1136 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 23, NO.

3, AUGUST 2008

Design of Wide-Area Damping Controllers for

Interarea Oscillations
Yang Zhang, Student Member, IEEE, and Anjan Bose, Fellow, IEEE

Abstract—This paper develops a systematic procedure of de- power system parameters, neglected high frequency dynamics
signing a centralized damping control system for power grid and invalid assumptions made in the modeling process. Second,
interarea oscillations putting emphasis on the signal selection local controllers lack global observation of interarea modes. It
and control system structure assignment. Geometric measures of
controllability/observability are used to select the most effective has been proved that under certain operating conditions an inter-
stabilizing signals and control locations. Line power flows and area mode may be controllable from one area and be observable
currents are found to be the most effective input signals. The syn- from another [1]. In such cases, local controllers are not effec-
thesis of the controller is defined as a problem of mixed 2 tive for the damping of that mode.
output-feedback control with regional pole placement and is The recently developed robust control theory and wide-area
resolved by the linear matrix inequality (LMI) approach. A tuning
process and nonlinear simulations are then used to modify the control system technologies offer a great potential to overcome
controller parameters to ensure the performance and robustness the shortcomings of conventional local controllers. Robust
of the controller designed with linear techniques. The design control techniques have been applied to design controllers that
process is tested on the New England 39-bus ten-machine system. formally guarantee the system stability with an acceptable
Index Terms—Damping controller, interarea oscillations, robust performance for a wide range of operating conditions [2]–[4].
control, wide-area control, wide-area measurements. With the technology of phasor measurement units (PMU), syn-
chronized dynamic data of power systems can be transferred
across the whole power system [5], [6]. The availability of
I. INTRODUCTION
wide-area measurements enables the real time detection and
ARGE power systems typically exhibit multiple domi-
L nant interarea swing modes on the order of 0.1–1.0 Hz.
This type of oscillations limits the amount of power transfer on
control of small signal instability in large scale power systems
[7]. Bonneville Power Administration (BPA) developed a
wide-area control system (WACS) for the transient stability
the tie-lines between the regions containing coherent generator and voltage support of their power system [30]. For some
groups. Today, oscillatory stability control has become more im- specific grid structures and operating conditions, wide-area
portant because these low-frequency interarea oscillations are control is more efficient than local control in preventing loss
often poorly damped with the increase of energy interchanges of synchronism. References [8] and [9] show that for a study
across the power grids. system resembling the Hydro-Québec grid, local controls need
The traditional approach to damp out interarea oscillations from four to 20 times larger gain than wide-area control to
is to install power system stabilizers (PSS) that provide supple- achieve a similar damping effect.
mentary control action through the generator excitation systems. Many researchers achieved good results by applying
In recent years, supplementary modulation controllers (SMC) wide-area measurements and robust control techniques to the
are added to FACTS devices to damp the interarea oscillations. design of wide-area control system for power system oscilla-
These controllers usually use local inputs and cannot always be tions damping. One promising approach is to design wide-area
effective in easing the problem due to two main shortcomings. measurements based controllers that provide control actions
First, based on a linearization of the system model in a nom- through generator excitation systems supplemental to the ac-
inal operating point, conventional local controllers designed by tion of local PSSs. In [10], a remote feedback controller (RFC)
the classical control techniques have their validity restricted to design methodology using PMU measurements is described.
a neighborhood of this point. But power systems constantly Reference [11] proposes a design of supervisor PSS (SPSS)
experience changes in operating conditions due to variations that exchanges information with local PSSs to improve power
in generation and load patterns and changes in transmission systems oscillatory stability. Reference [12] proposes a decen-
networks. In addition, some uncertainty is introduced into the tralized/hierarchical structure for a wide-area control system.
power system model due to inaccurate approximation of the Wide-area signals based PSSs are used to provide damping
additional to the local ones.
Manuscript received March 14, 2007; revised July 14, 2007. This work was In this paper we develop a systematic design procedure of
supported in part through the Consortium for Electric Reliability Technology wide-area damping control systems by combining stabilizing
Solutions (CERTS), funded by the Assistant Secretary of Energy Efficiency
and Renewable Energy, Office of Distributed Energy and Electricity Reliability,
signal selection and LMI based robust control design together,
Transmission Reliability Program of the U.S. Department of Energy under In- with particular attention to several issues. One such issue is the
teragency Agreement No. DE-AI-99EE35075 with the NSF and in part through selection of input stabilizing signals. The number of input sig-
PSerc, an NSF I/UCRC. Paper no. TPWRS-00171-2007. nals and the effect of different types of signals on control perfor-
The authors are with the Washington State University, Pullman, WA 99163
USA (e-mail: [email protected]; [email protected]). mance are important design considerations. For example, gen-
Digital Object Identifier 10.1109/TPWRS.2008.926718 erator speed deviations have often been used as controller in-
0885-8950/$25.00 © 2008 IEEE

