Sei sulla pagina 1di 8

1938

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

An Improved Methodology for the Design of Power


System Damping Controllers
Rodrigo A. Ramos, Member, IEEE, Andr C. P. Martins, and Newton G. Bretas, Senior Member, IEEE

AbstractThis paper presents a set of significant improvements


made over a previously developed methodology for the design
of controllers to damp electromechanical oscillations in power
systems. This previous methodology was able to fulfill several practical requirements of the oscillation damping problem, related
to robustness, decentralization, and output feedback structure.
However, problems related to an adequate controller gain (to
avoid interactions with unmodeled dynamics) and disturbance
rejection (allowing the use of noisy input signals, such as the rotor
speed measurements) were not treated in this previously reported
methodology, and are now addressed in this work. Moreover, the
requirement of zero gain in steady-state conditions is now met
in a more efficient way. The results of the nonlinear simulations
show the satisfactory performance of the designed controllers with
respect to the mentioned practical requirements.
Index TermsDamping controllers, electromechanical oscillations, linear matrix inequalities, power system stabilizers.

I. INTRODUCTION

LECTROMECHANICAL oscillations in power systems


are a problem that has been challenging engineers for
decades. These oscillations may be very poorly damped in
some cases, resulting in mechanical fatigue at the machines and
unacceptable power variations across important transmission
lines. For this reason, the use of controllers to provide better
damping for these oscillations is of utmost importance.
The most used type of controller for this purpose is known as
a power system stabilizer (PSS), which consists of a classical
phase compensator. Due to its simplicity and well-established
commissioning procedures, the PSSs have been the preferred
controllers to provide damping for electromechanical oscillations in many countries. Nevertheless, these controllers present
some drawbacks that are inherent to their design procedure. One
of these drawbacks comes from the fact that power systems experience significant variations in their operating conditions. To
deal with this problem, empirical adjustments of the PSS parameters are carried out, in a process usually known as tuning [1].
The tuning techniques have been significantly improved over
the past years, and have achieved good results in many cases
([2] is an example). However, due to their empirical nature, these
techniques have some limitations. So, motivated by the perspective of obtaining more effective controllers, several researchers

have proposed the application of robust control techniques to the


design of damping controllers for power systems ([3], [4] and
[5] can be cited as recent examples). Following the same path,
this work will be focused on the the design of robust PSS-type
controllers.
As in most other practical control problems, the task of
electromechanical oscillation damping in power systems has
a series of requirements that must be fulfilled by the designed
controllers. When attempting to apply a control technique, the
designer must be certain that the assumptions contained in the
technique are physically implementable (with a cost-compatible technology), otherwise the obtained controllers will not
reach the field. Some of these practical requirements are the
previously mentioned robustness, an output feedback decentralized structure and the guarantee of a minimum damping ratio
for all modes of oscillation. These requirements were already
successfully treated by a methodology proposed in [6].
This paper describes some important improvements made
over the methodology presented in [6], aiming to enhance
the controller performance with respect to disturbances in the
input signal and to obtain controllers with adequate gains in
the frequency range of interest. A good disturbance rejection is
norm of the
accomplished through a minimization of the
controller, allowing the use of the rotor speed as an input signal.
The adequate controller gains are sought with a reformulation
of the design procedure, which is recast as an LQR problem
(aiming to minimize the controller effort), and help to avoid the
interference of the controller with unmodeled dynamics of the
plant. Both improvements can be viewed as additional practical
requirements, not treated in [6].
The structure of this paper is as follows: Section II reviews
the fundamentals of the methodology proposed in [6], which
composes the basis of the design procedure; Section III brings a
detailed explanation of the improvements focused on this paper,
and the complete design algorithm is given in Section IV; some
tests of the design procedure, as well as an analysis of their
results, are reported in Sections V, and concluding remarks are
made in Section VI.

