can be used to provide additional damping to
inter-area oscillations by inclusion of a supplementary
damping controller (SDC) to the device [4].
From the practical viewpoint, there are a number of
features that any damping controller has to possess. The
following list depicts some of them:
1. the controllers must be robust with respect to the
uncertainties in the system operating point;
2. multiple damping controllers operating simultaneously in
a system must have a coordinated action;
919

& The Institution of Engineering and Technology 2009

www.ietdl.org
3. the action of the damping controllers must vanish in
steady state, so the controllers do not change the operating
point dened by the load ow;
4. a control structure based on dynamic output feedback
must be used due to difculties in obtaining measurements
of all state variables of the system;
5. whenever possible, it is preferable to use local control
input signals, because the use of remote signals generally
increases the cost of the control scheme [wide-area control
systems (WACS) have been recently proposed and are still
under development, but they are still not widely available
and satisfactorily reliable at a reasonable cost].
It is an usual practice to employ classical phase compensation
to design SDCs for FACTS devices. These controllers consist
basically of a static gain, a washout lter and a phase
compensation network [4]. Such a control structure satises
the requirements (iii)(v). However, in order to fulll the
requirements (i) and (ii), a careful selection of the phase
compensators has to be made, which is not a trivial task.
The experience accumulated over years of practice has
generated a number of guidelines for the selection of these
parameters. The rst widely used set of guidelines for this
purpose is reported in [5], and the application of them to
power system stabiliser (PSS) design is usually called a
tuning process. The advent of selective modal analysis [6],
and later on the introduction of residue analysis [7] and
induced torque coefcients [8] have signicantly improved
the efciency of the tuning methods. However, these
problems are still very difcult to handle, because they
involve several variables and degrees of freedom, and
therefore only highly skilled engineers are able to perform
an effective tuning of controller parameters for several
operating conditions.
A number of alternative techniques, based on robust control
theories, have been proposed to address these issues. H1 mixedsensitivity formulation [9] and regional pole placement using
linear matrix inequalities (LMIs) [1012] can be cited as
examples. These approaches generate very efcient
controllers, but most of them provide high-order controller
structures. The industry seems to prefer conventional phase
compensation structures, as several industry-based researchers
adopt them as their structures of choice ([13] is an example).
Based on this consideration, this paper proposes an
algorithm that aims at relieving the engineer in charge of the
controller design from the burden of the trial-and-error
process involved in this problem. The designer only has to set
up a proper range of controller parameters, based on welldened equations, and to supply this range as input data to
the algorithm, which then applies an iterative method (based
on the solution of LMIs) to nd an acceptable tuning
considering all the operating conditions of interest.
920
& The Institution of Engineering and Technology 2009

The paper is structured as follows: Section 2 depicts the

formulation of the tuning problem in terms of bilinear matrix
inequalities (BMIs) and Section 3 presents the algorithm that
tunes the controller parameters with an iterative LMI solving
process, also presenting what is required from the designer and
how the burden of manually tuning these parameters is taken
off him/her; the results of the application of the proposed
algorithm to the New England/New York benchmark system
are presented in Section 4, and Section 5 presents the conclusions.

2 Formulation of the tuning

problem in terms of BMIs
Since the objective of this paper is to tune SDCs for FACTS
devices, the following presentation will assume that the
formulation of the problem is suited to this type of design.
As will be seen, however, the formulation is general enough
to enable the simultaneous tuning of both PSSs and SDCs
for FACTS devices. The standard approach to power
system modelling for SDC tuning is based on a set of nonlinear differential-algebraic equations in the form
x_~ f (~x, u,
~ z, l)
0 g(~x, u~ , z, l)

(1)
(2)

y~ h(~x, u~ , z, l)

(3)

where x~ [ Rn is the system state vector, u~ [ Rp is the control

input vector, y~ [ Rq is the measured output, z [ Rm is a
vector of algebraic variables representing the transmission
network coupling among the state variables and l [ Rl is
a vector of parameters, representing the load levels and
other quantities dening the system operating condition.
The algebraic constraints (2) can be eliminated from (1) to
(3), and the resulting equations can be linearised around a
specic operating condition, which gives
x_ j Aj xj Bj uj

(4)

yj C j xj Dj uj

(5)

In (4) and (5), xj [ Rn represents a deviation from an

equilibrium point x~ je of (1) (3), obtained for a particular
value of the parameter vector l. In a similar way, uj [ Rp
and yj [ Rq represent deviations from u~ je and y~ je ,
respectively. To ensure the robustness of the designed
controllers with respect to variations in the parameter
vector l, the industry usually works with a set of models
(4) and (5) obtained by linearisation around several
different equilibrium points x~ je , j 1, . . . , L, each of them
corresponding to a respective value of interest for l.
As mentioned in Section 1, phase compensation is the typical
control approach used by the industry to implement both PSSs
and SDCs for FACTS devices. The block diagram of such a
structure is shown in Fig. 1, where the phase compensation
elements were lumped in the middle block, for simplicity.
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

www.ietdl.org
into a single one, given by
2

Ac1
^ ^
^
6 .
^
_
x c A c x c Bc y j 6
4 ..