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 ZHANG AND BOSE: DESIGN OF WIDE-AREA DAMPING CONTROLLERS FOR INTERAREA OSCILLATIONS 1137

puts but these are not easily obtained and need to be synchro-
nized, which often increases time delays. Our results show that
a comparison of different inputs can lead to a smaller number
of simpler inputs that provide the same control performance.
The method of geometric measures of controllability and ob-
servability introduced by Hamdan [13] is used to select the most
effective stabilizing signals and control locations.
In the literature, while single objective synthesis techniques
like optimization are often used for robust controller de-
sign [2]–[4], [11], [15], multiobjective synthesis is seldom ap-
plied in the design of wide-area damping controllers for power
system oscillations. Multiobjective synthesis has several advan-
tages over single objective synthesis as discussed in part IV. The
design method proposed in this paper for a wide-area damping
controller (WADC) to provide damping signals simultaneously
to automatic voltage regulators (AVRs) of several selected gen-
erators in addition to their local PSS control signals, is a multi-
objective synthesis, a mix of time- and frequency- domain spec-
ifications ranging from and performance to regional
Fig. 1. General structure of wide-area damping control system.
pole placement constraints.
The resulting WADC controller is a robust multiple-input-
multiple-output (MIMO) controller. Because linear techniques centralized wide-area damping control system, it still turns out
are used in designing this controller, it should be further tuned to be more cost-effective than installing new control devices.
and its performance and robustness should be tested on a In most power systems, local oscillation modes are often well
test-bed that represents the realistic power system with all its damped due to the installation of local PSSs, while interarea
nonlinearities, preferably the same type of simulation programs modes are often lightly damped because the control inputs used
that are used to study the dynamics of power systems. In our by those PSSs are local signals and often lack good observa-
design process, nonlinear simulations are conducted using the tion of some significant interarea modes. This suggests that a
Transient Security Assessment Tool (TSAT) [26] to test the wide-area controller using wide-area measurements as its inputs
efficacy of the designed WADC, which is modeled as a TSAT to create control signals supplement to local PSSs may help to
User-Defined Model (UDM). damp interarea oscillations out. A centralized controller struc-
This paper is structured as follows: Section II presents the ture is thus proposed and shown in Fig. 1.
architecture of the proposed wide-area damping control system; In the proposed wide-area damping control system, selected
Section III presents techniques used in the selection of stabi- stabilizing signals are measured by PMUs and sent to the con-
lizing signals and location of control devices; Section IV briefly troller through dedicated communication links. The wide-area
discusses the mixed output feedback control with damping controller calculates modulation signals and sends
regional pole placements in the LMI framework; Section V them to the selected generator exciters.
describes the general procedure of the design of wide-area Normally, all the local PSSs are still conventional controllers
damping controller; Section VI gives one numerical example designed by classical methods. In this design, they are modeled
and Section VII presents the conclusions. in the open loop state-space representation, on which the design
of the WADC is based. The whole damping system includes
II. WIDE-AREA DAMPING CONTROLLER STRUCTURE two levels. The first level is fully decentralized and consists of
The decentralized structure and the centralized structure are conventional PSSs. The second level is centralized and provides
the most often used approaches for the design of damping con- supplemental damping actions in addition to the first level for
troller. Based on local measurements, the first approach does not the lightly damped interarea oscillations.
need additional telecommunication equipment. But, decentral-
ized/local control alone may not be enough to meet the damping III. SELECTION OF MEASUREMENTS AND CONTROL LOCATIONS
needs of the future electrical networks, which are highly inter- Wide-area control is desirable for interarea oscillation
connected and stressed [9]. In contrast, centralized wide-area damping mainly because it provides better observability and
damping control provides more efficient solutions due to the controllability and thus better damping effects of those modes.
availability of large amount of system wide dynamic data and This benefit partly comes from the availability of system-wide
better observation of interarea modes. Wide-area controls in- dynamic information contained in remote stabilizing signals.
clude any control that requires some communication link to ei- In the selection of stabilizing signals and control locations, it
ther gather the input or to send out the control signal [14]. It is desirable to use as few measurements and control devices as
is found that if remote signals are applied to the controller, the possible to achieve satisfactory damping effects. This effort is
system dynamic performance can be enhanced with respect to worthwhile from an economic viewpoint because less measure-
interarea oscillations [1], [10]. Even though additional telecom- ment and control points mean less cost for communication links
munication equipment is needed for the realization of such a and/or controllers. Furthermore, less control loops are involved,

