Vous êtes sur la page 1sur 7

IEEE Transaction on Power Apparatus and Systems, Vol. PAS-103, No.

8, August 1984 1983


POWER SYSTEM STABILIZER BASED ON ADAPTIVE
CONTROL TECHNIQUES

A. Ghosh G. Ledwich O. P. Malik G. S . Hope


University of Calgary University of Queensland University of Calgary
Calgary, Canada Brisbane, Australia Calgary, Canada

Abstract:
adaptive stabilizer output is used to supplement the
Characteristics of certain adaptive control normal voltage regulator output. Simulation studies
techniques are dis'cussed in this paper with particular have been. performed with the adaptive'stabilizer based
emphasis on their application to power systems. A on two alternate, adaptive control algorithms and a
modified form of one of the more promising algorithms conventional fixed parameter stabilizer. Comparative
that makes it more suitable. for power systems is results of these studies are given.
developed and described. Comparative results of
studies with adaptive stabilizer based on two 2. Algorithms With Implicit Identification:
alternate adaptive control algorithms and a
conventional fixed parameter stabilizer show the Concept: A very useful adaptive control technique is
improvement in response obtained with the adaptive to specify a desired performance and measure the
algorithm. actual performance against this performance. This
type of controller, commonly known as a 'Model
1. Introduction: Reference Adaptive Controller' (MRAC) follows a model
designed for the desired performance. A reference
Adaptive control can be described as the changing model representing the desired.behaviour of the closed
of controller parameters based on the changes in loop system is driven by the same input as the
system operating conditions. Advantages of on-line controlled system. The regulator parameters are then
changes of controller parameters have been adjusted depending upon the error, between the system
demonstrated for various processes [1,2]. The output and the reference model output. A model
application of adaptive control strategy to power reference adaptive system is shown in Fig. 1.
system control is attractive because the effective
system response changes with load level and system
configuration. Whenever an adaptive controller >!Reference|
detects changes in system operating conditions, it model j{
responds by determining a new set of control
parameters. Several papers illustrating the
application of adaptive control to synchronous machine
excitation have appeared over the past two years.
Model reference adaptive control is presented in Refs.
[3 and 4]. Simulation studies with a gradient
following, adaptive approach appeared in Ref. [5]. Set
Simulations. and micromachine studies reported in Ref. pointr
[6] -used an adaptive controller based on minimum input behaviour

variance [MV] concept [7].


Although many aspects of power system control ero
problem.. are different from the problems for which the
XAdaptation
mechanism
adaptive control algorithms were originally developed,
the above studies indicate that the benefits of
adpative control can still be. realized in power
Fig. 1. A model reference adaptive control
systems. Fixed gain controllers are always a
scheme.
compromise between the best settings for light and
heavy load conditions. Adaptive technique, ensures
that the controller parameters are optimal for the Steps: 1) Choose a model representing the desired
operating conditions, and thus the system stability is performance.
enhanced. 2) Compute the error between the outputs of
Characteristics of few adaptive control the reference model and the actual system.
algorithms are discussed in this paper from the point 3) Based on the correlation between this error
of view of their applicability to power systems. A and the system states, update the feedback gains of
modified form of one of the more promising algorithms the system.
that makes it more suitable for power systems is
developed and presented. Certain algorithms exhibit Application: Let the plant, X (t), and the reference
P) th
characteristics which influence their applicability to model, Xm(t), behaviours be given by the nth order
power systems. To illustrate the-effects of these state equations
characteristics generator excitation control through k (t) = A (t) X (t) + B (t) u(t) (1)
p p p p
adaptive stabilizer loop is used as an example. The
k
m
(t) = A
m
(t) Xm (t) + B
m
(t) u(t) (2)
where
is the state vector
x

84 WM 018-8 A paper recommended and approved is the control (input)


u
by the IEEE Power System Engineering Committee of A, B are matrices
the IEEE Power Engineering Society for presenta- and the subscripts p,m refer to plant and model
tion at the IEEE/PES 1984 Winter Meeting, Dallas, respectively.
Texas, January 29 February 3, 1984. Man"lscript .An error function can be formed by writing
submitted July 6, 1983; made available for print-
ing December 13, 1983. e(t) =
Xm(t) -
Xp(t) (3)

0018-9510/84/0800-1983$01.00()1984 IEEE
1984
The aim of the control is to force this error in Fig. 2. An MV controller detects the output error
function to zero. This can be done by choosing a at the first sample, and thereafter endeavours to keep
Liapunov function [81 or using Popov' s the output error identically at zero 4s shown in Fig.
'hyperstability' theory [3]. 3. Note that for the system is eqn. (4), the variance
of the noise, ek is taken as zero.
Characteristics: The control designer must have a
good idea of the performance achievable on the system, 1 ,.0
otherwise control limiting occurs and the resulting
response will be substantially different to that of
the reference system. In the case of an unpredictable
external disturbance, the reference system will not
respond immediately while the actual system will.
This performance difference will be interpreted as a -t -A

system error, and the adaption of system gains will 10 00


occur. Using the series parallel method, as in Ref.
(3] for excitation control, guarantees some model -l
response to the disturbance, but the choice of model
is still a problem particularly in multimachine .0
/Samples
systems.
Note that the controllers reported in Refs. [3
and 81 are analyzed in continuous time domain only, Fig. 2. Impulse response of the uncontrolled
and thus require accurate high speed analog computers system
to perform complex computation.