Figure 1 Typical PSS or SDC block diagram

The subscript i in all controller parameters of Fig. 1 denote
that this controller is added to the ith FACTS device. The
rst block in Fig. 1 is known as a washout lter and
guarantees that the controller gain is zero under steadystate conditions. This block possesses a derivative action
represented by the term sTwi , which can be incorporated
into the plant to simplify the design formulation. Note that
this does not imply that an ideal derivative signal must be
measured in the plant, since this term can be combined
with one of the poles of the controller after the design is
carried out, and the result can be implemented as a typical
washout block, as explained in [11]. To do so, we can
dene a new system output vector y j [ Rq as
 j uj
y j y_ j C j xj D

(6)

 j C j Bj are matrices with

where C j C j Aj and D
dimensions determined by Aj , Bj and C j . We remark that
 j can be a non-zero matrix for some type of FACTS
D
devices due to the sensibility of the line active power ow
to variations in the FACTS controllable parameter.
Considering a controller with two lead-lag blocks (which
corresponds to n 2), the transfer function corresponding
to the block diagram in Fig. 1 can be put in state-space
form as follows
2

0
gi

ai
6
x_ ci 4 gi  ai bi
ai b2i  bi gi


uji 0 0

gi  bi gi

3
2 3
1
0
7
6 7
0 5xci 4 bi 5y ji (7)
g i
b2i

(8)

Ki xci

where


..
.

0 
3

3
0
.. 7
. 7
5xc
Acnc

0
.. 7
. 7
5y j
   Bcnc
2
3
0
C c1   
^ ^
6 .
^
^
.. 7
..
.
uj C c x c 6
. 7
.
4 .
5xc
0    C cnc
Bc1
6 .
6
4 ..
0


..
.

(12)

(13)

The nal closed-loop models of the controlled power system

can be obtained from the combination of the open-loop
system models given by (4) and (6) with the SDCs models
given by (12) and (13), which can be written in the form
h ^ iT
^ ^
^
_
^
x j Aj x j , where x j xj x c and
^

Aj

A j 4 ^
BC C j

Bj C C
^

 j CC
A C BC D

(14)

for j 1, . . . , L. From (14), it becomes clear that the

problem of tuning the SDC parameters consists in nding
^

a matrix triplet (Ac , Bc , C c ) ensuring that matrices Aj ,

j 1, . . . , L, fulll some desired performance criterion.
The criterion that is most widely accepted by the industry
states that the performance of the closed-loop non-linear
system can be considered as satisfactory if all eigenvalues of
^

all matrices A j , j 1, . . . , L, present a damping ratio

greater than a certain pre-specied minimum value zmin .
^

This specication denes the loci of eigenvalues of Aj for

acceptable performance as a conic sector in the complex
plane, as shown in Fig. 2, where u arccos zmin .
Considering the previous background, it is possible to see
^

1
ai
,
Twi

T
bi 1i ,
T2i

1
gi
T2i

(9)

that the problem of nding a matrix triplet (Ac , Bc , C c ) that

fullls the minimum damping ratio specication can be cast

In (7) and (8), xci [ Rnci is a vector composed by the state

variables of the ith SDC. Equations (7) and (8) can be
written in a more compact form
x_ ci Aci xci Bci y ji

(10)

uji C ci xci

(11)

Since we have one set of Equations (10) and (11) for each
SDC, we can lump all these nc sets of equations (nc being
the number FACTS in the system equipped with SDCs)
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

