You are on page 1of 13

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO.

3, AUGUST 2005

1465

Power Systems Stability Enhancement Using a


Wide-Area Signals Based Hierarchical Controller
Francis Okou, Member, IEEE, Louis-A. Dessaint, Senior Member, IEEE, and Ouassima Akhrif, Member, IEEE

AbstractA two-level hierarchical structure is proposed to


improve the stability of multimachine power systems. The solution
consists of a local controller for each generator at the first level
helped by a multivariable central one at the secondary level. The
secondary-level controller uses remote signals from all generators
to synthesize decoupling control signals that improve the local
controllers performances. The first-level controllers, on the
other hand, use only local signals to dampen local oscillations.
A systematic procedure for the design of the wide-area signals
based central controller is given. It is based on a reformulation of
the multimachine power system model into a suitable and closed
form. Our control strategy leads to a robust wide-area controller
having two major components: a model-based decoupling nonlinear control and an adaptation module. The nonlinear control
parameters are updated so that the controller sensitivity to power
systems operating point and structure variation is considerably
reduced. The hierarchical structure is used on a large power
system and simulation results reveal that the system stability is
considerably improved: a power system, unstable under conventional controllers (AVR-PSS/GOVERNOR), is rendered stable
when combining the central and local controllers actions.
Index TermsHierarchical systems, power system modeling,
voltage and speed regulation, wide-area control.

I. INTRODUCTION

URING the past two decades, tremendous efforts have


been made by power engineers as well as control
researchers to improve power systems stability. The standard controllers (automatic voltage regulator (AVR), power
system stabilizer (PSS), and speed governor) are single-input
single-output noncoordinated linear controllers, which are not
always able to guarantee stability when severe disturbances
or contingencies occur. The complexity of the problem comes
from the fact that power systems are highly nonlinear, large
scale, and multivariable.
Two classes of solutions are distinguishable in the literature.
The decentralized approach and the solution based on wide-area
measurements. Since the first approach is based on local measurements, additional communication facilities are not needed.
The proposed solutions are very often conservative however
since remote interactions among generators have to be dominated by local control laws. Modern linear as well as nonlinear
control methods have been used to propose well-tuned PSS [1],
[2] or novel power systems regulators [3], [4].
Manuscript received July 19, 2004; revised February 1, 2005. This work was
supported by the Transnergie Chair on Power Systems Simulation and Control.
Paper no. TPWRS-00057-2005.
The authors are with the cole de Technologie Suprieure, Montreal, QC H3C
1K3, Canada (e-mail: dessaint@ele.etsmtl.ca).
Digital Object Identifier 10.1109/TPWRS.2005.852056

The second-class approach on the other hand uses a large


amount of information from the system to be controlled thus
providing less conservative and more efficient solutions. However, additional communication facilities are needed for their
implementation. [5] proposes an optimal two-level structure for
the transient stability problem in multimachine power systems.
The solution involves a number of independent local controllers communicating with a central coordinating controller,
which optimizes a global cost function. In [6], a decentralized/hierarchical structure is proposed. Wide-area signals based
PSS is used to provide additional damping. Observability and
controllability concepts are used to select the optimal signal
to be measured and the appropriate location for the PSS,
hence reducing the complexity of the system. A sequential
optimization procedure is used to tune the proposed controller.
[7] uses multi-agent concepts to coordinate several supervisor
PSS (SPSS) based on remote signals while exchanging information with local PSS to improve power systems stability. The
SPSS uses a rule-based fuzzy-logic system and robust PSS
to deal with uncertainties introduced by nonlinear terms and
operating conditions. Despite the relative good performance
of these solutions, they remain limited since the multivariable
and nonlinear characteristics of the power system are not taken
into account explicitly.
This paper proposes a two-level hierarchical controller based
on wide-area measurements. The problems of voltage regulation
and rotor oscillations damping are addressed simultaneously.
The main characteristic of our approach is that the second-level
control action maximizes the performance of local controllers
by decoupling each generator dynamics from the rest. The controller structure and the signals needed for its implementation
are systematically obtained from the system model and the design method. The proposed secondary-level controller is robust
and combines a nonlinear decoupling control derived from the
inputoutput linearization control method with an adaptation
algorithm. The latter continuously adapts controller parameters during a change in operating conditions or structure, thus
considerably increasing the controller robustness. Furthermore,
the secondary-level contribution vanishes in steady state. Although communication delays are not taken into account explicitly during the design, the solution is tested in a realistic situation
with nonlinear loads and delays up to 400 ms.
The paper is organized as follows. Section II derives a suitable
multimachine model for the design of multivariable voltage and
speed controllers. The hierarchical controller design methodology is exposed in Section III. Simulation results from validation tests are presented in Section IV. The paper ends with a
conclusion.