II. REVIEWING THE FUNDAMENTALS


Manuscript received March 7, 2005; revised June 9, 2005. Paper no. TPWRS00138-2005.
The authors are with the Escola de Engenharia de So Carlos/USPSo
Carlos, So PauloBrazil (e-mail: ramos@sel.eesc.sc.usp.br; martins@sel.eesc.sc.usp.br; ngbretas@sel.eesc.sc.usp.br).
Digital Object Identifier 10.1109/TPWRS.2005.857280

The methodology presented in [6] was developed to fulfill several practical requirements of the oscillation damping
problem in power systems. In this section, these requirements
are presented and the mathematical fundamentals of the treatment given to them in [6] are briefly depicted.

0885-8950/$20.00 2005 IEEE

RAMOS et al.: AN IMPROVED METHODOLOGY FOR THE DESIGN OF POWER SYSTEM DAMPING CONTROLLERS

1939

A. Output Feedback Control Structure and


Controller Coordination
Considering the major difficulties for implementing state
feedback damping controllers in power systems (mainly related
to the measurement of the rotor angles within the same reference frame), the output feedback structure is usually preferable.
This subsection presents a formulation for the closed-loop
system with this type of controllers.
The linearization of the power system model around an operating condition provides a set of equations in the form

Fig. 1. Minimum damping criterion for pole placement.

and controller models). These models constitute the vertices


of the polytopic set, and the methodology will then search for
,
,
and
such that
matrices

(1)
(2)
where
is the state vector,
is the control input for
the system and
is the measured output of the system.
The matrices in (1)(2) and the other matrices in appearing in
this paper have matching dimensions with their respective multiplying vectors or matrices. For more details on the nonlinear
model of the power system used in this work, see [6].
A linear dynamic output feedback controller for the system
(1)(2) may have the structure
(3)
(4)
where
is the state vector of the controller. The closedloop system formed by the feedback connection of (1)(2) and
(3)(4) can be represented by
(5)
where
(6)
is a vector containing the states of both the system
and
and the controller. The problem of stabilizing the system (1)(2)
by the output feedback controller (3)(4) can then be solved if
, ,
and
are found, such that [7]:
matrices
(7)
It is important to remark that the use of the multimachine
model directly in the design formulation provides a natural coordination for the designed controllers, which is another interesting feature of the proposed methodology.
B. Stability Robustness
In order to fulfill the robustness requirements, a technique
called polytopic modeling was applied in [6]. Some of the
propositions for the design of robust damping controllers for
power systems represent the variations in the operating conditions as uncertainties over a nominal model ([8] is an example).
In contrast, the polytopic model is composed by a set of
typical operating points (which can be taken, for example, from
,
the load curves of the system) in the form
(representing the closed-loop connections between the system

(8)
As a benefit, given by the convex structure of the polytopic
,
, and
will
set, the controller described by matrices
stabilize all the linear models contained in this set (which
may correspond to intermediate operating points of the power
system, not considered in the design).
C. Performance Robustness
The most used index to evaluate the small-signal stability of
a power system is the minimum damping ratio among all modes
of oscillation [9]. The final performance evaluation of the controllers in such systems is carried out by inspection of their effects over this minimum damping, calculated for a number of
operating conditions [10]. However, in the classical PSS design, there is no guarantee that a certain overall damping will
be achieved. If this criterion is not met, the controllers must be
redesigned and rechecked, in a trial-and-error process that incorporates some heuristics from the experience of the designer (the
tuning process). One of the great advantages of the methodology
in [6] over the classical PSS design is the possibility to include
an overall minimum damping ratio as a design objective, so the
designed controllers can automatically guarantee a satisfactory
small-signal stability index for the system (although the selection of controller parameters for this methodology could also
involve a trial-and error process as well).
In the cited methodology, the performance criteria are defined
based on the concept of D-stability [11]. According to this conis D-stable if all of its eigenvalues are concept, a matrix
tained in a convex region of the left half of the complex plane,
usually called region D. This region can be shaped in a variety
of manners, but the region shown in Fig. 1 (which contains all
, where stands for the damping ratio and
modes with
is a predefined minimum value for this ratio) is particularly
well-suited for the problem under study, for the reasons mencan be
tioned above. In this case, D-stability of the matrix
, satisfying
assured by the existence of a matrix
(9)
where

is the angle shown in Fig. 1.