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 1138 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 23, NO. 3, AUGUST 2008

so that the interaction between loops can be reduced. According

to the findings of this research, only a few measurements and
control points are needed to achieve acceptable damping for
troublesome interarea modes under the proposed centralized
network configurations.
The most often used method to select locations and stabilizing
signals for PSSs and FACTS devices is controllability/observ-
ability analysis [16], [17]. This method is derived from modal
control theory of linear time-invariant system and calculates Fig. 2. Multiobjective damping controller synthesis configuration.
residue-based measures of modal controllability/observability.
The limit of residue-based measures is that they are only valid
for the signals of the same type. This approach suffers a scaling IV. MIXED OUTPUT-FEEDBACK CONTROL WITH
problem when comparing the strength of signals of a widely REGIONAL POLE PLACEMENTS VIA LMI OPTIMIZATION
differing physical significance, such as power flow in a tie-line Robust control techniques were introduced into power system
(MW), bus frequency (Hz), shaft speed (rad/s), and angle shift damping controller design in the last decade to handle mod-
(deg.) [12]. To overcome this shortcoming, the method used in eling errors and uncertainties. The most often used approach
[27], geometric measures of modal controllability/observability, is the single objective synthesis, in which all control require-
is also used in this research to evaluate the comparative strength ments are weighted and formulated in a single objective. It is
of a signal or the performance of a controller with respect to a well known that each robust method is mainly useful to capture
given mode. a set of special specifications [20]. control maintains good
After linearization around a given operating condition and robust performance in presence of model uncertainties. But it
elimination of algebraic variables, the state-space model of the is mainly concerned with frequency-domain performance and
studied system can be written as does not guarantee good transient behaviors for the closed-loop
system. control gives more suitable performance on system
(1) transient behavior and is often applied to meet performance
(2) specifications and impulsive disturbance rejection while guar-
anteeing closed-loop stability. In many practical applications,
where is the state vector and the measured output the trade-off between conflicting requirements has to be made
vector; is the input vector whose entries are control so that a single norm can represent all design requirements. In
signals sent out by WADC; , and this case, minimizing this performance index is not very effec-
are state, input and output matrices, respectively. tive because the resulting controller is often conservative and the
Suppose matrix has distinct eigenvalues achievable closed-loop performance is limited. What is more,
and the corresponding matrices of right and left the selection of weighting functions to meet the trade-off be-
eigenvectors, respectively, and . The geometric measures of tween conflicting requirements is hard and time consuming [25].
controllability and observability associated To overcome these limitations of a single objective synthesis
with the mode are technique, the multiobjective synthesis technique, which can in-
corporate various design specifications easily, is naturally con-
sidered. In our design process, the mixed output-feed-
(3) back control with regional pole placement is applied to design
a wide-area damping controller.
(4)
A. Multiobjective Damping Controller Synthesis Formulation

with the th column of input matrix (corresponding to The configuration of the multiobjective damping controller
the th input) and the th row of output matrix (corre- synthesis is shown in Fig. 2. The output channel is associ-
sponding to the th output). and are the modulus and ated with the performance and the channel is associ-
Euclidean norm of , respectively; is the geometrical ated with performance. is a low-pass filter in the
angle between the input vector and the th left eigenvector, performance channel for output disturbance rejection. is
while is the geometrical angle between the output a high-pass filter or some small constant in performance
vector and the th right eigenvector. channel that is used to reduce the control effort. is a
The joint controllability/observability measure is defined by high-pass filter in the performance channel to ensure ro-
bustness against model uncertainties.
Good design specifications ensure high-performance and easy
(5) implementation of designed controllers. In the single objective
synthesis approach, the performance is used to evaluate
In the proposed design procedure, only a few measurements all design specifications like disturbance rejection, robustness
with the highest observability of interarea modes are selected and control effort. In contrast, in this multiobjective synthesis
as stabilizing signals and only a few generators with the highest approach, the performance is only used to measure robust-
controllability of those modes are chosen as control locations. ness against dynamic uncertainty. The performance is used to