0885-8950/$20.00 2005 IEEE

1466

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

an ac-to-dc converter fed from the synchronous machine terminals or auxiliary power. The static exciter type is used as power
stage in this study. It is represented by a scaling gain which is
incorporated here into the controller gains.
C. Prime Mover Model

Fig. 1. Generator components.

II. POWER SYSTEMS MODEL


Consider a power system of generators and buses interconnected through lines and transformers, and providing active
and reactive powers to static or dynamic loads. Each generator
consists of a synchronous machine with its exciter and a turbine
controlled by a servo-valve. Fig. 1 depicts the generator structure and the main signals involved.

A steam turbine with a fossil-fuelled single reheat tandem


compound is used to provide the alternator with mechanical
power. The dynamics of such a turbine are highly nonlinear
when detailed models are adopted. However, a relatively accurate (in per unit) simplified linear model is commonly used for
stability and control studies. The turbine and its input servovalve are described by the following equations:

(7)

A. Synchronous Machine Model

(8)

The well-known two-axis model in per unit is used to represent the synchronous machine. The stator dynamics and the
subtransient dynamics from two of the three damper-windings
is then represented
are neglected. Generator ,
in its local - reference frame by a fourth order model consisting of the field voltage, one damper winding, the rotor and
the power angle dynamics [8]
(1)
(2)
(3)

,
, and
are the mechanical torque, steam presand
are
sure, and steam valve position respectively.
the re-heater main inlet volumes and steam chest time constants,
is the fraction of total power generated by
respectively.
the high pressure (HP) section.
D. The Network Model
generators through
The network interconnecting the
buses is represented by two sets of algebraic equations [8]. The
first one describes the power equation at a generator terminal
bus and the second set is for a bus without generation

(4)
Equation (1) stands for the slower damper dynamics while the
field dynamics are described by (2). The rotor and the power
angle dynamics are presented by (3) and (4), respectively. Since
stator dynamics are neglected, the latter are simply described by
the following two algebraic equations:

(9.a)

(5)

(9.b)

(6)
,
,
,
,
, and
represent the current, the terminal voltage, the damper-winding flux linkage, and the field
flux voltage in the (local) - reference frame, respectively.
,
,
, and
are the direct- and quadrature-axis synchronous and transient reactances, respectively.
and
are the generator open-circuit time constants.
and
stand for the field voltage and the mechanical torque to the shaft
, and are provided by the exciter and prime mover respectively.
is the rotor speed and the power angle expressed in the absolute reference frame.
B. Exciter Model
The main exciter may be either a dc generator or an inverted
ac generator driven off the main shaft, or a static device such as

and
are the active and reactive powers of loads at bus
. Parameter
is the
entry of the network bus
is expressed in
admittance matrix. The bus voltage
the absolute reference frame. The relation
is valid at the generator terminal bus. Fig. 2
gives the relationship between the absolute frame
and
position and variables.
machine local reference frame

E. Unified Description of Network Components


The various network components are usually represented by
three types of models, namely constant impedance, static and
dynamic models. The constant impedance model is actually a
special case of the static one.

OKOU et al.: PSS ENHANCEMENT USING A WIDE-AREA SIGNALS-BASED HIERARCHICAL CONTROLLER

1467

become time-varying parameters. Dividing both active and reactive powers by the square of the load terminal voltage leads
to the following expression:
(13)

Fig. 2.

represents the instantaneous equivalent impedance of the


is constant and the
dynamic aggregate loads. In steady state,
(dynamic or static) load is equivalent to a constant impedance
one. Similarity between (11) and (13) leads to the conclusion
that network components can be described in a unique fashion
by their instantaneous equivalent impedance.
by
in (9) and diReplacing
gives the current equation at each bus
viding by

Absolute and local reference frames.