D. Decentralization
The technological difficulties for implementing centralized
damping controllers in power systems are still significant, and

1940

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

therefore the decentralized structure is preferable in most cases.


Decentralization constraints can be easily handled in formulation (9) with the imposition of block diagonal structures of ap,
and . These block
propriate dimensions for matrices
diagonal structures will ensure that the controller of a particular
generator will be based only on its own input and output. However, to save space, the presentation of the approach for handling
decentralization will be based on (7) rather than (9). The extension of the presented formulation to (9) can be seen in [6].
To obtain this new formulation from (7), it is necessary to
define the following partitions:
(10)
New variables are also defined
(11)
(12)
It is possible to show [12] that, with these new variables, (7)
can be rewritten in the equivalent form
(13)
where indicates that the respective term is implicitly defined
due to the symmetry of (13). Matrix is given by
, where
is calculated a priori by
, with
and given by [6]
(14)
This approach has several advantages, both with respect to
the fulfillment of practical requirements of the problem and to
the reduction of the computational burden of the methodology
(more details can be obtained in [6]).

by reformulating the previous design procedure. Instead of only


searching for a feasible solution, each stage of the design now
performs the minimization of a suitable objective function. The
following subsections explain why these two features must be
considered as practical requirements of the oscillation damping
problem in power systems and how they were treated in this improved methodology.
A. Searching for Appropriate Gains in the Frequency Range
of Interest
The power system equations used in the analysis of low
frequency oscillations and design of damping controllers often
have a series of simplifications to reduce the complexity and
dimension of the model. Dynamical phenomena of less importance for the problem in question are often neglected. However,
care must be taken to avoid interaction of the designed controller with these unmodeled dynamics, and it is commonly
accepted that controllers with high gains have a bigger probability of interacting with these dynamics (specially with respect
to interplant modes). For this reason, acceptable gains are a
practical requirement for power system damping controllers.
A careful look at Section II-A reveals some insights about
the structure of the controller. Equations (3) and (4) show that,
and
define the dynamical part of the conwhile matrices
acts as a feedback gain, generating the controller, matrix
troller output from its states. As the system and controller states
converge to zero, these two state vectors approach each other. In
can be viewed as a kind of state feedback
this sense, matrix
matrix, and therefore designed as such, as can be seen in (14).
The well-known LQR theory provides an approach to reduce
the state feedback controller effort by minimizing a quadratic
performance index associated to the weighted square of the controller output. This approach can be used in the search of controllers with lower gains, since minimizing controller effort can
have a direct impact on its gain. The controller effort can be reduced by minimizing the performance index [7]

III. IMPROVED METHODOLOGY


In the previous section, the mathematical treatment that ensures the fulfilling of the considered practical requirements was
presented. This treatment can be summarized in inequality (9),
appropriately rewritten in the form of (14) and (13). The restrictions expressed by (9) define the set of possible triplets ( , ,
) that stabilize (5) meeting the desired performance and robustness criteria. In other words, this inequality defines the set
of all stabilizing controllers for all the considered operating conditions and satisfying a predefined minimum damping requirement.
The methodology proposed in [6] formulated the problem of
finding the triplet ( , , ) as a feasibility problem, which
implies that any controller satisfying (14) and (13) could be considered as a solution of the problem. However, finding a feasible solution for (9) does not automatically provide a physically implementable controller. Controllers with excessively
high gains or unacceptable noise rejection characteristics, for
example, may be contained in this set (and therefore may be provided by the application of the methodology in [6]). The main
purpose of this paper is to avoid these two undesirable features