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 ZHANG AND BOSE: DESIGN OF WIDE-AREA DAMPING CONTROLLERS FOR INTERAREA OSCILLATIONS 1139

Matlab [19]. All generators are represented by detailed

models, i.e., the two-axis model with exciter, governor
and conventional PSS with two lead-lag compensation
blocks. The nonlinear model is linearized around a chosen
operating point. Then, small signal analysis is conducted
with this linear model to get the frequencies, shapes and
damping ratios of critical interarea modes.
2) Selection of measurements and control device locations:
Measurements that can be easily obtained and synchro-
nized and have the highest observability of critical inter-
area modes are good candidates for input signals. Geo-
Fig. 3. LMI region for pole placement.
metric measures of modal controllability/observability in-
troduced in Section III are used to evaluate the comparative
measure control effort and output disturbance rejection because strength of candidate signals and the performance of con-
control gives more suitable performance on system transient trollers at different locations with respect to a given inter-
behavior and control cost can be more realistically captured by area mode. The most often used input signals are voltage
the norm [18]. Since our aim is to damp out interarea oscilla- magnitudes, voltage phase angle, line power or current, fre-
tions, the center-of-inertia (COI) differences between areas are quency and generator rotor speeds.
selected as controlled output associated with performance. 3) Linear model reduction: The controller obtained by the
Pole-placement constraint is added to the multiobjective con- LMI approach is of full order, that is, the same size as the
trol problem shown in Fig. 2 to ensure good transient response design model including weighting functions. A middle size
of the closed-loop system. This problem can be formulated in system usually has several hundreds of states. To design a
the LMI framework. controller with such a high order model is neither practical
nor necessary. Therefore, model reduction is often applied
B. LMI-Based Mixed Output-Feedback Control With to obtain a lower order model for controller design. The
Regional Pole Placements reduced order model should be assured to have the same
The multiobjective optimization problem discussed global characteristics as the original system [24]. In this re-
above is solved by LMI techniques. It has been shown in [22] search, the balanced model reduction via the Schur method
that LMI provides a natural framework to formulate the mul- provided by the robust control toolbox in Matlab [19] is
tiobjective control problems without additional conservatism. used for the model reduction task.
LMI offers more flexibility for combining several constraints on 4) Controller synthesis: An LMI approach to the mixed
the closed-loop system or objectives in a numerically tractable output-feedback control with regional pole
manner. The resulting controllers do not in general suffer from placement is applied to design a wide-area damping
the problem of pole-zero cancellation [21]. Reference [20] controller for interarea oscillations. The designed con-
gives a detailed description of an LMI approach to such a troller should meet the requirements of robust stability,
complex problem of mixed output-feedback control robust performance and acceptable transient response.
with regional pole placement. Sometimes the order of the obtained controller still needs
Good transient response can be achieved by placing all to be reduced for easy implementation. In this case, the
closed-loop poles in a prescribed region in the left half plane. balanced model reduction is applied again.
Pole constraints are also useful to avoid fast dynamics and 5) Closed-loop verification and nonlinear time domain sim-
high-frequency gain in the controller, which in turn facilitate ulation: The performance of the controller is evaluated in
its digital implementation. Excessively large controller gains the closed-loop system with the full-order linear model by
should be avoided because they could lead to controller output using Matlab. The controller is then tuned and its perfor-
saturation and a poor large disturbance response of the system. mance in the actual nonlinear power system is evaluated by
To avoid large feedback gains, the system poles should not be time domain simulation using TSAT. The robustness of the
shifted too far into the left half plane. Restricting the real parts designed controller is verified for different operating con-
of the closed loop poles to be greater than a suitable negative ditions and fault scenarios.
number inhibits such excessive shifting [23]. One LMI region
for all pole placement objectives discussed above is shown in VI. RESULTS OF CASE STUDIES
Fig. 3. When the closed-loop poles are in this region, it ensures A wide-area damping controller is designed for the New Eng-
minimum damping ratio , minimum rate of decay land 39-bus ten-machine system, which is shown in Fig. 4. De-
and acceptable controller gains. tailed model descriptions and all parameters including network
data and dynamic data for the generators, excitation systems,
V. WIDE-AREA DAMPING CONTROLLER DESIGN PROCEDURE PSSs can be found in [28]. Some modifications have been made
In this research, the design of wide-area damping controllers to create a simple system modal structure so that the controller
for interarea oscillations includes the following steps: design procedure can be illustrated more clearly. The system
1) System model and small signal analysis: The full-order is stressed by increasing the load and generation level. In this
nonlinear model of the studied system is calculated using design, the conventional PSSs are tuned for local modes first