Constant impedance loads are described by the following


equation, in which active and reactive powers are directly
proportional to the square of their terminal voltages
(14.a)
(10)
and
are constant parameters. Dividing both sides of
the equality by the square of the terminal voltage leads to

(11)
is the constant load impedance. Several results in the literature on load modeling have pointed out that the dominant behavior of large power systems loads can be accurately described
by static and dynamic aggregate models [9][11]. Without loss
of generality, let us consider the KarissonHill model [9] to represent dynamic loads

(12.a)
(12.b)

(14.b)
Parameter
is the
entry of the admittance matrix,
and
including all instantaneous impedances
on the
diagonal.
is a time-varying parameter since it depends on
.
Remark 1: The idea behind the loads unified description is
that both the constant impedance and the dynamic loads can be
and
which
represented by the two quantities
can be constant or time-varying depending on the nature of the
is
load. Equation (14) with time-varying parameter
a result of combining (9) and (12). Recall also that the two-axis
model which describes the synchronous generator, the network
model and the loads dynamical model (hence the loads unified
description) are valid under the assumption that fast stator (and
network) dynamics are neglected.
Equations (1)(14) are therefore the equivalent of the traditional multimachine power systems model. This model is not
appropriate for the design of a model-based controller since it
and
) with other signals (curcombines state variables (
rents and voltages). Next, a more suitable multimachine power
system model is derived from the so-called traditional multimachine model.
F. Power System Model for the Design of Decentralized and
Hierarchical Multivariable Control

(12.c)
and
The static load case is obtained by setting
, with
for the constant
impedance case.
and
are referred to as recovery active
and reactive powers respectively [9] (or the constant power comand
, in the dynamic load case,
ponents [8]). Note that

The main idea behind the proposed model is essentially a


change of variables. The objective is to get dynamic equations
in which currents and voltages at generator-free buses are eliminated. Generators terminal voltages are used as state variables
instead of internal field voltages and damper windings flux linkages. The following steps summarize the modeling procedure.
Step 1: Express generator-free bus voltages in terms of generators terminal voltages.
Step 2: Express each generator internal voltage in terms of
its terminal voltage.

1468

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

Step 3: Express each generator stator current in terms of its


terminal voltage.
Step 4: Replace the obtained internal voltage and current
equations in the traditional model [(1)(14)].
The result is the following dynamics for generator terminal
voltage, rotor speed, and power angle
(15.a)

(17)

(15.b)
(16)
The expressions of the parameters appearing in the model are
given in the Appendix .
and
denote the angle (
)
), respectively.
and speed difference (
Remark 2: Since generators terminal voltages can experience abrupt variations (or discontinuities) particularly during
short-circuits, they can not be used a priori as state variables.
The two-axis model assumption which neglects the fast dynamics makes this possible however. Model ((15)(17)) is valid
,
) and
therefore only after the discontinuity (i.e., for
the instant of the discontinuity. Generators terminal
not at
voltages when used as state variables are usually referred to as
pseudo-state variables.
The proposed model is in a form suitable for the application
of modern model-based control methods. Decentralized power
system voltage and rotor speed regulator design becomes more
tractable since each generator terminal voltage and rotor speed
are simply expressed in terms of state variables. (Recall that the
). Also, since nonlinear perturvoltage expression is
bation terms from other generators on generator - represented
by
,
and
are (explicitly) available, a decentralized
controller guaranteeing global stability of multimachine power
,
, and
system is possible. In [12], the perturbation terms (
) are dominated by the local controllers linear terms. The
controllers fixed gains are computed from a global stability condition. Although very good transient and stability performance
may be obtained, this technique remains conservative even if additional tuning conditions based on LMI optimization are used
to reduce this conservativeness [13].
This paper proposes an alternative solution which is less conservative and is based on cancellation of inter generators interactions. The basic idea is to explicitly cancel perturbation terms
on the terminal voltage and rotor speed dynamics using suitably
computed nonlinear terms provided by control inputs (excitation and servo-valve input). Since the perturbations to be cancelled depend on remote generators state variables, remote measurement is needed to implement the solution. Such a control
strategy is called wide-area control. The controller structure and
robustness as well as the transportation delay compensation are
challenging problems that need to be dealt with in this context.
Wide-area control benefits and the technology required for
its implementation, namely phasor measurement unit are amply
reported in recent power systems control literature. The contribution of this paper is to use model-based nonlinear control
technique to systematically derive the controller expression and
specify the remote signals to be measured. The main advantage
of our solution is that it can be used in conjunction with any