3. Algorithms With Explicit Identification: 1.0 ri

The other algorithms discussed in this paper all


perform an explicity identification of system
parameters. The identification algorithms are always
in discrete time domain, so the computation of control
is done using sampled data design techniques.
Recursive least squares (RLS) identification technique
(91 can be used for the system transfer function
identification. This identification technique updates
the parameter estimates at every sampling instant.
These updated estimates are then used to compute the
controller parameters. To correctly identify the
system there must always be some variations in the 0 10 30 50
system output. For adaptive excitation control Fig. 3. Impulse response of the system
problem this variation may be caused by continual with M.V. Controller, no limit
changes in load demand. However, if that variation applied.
does not register on the measurement system, a barely
perceptible amount of random noise may be injected
into the exciter. Steps: 1) Digital devices are normally used to
The presence of feedback or of correlation in the implement adaptive control algorithms. Thus, the
external disturbance may influence the accuracy of computations are done in discrete time, and the
identification. However, a so called 'self-tuning' designer must preselect the sampling frequency. As a
property was proven for minimum variance controller guide, choose a sample frequency approximately 10
(10] which states that if the closed loop controller times the normal system frequency of oscillation.
converges it must converge, to the optimal controller, 2) At each sampling instant update the
even when the measurements are corrupted by correlated parameter estimates using the RLS algorithm.
disturbances (colored noise). A similar proof also 3) Compute the control which makes the
exists for pole assigned controller [1.1], although an predicted error zero.
equivalent proof for linear quadratic optimal
controller is yet to be presented. Application: This form of adaptive controller was
used with an output consisting of a weighted sum of
3.1. Minimum Variance (MV) Controller: terminal voltage and speed error [6], and is very easy
to compute. For a plant identified as (101
Concept: The combination of minimum variance control
[7], and RLS identification has been shown (101 to Yk+1 -a= -1= (6)
give good regulation even when the system disturbances
were correlated. An MV controller first predicts the the MV control is given simply as (10]
next measurement for zero control, and then chooses
the control value so that the predicted output error I (7)
is zero. For example, consider the system given by uk layk+a2kk-l-b2uk-1 I
[a
linear difference equation This form of controller may be implemented using
comparatively slow microprocessors, and can easily
(1-1.386q- +0.96q2 )yk =
(q l+0.98q 2)uk + ek (4) achieve a sample rate of 50 ms for machine control.
where u and y are the input and output of the system Characteristics: One unfortunate characteristic of
ek is a sequence of pseudorandom noise, and q 1 is a this control algorithm is the possibility that the
delay operator, such that control computation may ge unstable [7]. In eqn. (7),
2
-1 if the pole at z = - b is on or outside the unit
q Yk~Y- (5)
circle (1z1:1), then control, uk increases without
The response of this system to a unit impulse is given bounds. This non minimum phase condition may occur
1985
when there is a delay associated with the process in iterations at every sample instant enables the
the continuous time domain. In addition to the non controller to track the parameter estimates, and the
minimum phase problem, the cancellation of large controller converges to the limit. To determine a
parameter errors in one sample time as in Fig. 3 is satisfactory r, simulations may be performed, or r may
impossible due to the control limits of the excitation be reduced from a large value till a satisfactory
system. If the control signal is limited, the control is achieved.
resulting control, which is always in phase with the To obtain the states xk of eqn. (9), or state
predicted error, can give rise to poor damping. Fig. estimate required in step 5, an observer is
4 shows the effect of control limit on MV control for required. The observer design as in Ref. (13) may be
ths process of eqn. (4). used, but the use of canonical forms can considerably
simplify the on-line computations.
1.0 Characteristics: The optimal control design technique
has the property that it always gives stable control
provided the parameter estimates are exact. Moreover,
the deterioration of performance with parameter error
is low when using an optimal regulator. However, to
0) obtain this desirable characteristic, a larger
..d O
I . . . I
computational burden must be undertaken.