(15)
Obviously, with no restrictions on this minimization, the an. However, if the restrictions given
swer to this problem is
by (9) are coupled to this minimization, the resulting procedure
then searches for the controller which will meet the desired practical requirements with the least controller effort. One additional
problem of this approach is that the minimization of (15) deof the system [13], which is highly
pends on the initial state
unpredictable. Fortunately, it is possible to find an upper bound
is known to belong to certain
for the quadratic cost (15) if
particular bounded regions, such as a polytope or an ellipsoid,
represent deviations from an
for example. Reminding that
initial operating point, it is possible to estimate a region which
will contain the maximum allowable values for these deviations.
In this work, this region is expressed in the form of an ellipsoid
, given by
(16)

RAMOS et al.: AN IMPROVED METHODOLOGY FOR THE DESIGN OF POWER SYSTEM DAMPING CONTROLLERS

If

, then an upper bound for (15) is given by


(where
stands for the
maximum eigenvalue of the matrix), provided that [13]

(17)
It can be seen that the inequalities (17) and (14) are very
is positive semi-definite,
similar. In fact, since the term
(17) implies (14). A low-effort state feedback controller for all
the initial conditions within the ellipsoid can be obtained by
the minimization of this upper bound, subject to the restriction
(17). With the application of the Schur complement formula, the
problem can be rewritten [13] as the minimization of subject
to
(18)
and
(19)
Coupling (18) and (19) to the inequalities related to the other
practical requirements (presented in Section II) and performing
the minimization, it is possible to transform the first stage of the
methodology in [6] (originally posed as a feasibility problem)
that fulfills the
into a search for a state feedback gain matrix
desired requirements (including the one related to low gain).
is then used to build matrix
,
This matrix
which will be used in the second stage of the design.
alone does not comIt is important to remark that matrix
pletely define the controller gain. This gain will be affected also
and . Howby the dynamic part of the controller, given by
is an imever, the minimization of the state feedback matrix
portant step toward gain reduction. Moreover, matrix
must
be appropriately defined to adequately shape the ellipsoid of initial conditions, which represent deviations from the initial operating point. Since these maximum deviations are hard to determine, conservative estimates may be used, but the process
can introduce a trial-and-error
of choosing the appropriate
process in the methodology.
B. Rejecting Noise in the Speed Signals
Another important practical requirement to be met by power
system damping controllers is a good disturbance rejection characteristic. This is particularly true for controllers that use rotor
speed measurements as input signals. For several reasons, the
speed signal obtained by a tachometric process has a significant
noise content that, if not appropriately treated, may severely degrade the controller performance.
The common approach to deal with this problem is the substitution of the speed signal obtained via tachometric process by
a different signal, often referred to as the integral of the accelerating power [14]. However, the simple exchange of the rotor
speed signal for the integral of accelerating power signal in [6],
to deal with the noise problems, would increase the size of the

1941

power system model, generating difficulties from the computational point of view.
control theory provides a suitable framework for the
The
norm of a
treatment of such problem. By looking at the
system, it is possible to evaluate the sensitivity of its output to a
disturbance (or class of disturbances) in its input. To apply the
theory to the methodology in [6], the closed-loop model (5)
is redefined as
(20)
(21)
In (20),
is a vector with the disturbances in the
speed signals. These disturbances are modeled as normally distributed random signals with zero mean and variance given by
, representing the noises. In (21),
is
a vector with the closed-loop system outputs, which also repreframework, it is
sent the speed signals (in a way that, in the
possible to evaluate the effect produced in the rotor speeds by
the noise in the speed measurements). With these definitions,
matrix remains in the form (6), while matrices and are
given by
(22)
With the extended closed-loop model (20)(21), it is possible
to calculate the output variance by the quadratic stochastic cost
[7]
(23)
where
by

is the observability Grammian of the system, given

(24)
The quadratic cost (23) is, in fact, equal to the
norm of
system (20)(21). However, the presentation of this definition
norm allows a better comprehension of the meaning
for the
of this norm in the context of the proposed improved methodology. The objective is to reduce the effects of noise in the speed
measurements by minimizing the variance of the system output
(in this case, the rotor speeds themselves) when the noise is
modeled in the closed-loop connection between the system and
the controller.
An upper bound for the cost (23) can be obtained by a convex
programming algorithm using the following formulation: minisubject to
mize
(25)
Applying the Schur complement formula, this problem can
subject to
also be rewritten as follows [12]: minimize
(26)