Figure 2 Loci of eigenvalues of A j indicating acceptable

performance of the closed-loop system
921

& The Institution of Engineering and Technology 2009

www.ietdl.org
as a search problem in Rcc  Rcp  Rqc , where
c  c, c  p and q  c are, respectively, the dimensions of
^

matrices A c , Bc and C c . Recent works solve this problem

using optimisation algorithms based on evolutionary
computing techniques (such as, for example, the genetic
algorithms used in [14]). These solutions generally achieve
good results, despite the intense computational burden
involved in these algorithms, which comes from the need
to investigate a large number of points in the search space,
in order to avoid getting stuck on local optima.
This paper opts for a different type of optimisation algorithm
that is based on Lyapunov stability theory. The idea is to nd,
^

together with the matrix triplet (A c , Bc , C c ), a set of matrices

^

P j , j 1, . . . , L, that can prove local stability of each

operating condition [15]. Based on Lyapunov theory, each
equilibrium x~ je of (1)(3) is locally stable if and only if there
^

exist matrices Ac , Bc , C c and P j such that

^T

Pj Pj  0

^ ^

^T ^

and P j A j Aj P j  0

^T

(15)

 T

^ ^
^ ^
6 sin u Aj P j P j Aj
6
6
 T
T
^ ^
^ ^
4
cos u Aj P j  P j Aj

linear in P j , j 1, . . . , L, if matrices A c and C c are xed.

The advantage of transforming (16) into a linear problem is
that the search space becomes convex, in such a way that
any available LMI solver can handle it. Based on this
consideration, it is possible to set up an algorithm to nd a
solution to (16) by iteratively solving the resulting LMIs
^

when either the P j or the Ac , Bc and C c matrices are xed,

in an alternate manner. Such an algorithm is known as
V K iteration and can be summarised as follows:
Algorithm 1 (V K iteration):
^T

Step 1: Choose a set of arbitrary matrices P j P j  0,

j 1, . . . , L, to initialise the algorithm;
^

previous step, minimise s over the entries of Ac and C c

subject to
2

 T

 T
3
^ ^
^ ^
^ ^
^ ^
u
A
P

P
A
u
A
P

P
A
cos
sin
j j
j j
j j
j j 7
6
6
7
6
 T
T
 T
 7  sI  0
^ ^
^ ^
^ ^
^ ^
4
5
cos u Aj P j  P j Aj
sin u Aj P j P j Aj
(17)
Step 3: Check whether s , 0; if yes, stop; if not, freeze the

A c , Bc , C c and P j such that P j P j  0 and

2

becomes linear in A c and C c . Conversely, the problem is

As mentioned earlier, local stabilisation is not enough to

ensure an adequate performance for the closed-loop power
system. To fulll the minimum damping ratio criterion
discussed in the previous paragraphs, we must nd
^

that if matrices P j , j 1, . . . , L, are xed, the problem

Step 2: Freeze the values of P j , j 1, . . . , L, obtained in the

for j 1, . . . , L, where the notations M  0 and N , 0

indicate positive and negative deniteness of matrices M and
N , respectively.

this in mind, and looking at (14) and (16), it is easy to see

 T
3
^ ^
^ ^
cos u A j P j  P j Aj 7
7
 T
70
^ ^
^ ^
5
sin u Aj P j P j Aj

in the previous step,

values for A c and C c obtained
^
minimise s over the entries of P j , j 1, . . . , L, subject to
^T

P j P j  0 and (17);
Step 4: Check whether s , 0; If yes, stop; if not, return to Step 2.

(16)
The proof of this statement can be traced from [15, 16] and
will not be repeated here due to space limitations. The
existence of cross products between the entries of the matrix
variables makes this search problem quite complex, because
the set of feasible solutions to it is not convex. In the next
section, an algorithm to deal with this kind of search will be
depicted and tailored to the SDC parameter tuning problem.

3 Algorithm to tune SDC

parameters
3.1 BMI solving algorithm
^

As will be seen later, the entries of matrix Bc depend only on

the maximum values of phase compensation allowed for the
SDCs and, therefore, this matrix is xed (i.e. its entries are
kept constant) during the whole tuning process. Having
922
& The Institution of Engineering and Technology 2009

It is easy to see that both Steps 2 and 3 involve solving a set of

LMIs, and that the value of s is non-increasing as the algorithm
iterates. The convergence properties of this algorithm are
discussed in [17]. Furthermore, it is clear that, once the
condition s , 0 is reached, a solution to the original BMI
was found, because (17) becomes equivalent to (16) if s , 0.
However, global convergence to a solution of (16) might not
be expected from this algorithm, as well as from the vast
majority of search methods working over non-convex feasible
sets. Since the set of feasible solutions is open, additional
restrictions must be included to restrict this set to physically
and practically meaningful solutions. The engineering part of
our proposed approach starts at this point, with the denition
of additional constraints on the set of feasible solutions to
capture the ranges of SDC parameters that would be suitable
for practical implementation. The next section describes the
process and tools for determining these constraints.
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