OKOU et al.: PSS ENHANCEMENT USING A WIDE-AREA SIGNALS-BASED HIERARCHICAL CONTROLLER

1469

type of local control, e.g., conventional AVR-PSS and Speed


Governor in order to increase the performance of the latter.
III. WIDE-AREA SIGNALS BASED CONTROLLER
The control strategy is illustrated in Fig. 3. Our solution is
a two-level structure consisting of local controllers (based on
local measurements) for instance the conventional AVR-PSS
and Speed Governor at the primary level and a wide-area central controller (based on the system wide signals or remote signals) at the secondary one. The local controllers are dedicated
to dampen local oscillations to ensure generators stability. The
wide-area controller is used to enhance each local controller
performance by compensating the interactions among generators and inter-area oscillations. More specifically, the wide-area
controller action is to decouple the subsystems dynamics so that
a conventional AVR-PSS and Speed Governor or any modern
linear/nonlinear technique can be applied in a decoupled environment to stabilize the generators local oscillations. One of the
advantages of such a hierarchical structure is that the systems
basic operation and stability are preserved when the wide-area
controller fails.
The objective of this section is to derive wide-area control
laws for each subsystem consisting of a generator and its local
controller. Delay-free communication channels are assumed
during control design. Communication delays are taken into
account however in simulations (Section IV-C).
A. Suitable Model for Wide-Area Control Synthesis
To derive a model-based control law (for the central wide-area
controller) that vanishes when the system is in steady state, we
introduce the following change of variables:
(18)
(19)
(20)
and
are, respectively, the terminal
The constants
voltage RMS reference value and generator local - reference
frame position in the absolute reference frame. Equations (18)
and (19) are motivated by the fact that the steady state values of
,
are
and
, respectively, where
is the steady-state power angle of generator
at a given operation condition. System dynamics [(15)(17)] in
combination with prime mover ones [(7)(8)] in the new coordinates have the following form:

Fig. 3.

Hierarchical controller structure.

where

,
,
, and
are the steady-state values of the
excitation, valve-input, mechanical power, and steam chest, re,
,
, and
can be
spectively. The expressions of
derived in a straightforward manner but are omitted here for sake
of simplicity. The sequel gives the step-by-step procedure to derive the multivariable wide-area control laws for each generator.
B. Wide-Area Controller Equations
Each control input (
or
) is decomposed into a local
or
which depends only on locally measurable
term (
or
which depends on
variables) and a global term (
remote signals). That is
(27)
(28)
Recall that our objective is to compute
and
in order
to cancel interaction terms in the voltage and rotor speed dynamics. We adopt the input-output linearization approach for
the design. The method consists of first finding the dynamics of
each variable to be controlled by differentiating it until one input
appears explicitly. Secondly, the inputs expressions are selected
in order to cancel terms which depend on remote signals. To this
end, let us now introduce the variable , which represents the
terminal voltage deviation
(29)
Differentiating

gives

(21)
(22)
(23)
(30)

(24)

From this equation, we choose the global control law for generator as
(25)
(26)

if
otherwise

(31)

1470

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

The remaining dynamics become

(32)
Note that these dynamics depend exclusively on locally measurable signals
and . Also,
depends on all generators
variables hence its wide-area characteristic. A linear controller
such as the conventional AVR-PSS or a nonlinear one [4] can
be used at this level to stabilize local dynamics (32).
, we introduce the rotor speed deviFor the synthesis of
ation

Fig. 4. Example of local controller with global signals from the central
controller.

(33)
Differentiating twice
(i.e., until the control input appears
explicitly) leads to the following dynamics:

Fig. 5. Strategy to improve the hierarchical controller robustness.