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 1140 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 23, NO. 3, AUGUST 2008

TABLE II
MAXIMUM CONTROLLABILITY AND OBSERVABILITY MEASURE
FOR 39-BUS TEN-MACHINE SYSTEM

powers and currents, generator rotor speeds and voltage

phase angle difference, are compared. Table II shows the
control locations with maximum controllability and signals
with maximum observability with respect to different inter-
area modes. in Table II, is the current magnitude of the
line connecting bus and ; is the active power of the
line connecting bus and ; is the voltage angle dif-
ference between bus and . Reference [29] shows that for
high stress conditions, current magnitude is a better input
Fig. 4. The 39-bus ten-machine test system. signal for the modulation of the parallel Pacific HVDC In-
tertie to damp Pacific AC Intertie oscillations. It is found in
this research that the active power and current of the same
TABLE I line have nearly the same observability with respect to a
CLASSIFICATION OF ELECTROMECHANICAL OSCILLATION MODES
specific interarea mode. Line active power is selected as
the input signal of the designed controller in this research.
According to Table II, G3, G7 and G8 are selected as con-
trol locations. The selection of input signals is not very ob-
vious. A large number of measurements is undesirable be-
cause this will increase possible interaction between con-
trol loops and the cost for communication links.
is selected as an input signal for the damping of mode 2.
Since has good observability for both mode 1 and
mode 3, it is also selected as an input signal. As shown
in simulations, these two signals are enough. Each tie-line
power contains information of all interarea modes in dif-
ferent levels. In fact, itself contains enough infor-
mation for all three interarea modes and could be the only
input signal for the controller at the cost of a higher gain
and a little bit worse but still acceptable damping effects.
by the method used in [1]. The wide-area controller is then de- • Model reduction: The original linear model order is 88. It
signed for the interarea modes. is reduced to a 12th-order model by the method of balanced
• Full-order model and small signal analysis: All generators model reduction.
are represented by the detailed model except generator 1, • Controller synthesis: The hinfmix function available in
which is an equivalent unit. The model is linearized around the LMI Control Toolbox of Matlab [19] was used to
a nominal operating point. Small signal analysis shows perform the necessary computations. Two controllers are
that this system has several local and interarea modes with designed. The first controller C1 has two input signals,
damping ratios less than 10%. The classification of these and . The second controller C2 has only
lightly damped modes is shown in Table I. Modes 4 to 8 one input . Both controllers provide control signals
are local ones. Even though their damping ratios are low, to generator G3, G7 and G8, as shown in Fig. 4. The
they will not last beyond 10 s because of their relatively orders of two resulting controllers are reduced to 10. The
large frequencies. It is not necessary to provide supple- frequency response of controller C1 is shown in Fig. 5.
mental damping to these modes. In each subplot, the upper curves show the amplitude
• Selection of measurements and control device locations: change and the lower curves show the phase shift, both as
The system has three coherent generator groups except functions of frequency. Weighting functions are given by
the equivalent generator G1. (G2, G3), (G8, G9, G10) and
(G4, G5, G6, G7) are nuclei of these groups correspond-
ingly. Several kinds of input signals, such as line active

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 ZHANG AND BOSE: DESIGN OF WIDE-AREA DAMPING CONTROLLERS FOR INTERAREA OSCILLATIONS 1141

Fig. 6. Rotor speed response of generator 5 to impulse disturbance.

Fig. 7. Performance improvement of controller C1 after tuning.