www.ietdl.org
3.2 Dening the ranges of acceptable
SDC parameters
As previously mentioned, it is necessary to include additional
constraints to the search problem presented in Section 3.1 to
restrict the set of possible solutions to physically and
practically meaningful ones. The phase compensation range
that provides additional damping for a particular set of
electromechanical modes can be determined by means of
residue analysis [7] or induced torque coefcients [8], for
example. In this paper, we chose to apply the former,
although similar results could be achieved with the latter.
In order to enable the application of residue analysis, we
initially set the parameters ai and bi , i 1, . . . , nc,
appearing in (7), to constant, pre-determined values.
The determination of the values for parameters ai follows
well-known industry rules, since these parameters are related
to the washout lters as can be seen in (9) and dened
according to typical values used in most SDC and PSS
designs. Common values for the washout time constant are
in the range of 3 20 s [5], which leads to

ai ai
where ai

const

const ,

0:05 , ai

const

, 0:33

(18)

is the constant value set for ai , i 1, . . . , nc.

For the determination of bi parameters, we take into account

that these parameters are related to the maximum phase
compensations wi max that each corresponding SDC can give
to its respective input signal. This relation is given by [18]
T
1 sin(wmax i )
bi 1i
T2i 1  sin(wmax i )

(19)

for i 1, . . . , nc. This maximum phase compensation can be

specied from a residue analysis of the transfer function relating
the output and input signals to the ith SDC, considering the
oscillation mode of interest lk (which is usually an inter-area
mode to be damped by this SDC).
With these xed values for ai and bi , the actual phase
compensation that will be given to the input signal for the
ith SDC will be ultimately determined by the gi parameter.
Therefore imposing limits on this parameter will result in
restricting the set of controllers in the form (7) and (8) that
can be provided by our proposed algorithm to contain only
practically meaningful ones. These limits can be written as

gi

min

, gi , gi

max

(20)

for i 1, . . . , nc, where gi min and gi max are, respectively,

upper and lower bounds on the allowed values for gi . The
upper bounds gi max are calculated by [18]

gi

max

p
vk bi

mode of interest lk . It is worth mentioning that, if gi is set to

gi max , the phase compensation provided by the ith SDC to lk
will be wi max . In the other extreme of the range, we must also
dene a minimum phase compensation wi min to be used in
the determination of the lower bound gi min , which can be
done using the relation [18]
(

1 j vi bi g1
i min
arg
1 j vi g1
i min

wi

min

(22)

or, equivalently

g2i min tan(wi

min ) gi min(vi

 vi bi ) v2i bi tan(wi

min )

(23)
The static gains Ki , i 1, . . . , nc, are then the only
remaining parameters in (7) and (8) to be constrained.
According to [19], using residue analysis it is possible to
calculate the gain Ki residue of an SDC as a function of the
desired location for the eigenvalue corresponding to lk by
Ki

residue



l
 lk 
 k des
Ri HLD (lk )

(24)

where lk des is the location that provides the desired damping

to the eigenvalue corresponding to lk and


HLD

sGwi

1 sGwi



1 sG1i
1 sG2i

n
(25)

In (25), parameters Gwi , G1i and G2i must be appropriately set

to provide the ideal phase compensation for the residue
corresponding to lk . Note that (25) is only used to dene
the bounds on parameters Ki and, therefore, Gwi , G1i and
G2i are only instrumental to the process. Considering the
previous analysis, it is possible to dene a range for lk des
that will translate into upper bounds Ki max and lower
bounds Ki min for parameters Ki , which can be written as
Ki

min

, Ki , Ki

max

(26)

for i 1, . . . , nc.
With the xed values for ai and bi and the bounded ranges
for gi and Ki given by (20) and (26), respectively, the set of
feasible solutions to the SDC tuning problem is now closed
and practically meaningful, and therefore we are ready to
present our proposed algorithm to tune SDCs for FACTS
devices, which will be done in the next section.

3.3 Proposed SDC parameter tuning

algorithm
Algorithm 2 (SDC tuning algorithm):

(21)

for i 1, . . . , nc, where v k is the frequency of the oscillation

IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

Step 1: Select an oscillation mode lk to which the SDCs have

to provide additional damping and determine the values of
923

& The Institution of Engineering and Technology 2009

www.ietdl.org
the maximum and minimum phase compensations wi
and wi min , i 1, . . ., nc, using residue analysis;

max

Step 2: Calculate the values of parameters ai , using (18), and

bi , using (19), the bounds gi max and gi min , using (21) and
(23), and the bounds Ki max and Ki min , using (24) and (25),
for i 1, . . ., nc;
Step 3: With the values of ai and bi calculated in Step 2,
^

build matrices A j and then choose a set of arbitrary positive

^

denite matrices P j , j 1, . . . , L;
^

Step 4: Freeze the values of P j , j 1, . . . , L, obtained

^ ^in the
s
over
the
entries
of
A
previous
step,
and
minimise
c , B c and
^
C c subject to (20), (26) and
2