(34)
Again,
in (34)

is selected in order to cancel the interaction term

(35)
The remaining dynamics depend only on local variables and are
of the form

(36)
Linear controllers such as the conventional Speed Governor or
a nonlinear counterpart can be used to stabilize local dynamics
(36). Fig. 4 shows an example of interaction between wideand
) and local AVRarea controller output signals (
PSS/Speed Governor standard regulators.
and
deRemark 3: Note that global control laws
,
and
respectively. An append on derivatives of
proximation of the differentiation operation (which filters out
very high frequencies) is very often appropriate. To this end, we
used the following transfer function:
(37)
Remark 4: When a complete ninth-order nonlinear model is
used to represent the synchronous machine and turbine, controls
(31) and (35) will only provide additional damping instead of
exact cancellation. Furthermore, since all practical limitations
on excitation and valve input have not been taken into account
during the design, a scaling gain and saturation are added to
the central controller outputs to give
(38a)
(38b)

This parameter ranges from 0 to 1 and can be considered


as a weighting factor which determines the level of contribution
of the wide-area controller signal. It is tuned by the designer
considering more problematic contingencies.
C. Controller Parameters Adaptation Algorithm
The strategy depicted at Fig. 5 is proposed to reduce the widearea controller sensitivity to the plant operating conditions. The
adaptation method adopted is motivated by the unified loads description of Section II-E which leads to the time-varying power
systems model described by (15)(17). When the systems operating point or structure changes, the parameters appearing in
,
,
and
change since they depend on network
parameters.
An estimation algorithm is therefore used to match the controller parameters to the ones of the system. This class of adaptive methods is commonly referred to as gain scheduling and is
largely used in aircraft autopilot design [14]. The adaptation law
has the following form:
(39)
where
denotes the vector of estimated parameters.
,
and
are the loads terminal voltage, active, and reactive
stands for the nonlinear expression
powers respectively.
relating estimated parameters to loads parameters.
Other types of estimation or identification methods reported
in the literature could be used as well. The gain scheduling approach has been adopted in this paper for its simplicity and its
rate of convergence. Indeed, integrations of nonlinear differential equations are not required since only algebraic computations
are performed. This considerably reduces controller complexity
parameand computational delays. Note that about
ters need to be computed for a power system with generators.
The flowchart at Fig. 6 summarizes the different steps involved
in the adaptation algorithm.

OKOU et al.: PSS ENHANCEMENT USING A WIDE-AREA SIGNALS-BASED HIERARCHICAL CONTROLLER

1471

TABLE I
LOST SIGNAL EFFECT ON CCT (IN CYCLE)

TABLE II
DELAY EFFECTS ON CCT

B. Delay-Free System and Constant Impedance Loads Case


Fig. 6. Adaptation algorithm.

Fig. 7. Two-area power system.

Since lines and transformers admittances are negligible in


comparison to loads admittances, only significant loads are involved in the adaptation process. This also helps reduce the
amount of information required by the adaptation system.
Four sets of tests are presented in the next section to thoroughly study the hierarchical structure characteristics and
possible limitations. A power system with generators represented by ninth-order dynamical equations is used for the
tests. Saturation of the magnetic circuit is also included in
the synchronous machine model. The turbine is represented
by its nonlinear model [15]. The local controllers used are
conventional AVR-PSS and Speed Governor. The stabilizer is a
well-tuned multiband-PSS [16].
IV. SIMULATION AND RESULTS
A. System Description
The power system depicted in Fig. 7 is representative of a real
system and is used to validate the proposed control structure.
It consists of three plants in area 1 and the rest of the power
system represented by a generator, in area 2. In steady operation,
area 1 provides area 2 with 5000 MW. The synchronous machines parameters are listed in Tables III and Table VI. The conhave been tuned by trial
troller parameters ,
and
and error. We use
for all the generators.

The objective of this section is to compare the performance of


the hierarchical structure versus the local-controllers-only structure. This comparison is performed for different contingencies
and for lost signals impact on the system. Both the transient responses and the critical clearing time (CTT) are evaluated. To
this end, two sets of tests are performed. The first set consists of
two different contingencies: a 6-cycle three-phase short-circuit
at bus (B1) and a much more severe 5-cycle three-phase shortcircuit at bus (B1) with L3 and L4 tripped off. Table IV gives the
constant loads characteristics. In this particular case, controller
parameters remain constant.
Figs. 810 give generator G2 terminal voltage, rotor speed
profile and inter-area oscillations after the 6-cycle short-circuit
for both hierarchical and conventional controllers. It can be
noted that the central controller provides significant additional
damping. Figs. 1114 show simulation results for the 5-cycle
short-circuit. The excitation during the second test is illustrated
in Fig. 15. Note that the hierarchical controller prevents the
systems instability. Indeed, the power system is unstable in the
case of local controllers alone (Figs. 13 and 14) but remains
stable under the wide-area controller action (Figs. 11 and 12).
Our control strategy can therefore increase power systems
critical clearing time (CCT).
The objective of the second set of tests is mainly to study
lost signals impact on the system transient performance and to
quantify the actual increase in CCT provided by the wide-area
controller. Four different configurations of the latter are used
to study lost signals impact. In the first configuration, all the
outputs are effectively used. Outputs 1 and 3 are lost for the
second configuration (i.e., outputs 2 and 4 are still in service),
only output 2 is used in the third configuration and in the fourth
configuration the wide-area controller is out of service. The fault
is a three-phase short-circuit at bus (B1). The transient behavior
of Generator G2 rotor speed is illustrated in Fig. 16 for the first
three configurations. The second test investigates the CCT for
each configuration using a three-phase short-circuit at bus (B1)
with lines (L3) and (L4) tripped off. Table I gives the results
of the second test. We observe that even in lost signals situation, the system is quite damped. Hence, the controller can be
reconfigured to optimize the number of outputs used without
strongly affecting the performance. On the other hand, the CCT
strongly depends on the central controller configuration. It is