Fig. 5. Frequency response of controller C1. (a) Frequency response of control
loops from inputs to generator G3. (b) Frequency response of control loops from TABLE IV
inputs to generator G7. (c) Frequency response of control loops from inputs to DAMPING RATIOS AND FREQUENCIES OF INTERAREA
generator G8. MODES FOR DIFFERENT LINE OUTAGES

TABLE III
IMPROVED DAMPING OF INTERAREA MODES

nonlinear simulations. In our design process, controller pa-

rameters like gains, zeros and poles are manually modified
(tuned) from observation of the controller performance in
nonlinear simulations. Fig. 7 shows the performance im-
provement of the controller C1 after such tuning. The fault
• Closed-loop verification and nonlinear time domain simu- is a three-phase fault on bus 27 for four cycles and the
lation: The resulting reduced-order controllers are first ver- curves are the transient response of active power in line
ified by small signal analysis. Table III shows the improved 17–18.
damping of interarea modes with wide-area damping con- The eigen-analysis of the system was carried out for
trollers. Fig. 6 shows the impulse response of the rotor different operating points to verify the robustness of the
speed deviation of generator 5 without and with the con- designed controller. Table IV displays the robustness of
trollers. The impulse signal is added to the input mechan- the controller C1 in case of the outage of different heavily
ical torque of generator 5. loaded lines. Table V shows the performance of the con-
When implemented in the actual nonlinear system, the con- troller C1 for different tie-line flows between the area
troller designed with linear techniques may not have as containing generator G4, G5, G6, G7 and the rest of the
good performance and robustness as in linear simulation system. The same analysis conducted for the system with
because of the loss of some system properties in model controller C2 showed that controller C2 also improved the
linearization and model reduction. It is necessary to tune damping of the interarea modes satisfactorily but not as
the controller parameters and verify its effectiveness with well as C1.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 1142 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 23, NO. 3, AUGUST 2008

TABLE V
DAMPING RATIOS AND FREQUENCIES OF INTERAREA
MODES FOR DIFFERENT POWER FLOWS

Fig. 9. Active power of line 15–16 response to a three-phase fault on line

16–24.

placement and is resolved by the LMI approach. The final step

of the design method is the tuning and testing of this linearly
synthesized controller on full nonlinear simulations to obtain
the desired robustness and control performance. This design
method was tested on the 39-bus New England system. From
the simulation results, the following conclusions can be drawn:
• Geometric measures of controllability/observability are ef-
Fig. 8. Active power of line 16–17 response to a three-phase fault on bus 16. fective in evaluating the comparative strength of candidate
stabilizing signals of widely differing types.
• Active powers and current magnitudes on tie-lines are good
To evaluate the performance and robustness of the designed choices for stabilizing signals with respect to critical inter-
controllers in different fault scenarios, nonlinear time do- area oscillation modes.
main simulations are conducted using TSAT. Controller • For the small size system considered, one stabilizing signal
output limits are 10% of the synchronous machine ter- is enough for the input of a WADC. Multiple inputs im-
minal voltage. prove the control performance only slightly for such small
Two types of faults are simulated. The first type is a three- systems but are expected to be necessary for acceptable
phase short circuit fault applied to buses for four cycles. control performance in large systems.
Several critical buses connected with heavily loaded trans- • Mixed output-feedback control with regional
mission lines were tested. The second type is a three-phase pole placement can be applied to the wide-area damping
short circuit fault applied to transmission lines for four controller synthesis with good results which cannot be
cycles. The fault was cleared by taking out the faulted obtained by using only one of them.
line. Several critical heavily loaded transmission lines were • For a controller designed with linear techniques, a tuning
tested. The two controllers achieved satisfactory damping process is necessary to ensure its performance in the ac-
effects for all of these scenarios. Fig. 8 shows the transient tual nonlinear systems. Nonlinear simulation using a typ-
response of the active power in line 16–17 to a three-phase ical transient stability program like TSAT is required to
fault applied to bus 16. Fig. 9 shows the transient response show that such a design of a wide-area damping controller
of the active power in line 15–16 to a three-phase fault ap- is effective in a practical implementation.
plied to line 16–24. It can be seen that the damping effect The above results are promising but the results also show that
of controller C1 is a little better than controller C2. This the use of linear, continuous methods to design a controller, al-
shows the benefit brought by using more measurements, though very powerful, requires considerable tuning, testing and
which provide more system dynamic information. further development in the nonlinear, discontinuous real world.
As shown in this paper, the testing by nonlinear simulation using
VII. CONCLUSIONS a production grade transient stability program can establish the
In this paper, a systematic design procedure for wide-area robustness of the controller design, i.e., the effectiveness of the
damping control systems is described. A centralized structure is controller over a range of operating conditions. As wide-area
proposed for such systems. The comparative strength of candi- controllers have to be discontinuous by necessity, the digital
date input signals and the performance of output control signals sampling rate and possible communication latency are variables
at different locations with respect to interarea modes are evalu- that affect performance. Both the design process and the testing
ated by geometric measures of controllability/observability. The should take these into account. Research is still needed to do so
synthesis of the robust MIMO controller is defined as a problem as the present day production grade simulations like TSAT are
of mixed output-feedback control with regional pole not capable of representing such communication variables. In