 T

 T
3
^ ^
^ ^
^ ^
^ ^
cos u Aj P j  P j Aj 7
6 sin u Aj P j P j Aj
6
7
6
 T
T
 T
 7  sI  0
^ ^
^ ^
^ ^
^ ^
4
5
cos u Aj P j  P j Aj
sin u Aj P j P j Aj
(27)

4 Example of application of the

proposed algorithm
To demonstrate the effectiveness of our proposed procedure,
tests were carried out on the benchmark New England/
New York interconnected system shown in Fig. 3. All the
generators were described by a sixth-order model [20] and
equipped with a rst-order model of a static type AVR,
with a gain of 50 p.u./p.u. and a time constant of 0.01 s.
The transmission system was modelled as a passive circuit
and the system loads as constant impedances. Classical
PSSs based on rotor speed signals were placed at generators
G2, G3, G5, G7, G9, G10, G11, G12, G14 and G15.
These PSSs were tuned based on the residue method given
in [7], aiming purely at enhancing damping of the local
modes. These conditions are a good representation of what
happens in practice, since we have two options to increase
damping of the inter-area mode: a major retuning of all
stabilisers in the system or the placement of FACTS
devices, with their respective SDCs, in strategic points of
the system. This paper will assume that the second option
was chosen, and thyristor-controlled series capacitors
(TCSCs) were selected to perform this task.

Step 5: Check if all eigenvalues of Aj , j 1, . . . , L, have a

damping ratio greater than zmin ; If yes, go to Step 6; If not,
^

freeze the values for A c , Bc and C c obtained in the previous

^

step, and minimise s over the entries of P j , j 1, . . . , L,

^T

subject to P j P j  0, (17), (20) and (26); then, return to

Step 4;
Step 6: From the values obtained for ai , bi , gi and Ki , for
i 1, . . ., nc, calculate the values of the SDC parameters
Twi , T1i , T2i and KPSSi using (9).

Typical simplifying assumptions [3] were used to model

the TCSCs in this example. The input signal to the TCSC
is a desired reactance for the device, which compensates the
line to generate the desired power ow under steady-state
conditions. This input signal is modulated by the output of
the SDC, and the net effect of the whole TCSC dynamic
behaviour (involving measurement and processing of the
input signals to calculate the ring angle and the ring of
the thyristors at the calculated angle) over the equivalent
reactance of the device was modelled by a rst-order linear
block, as can be seen in Fig. 4. It is important to remark

Figure 3 New England/New York benchmark test system

924
& The Institution of Engineering and Technology 2009

IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

www.ietdl.org

Figure 4 TCSC dynamic model used in this paper

here that a more detailed model of the TCSCs could be used
if the study required more precision in the representation of
their dynamics. Similarly, other types of FACTS devices
(including combined series/shunt devices like the UPFC,
for example) could also be handled by our proposed
algorithm in any level of detail, the only requirement being
that a state-space model is given for them.

Figure 5 Daily load curve of the test system

We remark that, although the topology of the network was
not varied in these operating points, the proposed
algorithm can handle such type of variation as well. The
non-linear simulations at the end of this section also show
that the tuned SDCs are robust to these network topology
variations.

After a careful mode controllability/observability analysis

(via residues and participation factors), two TCSCs were
placed in the system, one of them being installed between
areas #1 and #2, in one of the lines that connect buses #60
and #61. The other was installed between areas #2 and #5,
in the line connecting buses #18 and #49. The steady-state
compensation levels of both TCSCs correspond to 50% of
the impedance of their respective lines.

The results of the eigenvalue analysis show that two interarea modes are not adequately damped in all considered
operating conditions without the inclusion of SDCs, as can
be seen in Table 1. This table shows that modes 3 and 4
are well damped, while mode 1 is poorly damped in the
operating conditions from 7 a.m. to 10 p.m. Mode 2 is
poorly damped for all points in the entire load curve.

The tests in this paper considered a daily load curve, which

is shown in Fig. 5, composed by 24 operating points, each
corresponding to a respective hour of the day. To construct
this load curve, we assumed that the data given in [20]
corresponds to a base case, which represents the system
operation at 3 p.m. in Fig. 5 (in which the load level is
approximately 62% of the maximum loading of the system).