1472

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

TABLE III
SYNCHRONOUS MACHINE PARAMETERS

Fig. 9. G2 rotor oscillations after a 6-cycle short-circuit hierarchical (solid


line) versus conventional (dash line).

TABLE IV
TRANSFORMERS, LINES, AND LOADS PARAMETERS

Fig. 10. Inter-area oscillation after a 6-cycle short-circuit hierarchical (solid


line) versus conventional (dash line).

Fig. 11. G2 voltage after a 5-cycle short-circuit with hierarchical controller.

In the next two sections, our solution is tested in a more realistic context. Delays both in the incoming and outgoing communication channels (between generators and the central controller) are now taken into account. A power system with nonlinear loads namely an induction motor and a voltage dependent
load are considered.
Fig. 8. G2 terminal voltage after a 6-cycle short-circuit hierarchical (solid line)
versus conventional (dash line).

C. Communication and Computation Delays Impact

therefore important to select the appropriate configuration for a


given contingency in order to optimize the number of outputs
used, particularly in a lost signals situation.

Constant delays are introduced during the simulation at the


wide-area controller inputs and outputs. Our objective is to comment on the effects of those delays on the transient performance
and on the CCT.

OKOU et al.: PSS ENHANCEMENT USING A WIDE-AREA SIGNALS-BASED HIERARCHICAL CONTROLLER

Fig. 12.

G2 speed after a 5-cycle short-circuit with hierarchical controller.

Fig. 13.

G2 voltage after a 5-cycle short-circuit with conventional controller.

Fig. 14.

G2 speed after a 5-cycle short-circuit with conventional controller.

A three-phase short-circuit at bus B1 (with L3 and L4 tripped


off for the CCT test) is used for this purpose. Figs. 17 and 18
compare the results of 400 ms of total delay to the ones without
delay. Table II also gives two different values of delays and the
corresponding CCT. We note that, up to 400 ms of total delay
(two ways), the hierarchical controller performs quite well and
continues to provide additional damping effects. Furthermore, it
improves the systems CCT for up to 400 ms of uniform delay
in communication channels.
Remark 4: Only the case of constant delay is considered in
this paper. It can be thought of as the worst-case scenario for the

Fig. 15.

1473

G2 excitation after a 5-cycle short-circuit with hierarchical controller.

Fig. 16. G2 rotor oscillations after a 6-cycle short-circuit conFig. 1 (solid line),
conFig. 2 (dashed line), conFig. 3 (dot-dash line).

Fig. 17. G2 terminal voltage after a 4-cycle short-circuit 0 ms delay (solid line)
400 ms delay (dashed line).

time-varying delay one. The constant delay value being the maximum value over the range of variation of the delays. The case
of nonuniform delays (i.e., different delay for each generator) is
out of the scope of this paper and left for further investigation.
D. Aggregate Dynamic Loads Case
A 20 000 MW induction machine supplying 0.9 p.u. of
mechanical power and a voltage dependent nonlinear load
of 10 000 MW and 3000 MVAR are included in load Chr3

1474

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

Fig. 18. G2 rotor speed after a 4-cycle short-circuit 0 ms delay (solid line) 400
ms delay (dashed line).

Fig. 20. G2 voltage profile in a power system with nonlinear loads hierarchical
(solid line) versus conventional (dashed line).