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.
 ZHANG AND BOSE: DESIGN OF WIDE-AREA DAMPING CONTROLLERS FOR INTERAREA OSCILLATIONS 1143

our future work we intend to do such development of the design [16] N. Martins and L. T. G. Lima, “Determination of suitable locations for
capabilities as well as move the testing to a simulation of larger power system stabilizers and Static VAR Compensators for damping
electromechanical oscillations in large scale power systems,” IEEE
systems such as the WECC system. These types of design and Trans. Power Syst., vol. 5, no. 4, pp. 1455–1469, Nov. 1990.
testing tools need to be developed before such controllers be- [17] P. Zhang, A. R. Messina, A. Coonick, and B. J. Cory, “Selection of lo-
come robust enough for implementation in real systems. cations and input signals for multiple SVC damping controllers in large
scale power systems,” in Proc. IEEE Power Eng. Soc. Winter Meeting,
1998, pp. 667–670.
REFERENCES [18] B. C. Pal, A. H. Coonick, and B. J. Cory, “Linear matrix inequality
[1] Aboul-Ela, A. Sallam, J. McCalley, and A. Fouad, “Damping con- versus Root-locus approach for damping inter-area oscillations in
troller design for power system oscillations using global signals,” IEEE power systems,” Elect. Power Energy Syst., vol. 23, no. 6, p. 48149,
Trans. Power Syst., vol. 11, no. 2, pp. 767–773, May 1996. Aug. 2001.
[2] R. A. Ramos, L. F. C. Alberto, and N. G. Bretas, “A new methodology [19] “Matlab Users Guide,” MathWorks, Natick, MA, 1998.
for the coordinated design of robust decentralized power system [20] C. Scherer, P. Gahinet, and M. Chilali, “Multiobjective output-feed-
damping controllers,” IEEE Trans. Power Syst., vol. 19, no. 1, pp. back control via LMI optimization,” IEEE Trans. Autom. Control, vol.
444–454, Feb. 2004. 42, no. 7, pp. 896–911, Jul. 1997.