To determine the allowable ranges for SDC parameters, a

residue analysis of the operating point at 3 p.m. was
conducted. The compensation angles required under this
condition are 2107.88 for mode 1 and 293.38 for mode
2. Deviations of +308 with respect to these angles
were used to set the ranges for phase compensation of
both SDCs, and the results were rounded up to the

Table 1 Damping ratios (%) of inter-area modes for the system without the proposed SDCs
Hour (a.m.)

Inter-area modes

Hour (p.m.)

#1

#2

#3

#4

6.02

3.66

7.86

8.50

7.40

3.75

7.80

6.30

3.74

6.21

Inter-area modes
1

12

4.73

3.10

8.76

9.87

8.34

4.66

3.08

8.79

9.87

7.76

8.39

4.63

3.06

8.82

9.93

3.68

7.77

8.15

4.51

3.00

8.91

10.01

6.27

3.61

7.46

8.33

4.24

2.90

9.05

10.13

6.05

3.12

7.71

8.56

3.84

2.72

9.29

10.31

5.43

3.38

8.33

9.23

3.58

2.62

9.44

9.53

4.97

3.21

8.61

9.65

3.73

2.68

9.31

10.30

4.44

2.50

8.16

9.50

3.99

2.79

9.20

10.25

4.40

2.98

8.94

10.04

4.33

2.94

8.99

10.10

10

4.54

3.03

8.88

9.98

10

4.77

3.12

8.73

9.83

11

4.82

3.14

8.70

9.77

11

5.27

3.33

8.42

9.45

IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

925

& The Institution of Engineering and Technology 2009

www.ietdl.org
nearest decade, resulting in 21208 , wSDC_1 , 2608 and
21408 , wSDC_2 , 2808.
After that, the maximum allowed phase compensations
(w1 max 120 and w2 max 140) were employed to
calculate the bi parameters using (19), which resulted in
b1 0:072 and b2 0:031 . The bounds on the gi variables
were calculated from (21) and (23), using the frequencies of
the inter-area modes of interest in the operating point at
3 p.m. (which are, respectively, v1 4:63 rad/s and
v2 3:53 rad/s). The ranges 0:95 , g1 , 5:86 and
0:82 , g2 , 5:22 were obtained from these calculations.
Based on (24), and again using the data corresponding to
the operating point at 3 p.m., it was possible to conclude
and
that
gains
around
KSDC 1 1:28 p.u./p.u.
KSDC 2 0:41 p.u./p.u. would be necessary to provide a
damping ratio of 5% for the two inter-area modes of
interest, under these operating conditions. The maximum
values for the gains were set to be at most twice the values
given for the operating point at 3 p.m., and the minimum
values were set to at least half these values, resulting in
0:65 , KSDC 1 , 2:6 and 0:2 , KSDC 2 , 0:8, respectively.
The time constants of the washout lters were set as 10 s
for both SDCs (Twi 10, i 1, 2), so they do not
interfere signicantly with the phase compensation.
A minimum damping ratio of 5% for all modes in all 24
operating conditions was specied as the objective of the
tuning algorithm, which was performed using the mincx
solver (available in Matlab LMI Control Toolbox) in a
laptop with an Intel Dual Core 1.6 GHz processor and
1024 MB of RAM memory. The whole design process
took about 15 h and 20 min, nishing after four iterations.

Figure 6 Inter-area modes of interest in the 5 p.m.

operating point, as the algorithm iterates

The transfer functions of the designed controllers are

presented in appendix.
The location of the closed-loop system poles associated
with the two least damped inter-area modes, as the
algorithm iterates, are presented in Fig. 6. It can be seen
that a solution is reached right after the least damped mode
crosses the boundary of the region of acceptable performance.
The damping ratios of the inter-area modes with the
inclusion of the designed SDCs, in all 24 operating
conditions, are shown in Table 2. It is possible to see in
this table that all modes present damping ratios greater
than 5%, thus fullling the design objective.

Table 2 Damping ratios (%) of inter-area modes for the system with the designed SDCs
Hour (a.m.)

Inter-area modes

Hour (p.m.)