TABLE V
NONLINEAR LOADS PARAMETERS

Fig. 21. Resistor part of load Chr3 instantaneous impedance after a short
circuit at B3.

Fig. 19. G2 speed profile in a power system with nonlinear loads hierarchical
(solid line) versus conventional (dashed line).

connected to bus B3. The characteristics of the loads are summarized in the appendix section (Table V). Only Chr3 which is
the most significant load is involved in the adaptation process.
The main objective during the test is to assess the proposed
control strategy performance in a realistic power system with
nonlinear and dynamic loads.
Figs. 19 and 20 compare the hierarchical structure performance to the one of the local controller for a three-phase shortcircuit at the dynamic loads bus B3. Once again, the transient

Fig. 22. Controller parameter after a short-circuit at B3.


TABLE VI
OPERATING POINT (P.U.)

performance is considerably improved with wide area control.


The instantaneous impedance of the load and the controller pabehaviors are illustrated at Figs. 21 and 22, rerameter
spectively. Recall that
is
with
,
and

OKOU et al.: PSS ENHANCEMENT USING A WIDE-AREA SIGNALS-BASED HIERARCHICAL CONTROLLER

appears in
. The contingency induces a large variation of
the load impedance. These dynamics could considerably affect
the wide-area controller performance if its parameters were left
constant (i.e., without the adaptation process). Updating the parameters helps preserve the model-based hierarchical structure
performance in a realistic power system.
V. CONCLUSION
A wide-area signals-based hierarchical control structure is
proposed to improve power systems stability. Our solution is a
two-level voltage and rotor speed regulator. It consists namely
of local controllers damping local oscillations and a centralized
controller decoupling the subsystems interactions to maximize
the local controllers performance. We propose a multivariable
design approach where wide-area control laws are systematically derived from a reformulation of the multimachine model.

1475

The loads instantaneous impedances are estimated online and


used to update the central controller parameters in order to increase the hierarchical structures robustness. The remote signals needed are the generator terminal voltages, power angles
and rotor speeds. The hierarchical structurewith the classic
AVR-PSS and Speed Governor as local controllersis tested on
a realistic power system. The results confirm that the proposed
controller is robust, increases considerably the power system
stability limit and enhances the performance of local controllers.
In addition, the hierarchical structure performs well up to 400
ms of communication delays. Further investigations will address
the problem of reducing the controller sensitivity to large and
nonuniform communication delays.
APPENDIX
See the equations below. See also [17] for the definition of
, and
,

1476

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 3, AUGUST 2005

ACKNOWLEDGMENT
The authors are grateful to I. Kamwa and R. Grondin from the
Hydro-Quebec Research Institute for their helpful comments
and invaluable insight. The authors would like to thank the Editor for the unreserved help and patience, the Associate Editor
for giving them the opportunity to improve this paper to its
present quality, and the anonymous reviewers for their constructive comments.
REFERENCES
[1] S. Chen and O. P. Malik, Power system stabilizer design using  synthesis, IEEE Trans. Energy Convers., vol. 10, pp. 175181, Mar. 1995.
[2] M. Klein, L. X. Le, G. J. Rogers, S. Farrokhpay, and N. J. Balu, H
damping controller design in large power systems, IEEE Trans. Power
Syst., vol. 10, no. 1, pp. 158166, Feb. 1995.

[3] O. Akhrif, F. A. Okou, L.-A. Dessaint, and R. Champagne, Application


of a multivariable feedback linearization scheme for rotor angle stability
and voltage regulation of power systems, IEEE Trans. Power Syst., vol.
14, no. 2, pp. 620628, May 1999.
[4] F. Okou, O. Akhrif, and L.-A. Dessaint, A novel modeling approach
for decentralized voltage and speed control of multi-machine power systems, Int. J. Control, vol. 76, pp. 845857, May 2003.
[5] A. Rubaai and E. Villaseca, Transient stability hierarchical control in
multimachine power systems, IEEE Trans. Power Syst., vol. 4, no. 4,
pp. 14381444, Oct. 1989.
[6] I. Kamwa, R. Grondin, and Y. Hebert, Wide-area measurement based
stabilizing control of large power systems A decentralized/hierarchical
approach, IEEE Trans. Power Syst., vol. 16, no. 1, pp. 136153, Feb.
2001.
[7] H. Ni, T. Heydt, and L. Mili, Power system stability agents using robust wide-area control, IEEE Trans. Power Syst., vol. 17, no. 4, pp.
1123113, Nov. 2002.
[8] W. Peter, M. Sauer, and A. Pai, Power System Dynamics and Stability. Englewood Cliffs, NJ: Prentice Hall, 1998.