1942

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

and

Step 4: Build the computational representation of the matrix variables and (with appropriate block diagonal structures) and of the LMIs

(27)

(30)
The formulation given by (26)(27) can be used to modify the
second stage of the methodology in [6]. Again, instead of using
a feasibility approach, the procedure can be driven to search for
controllers with a satisfactory noise rejection characteristic.

(31)
(32)

C. Ensuring Zero Gain in Steady-State Conditions


If the gain of the controller (3)(4) is not equal to zero when
the system is operating in steady state, the equilibrium points of
the closed-loop system (5) may be different from those of the
open-loop system (1)(2). In order to achieve the desired zero
gain during steady state, it is sufficient to ensure the existence of
a zero at the origin of the -plane (i.e., at
) in the controller
transfer function. So, with the inclusion of a pure derivative term
in the plant (1)(2) prior to the design, this practical requirement
is also fulfilled by the methodology.
In the state-space formulation, the proposed inclusion of this
derivative term only involves the replacement of equation (2) in
the power system model by [15]
(28)
At first, the inclusion of a pure derivative term may seem to
implicate difficulties from the implementation point of view.
However, this term can be implemented in a washout-like
block, when combined to another term in the denominator of the
controller transfer function, with no additional difficulty from
the implementation viewpoint. In other words, the final controller transfer functions can be implemented in the form

(33)

for
.
Step 5: By applying an LMI solver to minimize subject
to (30)(33), find and and then calculate
.
B. Stage 2
and
, appropriately
Step 1: Choose matrices
defining the noise signal variances, and matrix
, defining the closed-loop system output (rotor
speeds in this case).
and
for
Step 2: Calculate
.
Step 3: Build the computational representation of matrix
variables , , , (with appropriate block diagonal structures) and and of the LMIs

(34)

(29)
where

is the order of

(35)

IV. CONTROLLER DESIGN ALGORITHM


When all the fundamentals of the methodology in [6] and
the proposed modifications are combined, the following design
algorithm is produced.
A. Stage 1
Step 1: Choose typical operating points and linearize the
system equations around these points, obtaining the
, and which
jacobian matrices ,
will form the open-loop vertex systems.
appropriately defining the
Step 2: Choose matrix
bounds on the allowable deviations for the initial
conditions.
Step 3: Define the minimum damping ratio required and
.
calculate

(36)
where

RAMOS et al.: AN IMPROVED METHODOLOGY FOR THE DESIGN OF POWER SYSTEM DAMPING CONTROLLERS

Fig. 2.

1943

One-line diagram of the two-area test system.

TABLE I
OPEN-LOOP MODES OF THE TWO-AREA VERTEX SYSTEMS

Fig. 3. Closed-loop modes for the two-area test system with RDCs, in various
operating conditions.