#1

#2

#3

#4

7.43

5.81

7.96

9.35

7.83

5.89

9.31

6.50

5.84

7.69

Inter-area modes
#1

#2

#3

#4

12

6.21

5.34

8.83

10.67

9.19

6.14

5.33

8.86

10.66

7.88

9.26

6.12

5.31

8.89

10.71

5.84

7.86

9.00

6.01

5.25

8.97

10.79

7.71

5.77

7.91

9.15

5.74

5.16

9.12

10.90

7.82

5.86

7.95

9.31

5.36

5.05

9.35

11.06

6.92

5.59

8.40

10.05

5.01

5.01

9.50

11.11

6.46

5.43

8.69

10.45

5.06

5.26

9.37

11.03

5.94

5.24

9.00

10.82

5.50

5.07

9.26

11.00

5.89

5.22

9.03

10.85

5.85

5.20

9.06

10.86

10

6.03

5.27

8.95

10.76

10

6.26

5.36

8.80

10.63

11

6.27

5.36

8.86

10.69

11

6.73

5.54

8.50

10.26

926
& The Institution of Engineering and Technology 2009

IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

www.ietdl.org

Figure 7 Comparative analysis between the SDCs from [21] and the ones tuned by the algorithm proposed in this paper

Figure 8 Operating conditions at 4 p.m.; solid line: system operating with SDCs; dashed line: without SDCs
It is important to remark that the best efforts of an
experienced designed could produce similar results in about
the same time taken by our procedure to tune the SDCs.
IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

For this reason, we chose to compare the results of our

procedure with the ones given by another robust control
technique that does not employ the classical concepts of
927

& The Institution of Engineering and Technology 2009

www.ietdl.org
phase compensation for SDC tuning. The method chosen
for comparison was initially proposed for PSS design in
[10] and [11] and recently adapted to the design of SDCs
for FACTS devices in [21].
Fig. 7 presents this comparison, where it is possible to see
that, although the SDCs from [21] provide phase lead over
the entire range of possible electromechanical oscillations, they
do so at the expense of also providing large amplication of
signals in the high-frequency range. This is an important
drawback of the method proposed in [21], from the practical
viewpoint, since measurement noises and other highfrequency unmodelled dynamics will be substantially amplied
in the controller responses, deteriorating their performances.
In contrast, the SDCs tuned in this paper provide an
amount of phase compensation that is sufcient to enhance
the damping ratios of the electromechanical modes, and
their respective gains in the high-frequency range are
considerably lower than the ones of the controllers from
[21]. Given that both types of controllers fulll the desired
performance criterion and the ones tuned in this paper
have the mentioned practical advantage, the comparison is
clearly favourable to the controllers proposed in this paper.
Finally, to validate the conclusions taken from the linear
analyses and to ensure performance robustness for the
designed SDCs, non-linear simulations of the system
response to several perturbations were carried out. The
results of one of these simulations are shown in Fig. 8,
where a three-phase solid fault was applied to one of the
lines connecting buses 47 and 48, when the system
operates under the conditions corresponding to 4 p.m. in
Fig. 5. The fault was applied at t 2 s and lasted for
80 ms, being cleared by the disconnection of the faulted
line. Since the operating condition without this line was
not considered in the design stage, the good results shown
in Fig. 8 under these unpredicted conditions are an
indication that the controller performance is robust with
respect to variations in the operating point. Several other
simulations conrmed this statement.

Conclusions

time to perform other important tasks for his employer.

Furthermore, the equations dening the ranges of allowable
parameter variation for the SDCs are well-dened and
straightforward to solve, which further simplies the work
of the designer with respect to the manual tuning process.
The next step of this research is the investigation of the
effects of system model dimension over the computational
effort demanded by the algorithm. Owing to the large
number of extra variables involved in the search procedure,
this effort tends to increase exponentially if the dimension
of the model grows. An alternative to overcome this
difculty is the a priori application of model order
reduction techniques, which is a topic that is under
investigation by the authors.

Acknowledgment

The authors wish to acknowledge the support of FAPESP to

this research, under grants number 2004/04672-4 and 2006/
05191-5.

References

[1] KUNDUR P., PASERBA J., AJJARAPU V., ET AL .: Denition and

classication of power system stability, IEEE Trans. Power
Syst., 2004, 19, (3), pp. 1387 1401
[2] PASERBA J.J.: How FACTS controllers benet AC
transmission systems, IEEE Power Eng. Soc. General
Meet., 2004, 2 pp. 1257 1262
[3] HINGORANI N.G. , GYUGYI L. : Understanding FACTS:
concepts and technology of exible AC transmission
systems (IEEE Press, New York, 2000)
[4] DEL ROSSO A.D., CANIZARES C.A., DONA V.M.: A study of TCSC
controller design for power system stability improvement,
IEEE Trans. Power Syst., 2003, 18, pp. 1487 1496
[5] LARSEN E.V., SWANN D.A. : Applying power systems
stabilizers, parts i, ii, iii, AIEE Trans. Power Appar Syst.,
1981, 6, pp. 3017 3046

As shown throughout the text, an automatic tuning of SDCs

for FACTS devices can be carried out by the algorithm
proposed in this paper. Although a relatively skilled
engineer is still required to apply the software that will
perform residue analysis in order to determine the ranges of
allowable parameter variations that will be supplied to the
algorithm, the burden of manually iterating over a large
number of possible values for the SDC parameters is taken
away from the designer and placed on the computer.