OKOU et al.: PSS ENHANCEMENT USING A WIDE-AREA SIGNALS-BASED HIERARCHICAL CONTROLLER

[9] D. Karisson and D. J. Hill, Modeling and identification of nonlinear


dynamic loads in power systems, IEEE Trans. Power Syst., vol. 9, no.
1, pp. 157166, Feb. 1994.
[10] D. J. Hill, Nonlinear dynamic load models with recovery for voltage
stability studies, IEEE Trans. Power Syst., vol. 8, no. 1, pp. 166176,
Feb. 1993.
[11] Standard load models for power flow and dynamic performance simulation, IEEE Trans. Power Syst., vol. 10, no. 3, pp. 13021313, Aug.
1995.
[12] F. Okou, L.-A. Dessaint, and O. Akhrif, Global voltage and speed control of large power systems, in Proc. 2003 IEEE Conf. Decision and
Control, pp. 44814486.
[13] S. Elloumi and E. B. Braiek, Robust decentralized control for multimachine power systems The LMI approach, in Proc. 2002 IEEE Int.
Conf. Systems, Man, and Cybernetics, pp. 476480.
[14] K. J. Astrom and B. Wittenmark, Adaptive Control. Reading, MA: Addison-Wesley, 1995, p. 19.
[15] P. Kundur, Power Systems Stability and Control: McGraw-Hill, 1994.
[16] I. Kamwa, G. Trudel, and L. Gerin-Lajoie, Robust design and coordination of multiple damping controllers using nonlinear constrained optimization, IEEE Trans. Power Syst., vol. 15, no. 3, pp. 10841092, Aug.
2000.
[17] A. F. Okou, O. Akhrif, and L.-A. Dessaint, Large power systems model
for decentralized and hierarchical control, in Proc. 2003 IEEE Power
Engineering Society General Meeting, pp. 17491753.

Francis Okou (M04) received the Dipl.Ing. degree in electrical engineering


from cole Suprieure Interafricaine de llectricit, Cte dIvoire, in 1993,
and the M.Ing. degree and the Ph.D. in electrical engineering from cole de
Technologie Suprieure (TS), Montreal, QC, Canada, in 1996 and 2001, respectively.
Currently, he is a Postdoctoral Fellow in the Department of Electrical Engineering at TS, where his research interests include the application of robust
and nonlinear control techniques to power systems.

1477

Louis-A. Dessaint (M88-SM91) received the B.Ing., M.Sc.A., and Ph.D. degrees from the cole Polytechnique de Montral, Montreal, QC, Canada, in
1978, 1980 and 1985, respectively, all in electrical engineering.
He was a research Assistant with the Hydro-Qubec Research Institute
(IREQ) from 1980 to 1985, where he worked on the simulation and control
of a wind energy conversion system. He is currently a Professor of Electrical
Engineering at the cole de Technologie Suprieure, Montral, QC, Canada.
From 1992 to 2001, he was the Director of the Groupe de Recherche en
lectronique de Puissance et Commande Industrielle (GREPCI), a research
group on power electronics and digital control. Since 2002, he has held the
Transnergie Chair on Power Systems Simulation and Control. He is an author
of the MathWorks SimPowerSystems simulation software.
Dr. Dessaint received the Oustanding Engineer Award from the
IEEE-Canada in 1997.

Ouassima Akhrif (M93) received the Diplme dIngnieur dtat degree from
the cole Mohammadia, Rabat, Morocco, in 1984 and the M.Sc.A. and Ph.D.
degrees from the University of Maryland, College Park, in 1987 and 1989, respectively, all in electrical engineering.
During 19891990, she was a Visiting Assistant Professor in the Systems
Engineering Department, Case Western Reserve University, Cleveland, OH. In
1992, she joined the cole de Technologie Suprieure, Montral, PQ, Canada,
where she is currently a Professor in the Electrical Engineering Department.
Her research interests are nonlinear geometric control, nonlinear adaptive control, and their applications in electric drives, power systems, and flight control
systems.

You might also like