Step 4: By applying an LMI solver to minimize


ject to (34)(36), find , , and .
Step 5: Calculate
and
.
and
Step 6: Calculate

sub-

V. CONTROLLER TESTS AND EVALUATION OF RESULTS


To save space, the presented tests are made over the same
systems used in [6], under the same conditions of operation and
with the same perturbation sequences for the nonlinear simulations. Only the most significant data related to the systems
and perturbations will be given, and other details (including the
power system equations used in the design) can be obtained
from [6] and [16].
The first system to which the design procedure was applied is
described in Fig. 2. At the first stage of the design, five different
operating conditions were chosen in Step 1. These conditions
are summarized in Table I. The same data provided in [16] were
used, except for generator 3. To provide an angular reference for
the system, this generator was considered as an infinite bus. For
this reason, the system is not perfectly symmetrical and some
differences between the modes reported in [16] and in Table I are
observed. Thyristor exciters with high transient gains (shown in
[16]) were used.
was built by choosing appropriate
In Step 2, matrix
bounds for the maximum deviations allowed for each of the
state variables in all the considered operating points. These
bounds are given by
(37)
(38)
was chosen in Step
A minimum damping ratio of
3, and the other steps of Stage 1 were automatically carried out

by a computer program, which used the available solvers in the


LMI Control Toolbox for MATLAB [17].
In Step 1 of Stage 2, the noises in all generator speed
measurements were assumed to have a normal distribution
with zero mean and standard deviation of 0.0015 p.u. From
and
were built according to
this data, matrices
the following rule: When is the row associated with the
speed equation of the th generator and is the column associated with the noise signal in the -th generator, we have
.
For all other values of and ,
.
The remaining steps of Stage 2 were also automatically
performed by the computer program. For this 12-state variable
model, the whole design process, carried out on a computer
equipped with a 2.4-GHz Pentium IV processor and 1 GB of
RAM, took approximately 2 min and 30 s to be completed.
The efficiency of proposed Robust Damping Controllers (denoted by RDCs in the remaining of the text), with respect to
robustness and to the minimum damping requirements can be
seen in Fig. 3. This figure shows the oscillation modes for the
closed-loop vertex systems, as well as 12 other intermediate op,
, and
erating conditions, generated by variations of
in the base case load levels. The RDC transfer functions
are given by (39)(41), and it can be seen that, as a result of
, these conthe size chosen for the diagonal blocks of matrix
trollers have order four.
A. Controller Transfer Functions

RDC of generator 1:

RDC of generator 2:

(39)

(40)

1944

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

Fig. 4. Bode diagrams of the RDC, ORDC and PSS controllers for generator 1.

Fig. 6.

Fig. 5.

effects of the noise in the speed measurements, when the system


is controlled by the designed RDCs or by the PSSs given in
Fig. 4. It can be seen that the response of the system with RDCs
and PSSs is quite similar (as expected from the Bode plots in
Fig. 4).
The same noise signals of Fig. 5 were applied in the simulation of Fig. 6. In this simulation, a short-circuit of 32 ms was
. After that, the protection schemes
applied to bus 8 in
, the fault is eliminated
open lines 78 and 89. In
and the lines are reconnected. It is possible to see that the RDCs
provide adequate damping for the oscillation modes, even in the
presence of noise in the speed measurements.
Tests were also carried out over the New England system, as
in [6], to evaluate the performance of the methodology when applied to larger sized systems. To save space, the result of these
tests will be omitted. The whole design process for this system
was completed in approximately 27 min. It is worth remarking
that model reduction techniques (such as [18], for example) can
be used in conjunction with this improved methodology to design controllers for larger-sized systems.
Full-sized power system models would probably impose a
prohibitive computational burden on the LMI solvers, and therefore the application of this methodology is only recommended if
a reduced-order model is available. In this sense, an interesting
area of new work would be the development of LMI solvers capable of exploring sparsity of the matrix variables. Other interesting feature verified in the tests is the fact that all the RDCs
were minimum phase and internally stable controllers, although
no special treatment to obtain this feature was included in the
methodology.

Comparison of noise rejection in the test system with RDCs and PSSs.