[7] PAGOLA F.L., PEREZ-ARRIAGA I.J., VERGHESE G.C. : On

sensitivities, residues and participations: applications to
oscillatory stability analysis and control, IEEE Trans.
Power Syst., 1989, 4, (1), pp. 278 285

The application of this algorithm signicantly enhances

the efciency of the SDC tuning process, since the manual
tuning is quite tedious and time consuming, and
automating this process gives the engineer much more free

[8] POURBEIK P. , GIBBARD M.J., VOWLES D.J. : Proof of the

equivalence of residues and induced torque coefcients
for use in the calculation of eigenvalue shifts, IEEE Power
Eng. Rev., 2002, 22, (1), pp. 58 60

928
& The Institution of Engineering and Technology 2009

[6] PEREZ-ARRIAGA J.I., VERGHESE G.C., SCHWEPPE F.C.: Selective

modal analysis with applications to electric power
systems, parts I and II, IEEE Trans. Power Appar Syst.,
1982, 9, pp. 3117 3134

IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

www.ietdl.org
[9] CHAUDHURI B., PAL B.: Robust damping of multiple
swing modes employing global stabilizing signals
with a TCSC, IEEE Trans. Power Syst., 2004, 19, (1),
pp. 499 506
[10] RAMOS R.A., ALBERTO L.F.C., BRETAS N.G.: A new methodology
for the coordinated design of robust decentralized power
system damping controllers, IEEE Trans. Power Syst.,
2004, 19, (1), pp. 444 454
[11] RAMOS R.A., MARTINS A.C.P., BRETAS N.G.: An improved
methodology for the design of power system damping
controllers, IEEE Trans. Power Syst., 2005, 20, (4),
pp. 1938 1945
[12] XUE C.F. , ZHANG X.P., GODFREY K.R.: Design of STATCOM
damping control with multiple operating points:
a multimodel LMI approach, IEE Proc. Gener. Transm.
Distrib., 2006, 153, (4), pp. 375 382
[13] KAMWA I., GRONDIN R., TRUDEL G.: IEEE PSS2B versus PSS4B:
the limits of performance of modern power system
stabilizers, IEEE Trans. Power Syst., 2005, 20, (2),
pp. 903 915
[14] ABDEL-MAGID Y.L., ABIDO M.A., AL-BAIYAT S., MANTAWY A.H. :
Simultaneous stabilization of multimachine power
systems via genetic algorithms, IEEE Trans. Power Syst.,
1999, 14, (4), pp. 1428 1439

[16] DAVISON E.J., RAMESH N.: A note on the eigenvalues of a real

matrix, IEEE Trans. Autom. Control, 1970, 16, pp. 252 253
[17] SKELTON R.E., IWASAKI T., GRIGORIADIS K.: A unied algebraic
approach to control design (Taylor & Francis, London,
1998)
[18] OGATA K.: Modern control engineering (Prentice Hall,
New Jersey, 2002)
[19] SADIKOVIC R., KORBA P., ANDERSSON G.: Application of FACTS
devices for damping of power system oscillations, Proc.
IEEE Power Tech., 2005
[20] PAL B., CHAUDHURI B.: Robust control in power systems
(Springer Science Business Media, New York, 2005)
[21] KUIAVA R. , RAMOS R.A., BRETAS N.G.: Robust control
methodology for the design of supplementary damping
controllers for FACTS devices, Revista SBA Controle
Automa., 2009, 20, (2), pp. 192 205

Appendix

SDC for TCSC installed between areas #1 and #2

SDC1 0:200

s10:0 1 s0:0170 1 s0:0170

1 s10:0 1 s0:5484 1 s0:5484

SDC for the TCSC installed between the areas #2 and #5

[15] CHIALI M., GAHINET P., APKARIAN P.: Robust pole placement
in LMI regions, IEEE Trans. Autom. Control, 1999, 44,
(12), pp. 2257 2270

IET Gener. Transm. Distrib., 2009, Vol. 3, Iss. 10, pp. 919 929
doi: 10.1049/iet-gtd.2008.0424

SDC2 0:885

s10:0 1 s0:0350 1 s0:0350

1 s10:0 1 s0:4871 1 s0:4871

929

& The Institution of Engineering and Technology 2009