H1 H1
[3] Z. Chuanjiang, M. Khammash, V. Vittal, and Q. Wenzheng, “Robust [21] P. Gahinet and P. Apkarian, “A linear matrix inequality approach to
power system stabilizer design using loop shaping approach,” control,” Int. J. Robust Nonlin. Control, vol. 4, pp. 421–448, 1994.
IEEE Trans. Power Syst., vol. 18, no. 2, pp. 810–818, May 2003. [22] S. Boyd, L. EI Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix
[4] G. E. Boukarim, S. Wang, J. H. Chow, G. N. Taranto, and N. Martins, Inequalities in System and Control Theory. Philadelphia, PA: SIAM,
“A comparison of classical, robust, and decentralized control designs Jun. 1994, vol. 15, Studies in Applied Mathematics.
for multiple power system stabilizers,” IEEE Trans. Power Syst., vol. [23] P. Shrikant Rao and I. Sen, “Robust pole placement stabilizer design
15, no. 4, pp. 1287–1292, Nov. 2000. using linear matrix inequalities,” IEEE Trans. Power Syst., vol. 15, no.
[5] W. A. Mittelstadt et al., “The DOE wide area measurement system 1, pp. 313–319, Feb. 2000.
(WAMS) project—Demonstration of dynamic information technology [24] J. J. Sinchez-Gasca and J. H. Chow, “Power system reduction to
for future power system,” in Proc. Fault and Disturbance Analysis & simplify the design of damping controllers for inter-area oscillations,”
Precise Measurements in Power Systems, Arlington, VA, Nov. 1995. IEEE Trans. Power Syst., vol. 11, no. 3, pp. 1342–1349, Aug. 1996.
[6] A. G. Phadke, “Synchronized phasor measurements—A historical [25] K. Zhou, J. C. Doyle, and K. Glover, Robust and Optimal Control.
overview,” in Proc. IEEE/Power Eng. Soc. Transmission and Distribu- Upper Saddle River, NJ: Prentice-Hall, 1996.
tion Conf. Exhib., 2002, pp. 416–479. [26] “TSAT Version 4.1 User’s manual,” PowerTech Labs, Sep. 2004.
[7] J. Quintero and V. Venkatasubramanian, “A real-time wide-area con- [27] A. Heniche and I. Kamwa, “Control loops selection to damp inter-area
trol framework for mitigating small-signal instability in large electric oscillations of electrical networks,” IEEE Trans. Power Syst., vol. 17,
power systems,” in Proc. 38th Annu. Hawaii Int. Conf. System Sciences, no. 2, pp. 378–384, May 2002.
Jan. 3–6, 2005. [28] Extended transient-midterm stability package, Jan. 1987, EPRI
[8] I. Kamwa, J. Béland, G. Trudel, R. Grondin, C. Lafond, and D. EL-4610, Project 1208, Fin. Rep.
McNabb, “Wide-area monitoring and control at Hydro-Québec: Past, [29] R. L. Cresap, D. N. Scott, W. A. Mittelstadt, and C. W. Taylor,
present and future,” in Proc. IEEE Power Eng. Soc. Summer Meeting, Damping of Pacific AC Intertie Oscillations Via Modulation of the
Jun. 18–22, 2006, pp. 1–12. Parallel Pacific HVDC Intertie, CIGRE 14-05, 1978.
[9] I. Kamwa, A. Heniche, G. Trudel, M. Dobrescu, R. Grondin, and D. [30] C. W. Taylor et al., “WACS-wide-area stability and voltage control
Lefebvre, “Assessing the technical value of FACTS-based wide-area system: R&D and online demonstration,” Proc. IEEE, vol. 93, no. 5,
damping control loops,” in Proc. IEEE Power Eng. Soc. General pp. 892–906, May 2005.
Meeting, Jun. 12–16, 2005, vol. 2, pp. 1734–1743.
[10] A. F. Snyder et al., “Delay-input wide-area stability control with
synchronized phasor measurements,” in Proc. IEEE Power Eng. Soc.
Summer Meeting, Jul. 16–20, 2000, vol. 2, pp. 1009–1014.
[11] H. Ni, G. T. Heydt, and L. Mili, “Power system stability agents using Yang Zhang (S’05) received the B.Eng. degree from North China Electric
robust wide area control,” IEEE Trans. Power Syst., vol. 17, no. 4, pp. Power University, Baoging, China, in 1997 and the M.Eng. degree from China
1123–1131, Nov. 2002. Electric Power Research Institute (CEPRI), Beijing, China, in 2001. He is now
[12] I. Kamwa, R. Grondin, and Y. Hebert, “Wide-area measurement based pursuing the Ph.D. degree at Washington State University, Pullman, WA.
stabilizing control of large power systems—A decentralized/hierar- He worked for CEPRI from 2001 to 2004. His special fields of interest in-
chical approach,” IEEE Trans. Power Syst., vol. 16, no. 1, pp. 136–153, cluded power system stability and real time control.
Feb. 2001.
[13] A. M. A. Hamdan and A. M. Elabdalla, “Geometric measures of
modal controllability and observability of power systems models,”
Elect. Power Syst. Res., vol. 15, pp. 147–155, 1988. Anjan Bose (F’89) received the B.Tech. (Honors) degree from the Indian In-
[14] K. Tomsovic, D. E. Bakken, V. Venkatasubramanian, and A. Bose, stitute of Technology, Kharagpur, the M.S. degree from the University of Cali-
“Designing the next generation of real-time control, communication, fornia, Berkeley, and the Ph.D. from Iowa State University, Ames.
and computations for large power systems,” Proc. IEEE, vol. 93, no. 5, He has worked for industry, academe, and government for 40 years in power
pp. 965–979, May 2005. system planning, operation, and control. He is currently Regents Professor and
[15] B. Chaudhuri and B. C. Pal, “Robust damping of inter-area oscilla- holds the endowed Distinguished Professor in Power Engineering at Washington
tions through controllable phase shifters using global signals,” in Proc. State University, Pullman, WA.
IEEE Power Eng. Soc. General Meeting, Jul. 13–17, 2003, vol. 3, pp. Dr. Bose is a member of the National Academy of Engineering and the recip-
1673–1679. ient of the Herman Halperin Award and the Millenium Medal from the IEEE.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on August 05,2010 at 17:40:06 UTC from IEEE Xplore. Restrictions apply.