RDC of generator 4:

(41)
The RDCs designed by this improved methodology were
compared with those presented in [6]. Fig. 4 presents a Bode
diagram for this comparison, where the controller designed
in [6] is referred to as ORDC. The frequency response of the
conventional PSS presented in [6] is also included in the figure.
It can be seen that the gain profile of the RDC is lower than the
gain for the ORDC, but is higher than the gain of the conventional PSS. However, as seen in Fig. 3, the RDC is designed to
fulfill a minimum damping requirement for several operating
conditions. On the other hand, the conventional PSS provides
this minimum damping guarantee only for the operating points
considered in the tuning process.
A series of nonlinear simulations were carried out to evaluate
the performance of the proposed RDCs with respect to noise
rejection. An equal noise signal was used for each of the generators in the system, with a normal distribution having zero mean
and 0.0015 p.u. of standard deviation. This noise signal was applied to the rotor speed measurements used as controller inputs.
Fig. 5 shows the rotor speed response (of generator 1) to the

Oscillations damped by the RDCs in off-nominal conditions.

VI. CONCLUSIONS
This paper presented an improved methodology for the design
of controllers to damp electromechanical oscillations in power
systems. Two major improvements were made over a previously
presented approach, consisting of suitable modifications in the
algorithm inequalities to allow the formulation of the controller
design in terms of an optimization problem (rather than a feasibility problem). The search for the controller is now guided by

RAMOS et al.: AN IMPROVED METHODOLOGY FOR THE DESIGN OF POWER SYSTEM DAMPING CONTROLLERS

an objective function, and it is possible to obtain a better control


over the achieved solution.
The proposed methodology has a series of advantages with
respect to the conventional approaches for PSS design. One of
the main advantages is the fact that, once a controller is found,
there is no need for a trial-and-error process (typical of PSS
tuning), because controller robustness is already guaranteed by
the quadratic stability approach (although some trial-and-error
could be involved in the process of choosing the adequate design
parameters).
The use of the polytopic model also presents some advantages over other robust control methodologies proposed in the
literature. By representing the system in a number of typical
operating conditions (rather than by uncertainties over a nominal operating condition), the polytopic model presents a more
comprehensive physical meaning when compared to other types
of uncertainty representation. A drawback of the polytopic approach is that all the system models must have the same dimension, which does not allow the representation of those conditions
where some generators go offline (although it may be possible
to represent such conditions by descriptor systems contained in
the polytope). Another shortcoming of the proposed methodology is the requirement that the controller must have the same
dimension of the plant model. Work is in progress to address the
problem of designing robust low order controllers with an output
feedback structure, but it still remains as an open problem.
An extension of the proposed procedure to allow the design
of supplementary stabilizing controllers for FACTS devices is
currently under development. Moreover, a detailed investigation
on the geometric features of the polytopic model, to allow the
construction of better suited descriptions of the power system, is
prioritized among the proposals for future research on this topic.

1945

[9] S. Gomes Jr., N. Martins, and C. Portela, Computing small-signal stability boundaries for large-scale power systems, IEEE Trans. Power
Syst., vol. 18, no. 2, pp. 747752, May 2003.
[10] M. J. Gibbard, Robust design of fixed-parameter power system stabilisers over a wide range of operating conditions, IEEE Trans. Power
Syst., vol. 6, no. 2, pp. 794800, May 1991.
[11] M. Chiali and P. Gahinet,
design with pole placement constraints:
An LMI approach, IEEE Trans. Autom. Control, vol. 41, no. 3, pp.
358367, Mar. 1996.
[12] M. C. Oliveira, J. C. Geromel, and J. Bernussou, Design of dynamic
output feedback decentralized controllers via a separation procedure,
Int. J. Control, vol. 73, no. 5, pp. 371381, 2000.
[13] S. Boyd, L. El Gahoui, E. Feron, and V. Balakrishnan, Linear Matrix
Inequalities in System and Control Theory. Philadelphia, PA: SIAM,
1994.
[14] F. P. de Mello, L. N. Hannett, and J. M. Undrill, Practical approaches to
supplementary stabilizing from accelerating power, IEEE Trans. Power
App. Syst., vol. PAS-97, no. 5, pp. 15151522, 1978.
[15] R. A. Ramos, L. F. C. Alberto, and N. G. Bretas, Decentralized output
feedback controller design for the damping of electromechanical oscillations, Int. J. Elec. Power & Energy Syst., vol. 26, pp. 207219, 2004.
[16] P. Kundur, Power System Stability and Control. New York: McGrawHill, 1994.
[17] P. Gahinet, A. Nemirovski, A. J. Laub, and M. Chiali, LMI Control
Toolbox Users Guide. Natick, MA: The Mathworks Inc., 1995.
[18] J. J. Sanchez-Gasca and J. H. Chow, Power system reduction to simplify
the design of damping controllers for interarea oscillations, IEEE Trans.
Power Syst., vol. 11, no. 3, pp. 13421349, Aug. 1996.

Rodrigo A. Ramos (S97M03) received the B.Sc.


degree in 1997, the M.Sc. degree in 1999, and the
Ph.D. degree in 2002 from Escola de Engenharia de
So CarlosUniversity of So Paulo (EESC/USP),
So Paulo, Brazil.
He is currently an Assistant Professor at
EESC/USP. His research interests are in the fields of
power system operation, small-signal stability, and
robust control.

REFERENCES
[1] E. V. Larsen and D. A. Swann, Applying power system stabilizers,
Parts I, II and III, IEEE Trans. Power App. Syst., vol. PAS-100, pp.
30173043, 1981.
[2] N. Martins, A. A. Barbosa, J. C. R. Ferraz, M. G. dos Santos, A. L. B.
Bergamo, C. S. Yung, V. R. Oliveira, and N. J. P. Macedo, Retuning
stabilizers for the North-South Brazilian interconnection, in Proc. PES
Summer Meeting, Edmonton, AB, Canada, 1999.
[3] G. E. Boukarim, S. Wang, J. Chow, G. N. Taranto, and N. Martins, A
comparison of classical, robust, and decentralized control designs for
multiple power system stabilizers, IEEE Trans. Power Syst., vol. 15,
no. 4, pp. 12871292, Nov. 2000.
[4] A. Elices, L. Rouco, H. Bourles, and T. Margotin, Design of robust controllers for damping interarea oscillations: Application to the European
power system, IEEE Trans. Power Syst., vol. 19, no. 2, pp. 10581067,
May 2004.
[5] W. Qiu, V. Vittal, and M. Khammash, Decentralized power system stabilizer design using linear parameter varying approach, IEEE Trans.
Power Syst., vol. 19, no. 4, pp. 19511960, Nov. 2004.
[6] R. A. Ramos, L. F. C. Alberto, and N. G. Bretas, A new methodology for the coordinated design of robust decentralized power system
damping controllers, IEEE Trans. Power Syst., vol. 19, no. 1, pp.
444453, Feb. 2004.
[7] K. Zhou, J. C. Doyle, and K. Glover, Robust and Optimal Control. Upper Saddle River, NJ: Prentice-Hall, 1996.
[8] B. C. Pal, A. H. Coonick, I. M. Jaimoukha, and H. El-Zobaidi, A linear
matrix inequality approach to robust damping control design in power
systems with superconducting magnetic energy storage device, IEEE
Trans. Power Syst., vol. 15, no. 1, pp. 356362, Feb. 2000.

Andr C. P. Martins received the B.Sc. degree


in 1998, the M.Sc. degree in 2000, and the Ph.D.
degree in 2005 from Escola de Engenharia de So
CarlosUniversity of So Paulo (EESC/USP), So
Paulo, Brazil.
His research interests include nonlinear systems
and power system stability.

Newton G. Bretas (M76SM89) received the


Ph.D. degree from University of Missouri, Columbia, in 1981.
He became a Full Professor with Escola de
Engenharia de So CarlosUniversity of So
Paulo (EESC/USP), So Paulo, Brazil (EESC/USP)
in 1989. His work has been primarily concerned
with power system transient stability using direct
methods, small-signal stability analysis and control,
as well as voltage collapse.

Potrebbero piacerti anche