3.3. Pole Assigned (PA) Controller:


Samples
Concept: This algorithm is similar to that of model
-1.0I reference in that the desirable response is pre-
specified. In this case, however, only the desired
Fig. 4. System response with band limited system closed-loop poles are specified, and the update
M.V. Controller. of controller parameters is based on explicit system
identification. Whereas the MV controller shifts all
the poles towards the origin, the PA controller has
3.2. Optimal Linear Quadratic (LQ) Controller: the freedom to place the poles at other locations.
This permits a trade-off between performance and
Concept: The concept behind this class of regulators control effort.
is to minimize a linear quadratic performance index of
the form Steps: Steps 1 and 2 are the same as for the MV
Go
2 2 controller.-
J = Yk +r uk-1] (8) 3) Compute control which will then place poles
k=1 to pre prescribed locations.
where uk and Yk are the input and output of the
process. Applications: For a single input single output system
For r . 0, the optimal control minimizes the the process is given by the transfer function
output error subject to the requirement that the
control effort be small. For r + 0, the penalty on B
Yk A Uk (14)
large control is reduced, and in the limit gives
minimum variance performance. For r + -, control where Yk is the sampled output of the system, uk is
action is taken only to stabilize an unstable the computed control at sample instants, A,B are
system. Thus, with a proper choice of r excessive polynomials given by
saturation in the controller may be avoided. -1 -na
(15)
Steps: Steps 1 and 2 are the same as for the MV -n
controller. B =b1q +....-I*bnnb qb
3) Describe the system in a state space form
as and q1 is the delay operator.
The control is computed from l11]
xk+1 = F xkg uk (9) G
(16)
h xk (1 0)
Yk =
where polynomials F and G are given by
where matrix F and vectors g and h are determined from
estimates in step 2. F l+flq 1+o**O+fn q f; nf = nb-l (1 7)
4) Obtain steady state solution of equations f
[12] G
-1
go+g1q +*ov+gn~g -n9g; ngg =
na-1
a
Kk+lk+1+rg
= -EgT ] 1 T k (11 ) Then choosing the closed-loop poles as roots of
the polynomial T, given by
k+1
rFklT S+Thh
l+9kk SkF+hI (1 2)
+000,tnt -nt
=
-1 (18)
5) Obtain the control from
T =
1tlq
the control parameters fi and gi are computed from
Mk = kkxk (13)
AF+BG = T (19)
Application: To obtain steady state solution of eqns.
(11) and (12) requires infinite number of where nt(na+nb-l (20)
iterations. This is not possible in real time
applications. A direct solution for steady-state is For na = nb = n, solution of eqn. (1 9) takes the
possible using eigenvalue technique [131, but it might form
be computationally expensive. A small number of
1986
1 0 .. . 0 b1 0 fi - tl -a, 4. Adaptive Controllers For Power Systems:

a1 1 b2 b1 f2 Desirable controller requirements for application


in power system environment may be described as i) The
controller should not saturate under transient
conditions. ii) The controller should not be unstable
under non minimum phase condition. iii) As an 'a
priori' decision on the parameters cannot be made,
an ... a2 bn f
n0n-1 preselection of required performance be avoided. iv)
Trade-off between best control effort and control
0 ... a3 0 tn+l action should preferably be made on a single
parameter.
Characteristics of the MV controller described in
the previous section indicate that it does not satisfy
... . . the first two properties. Similarly, the MRAC and PA
controller are not suitable from the point of view of
0 0 ... an t2n-1 the 3rd property above. For the MV controller the 4th
property is not a constraint, because it does not need
(21 ) any outside interference to influence the control.
Eqn. (21) can be written as The MRAC and PA controller do not possess the 4th
property, because in one case a model has to be
Cx = d preselected, and characteristic equation in the other.
The LQ controller does not suffer from the first
where C is (2n-1)x(2n-1) matrix, and x and d are three constraints and possesses the property of using
vectors each having (2n-1) elements. The solution of a single parameter for obtaining the trade-off. It
eqn. (21) is therefore seems to be suitable for application in
power system.
x = C d (22) In the PA controller a poor choice of T in eqn.
(18) may lead to an unstable F-polynomial, and thus an
where the inverse in eqn. (22) exists provided that A unstable controller. This can easily happen when 'a
and B have no common factors [14,p. 145]. Note that, priori' decision on the system transfer function
if the elements of d are very small, the coefficients cannot be made. It is proposed to bovercome this
of F and G are small. With the coefficients of F drawback by adding an engineering constraint so that
small, the roots of F are close to zero, and a stable the poles are shifted towards the origin radially,
controller is achieved. instead of choosing them arbitrarily. The PA
controller with- this proposed modification - Pole
Characteristics: Unlike the MV controller, the PA Shifting (PS) controller is described below.
controller places the closed-loop poles at
preprescribed locations. If the closed-loop poles are 4.1. Pole Shifting (PS) Controller:
moved a small distance from their open-loop locations,
only a small amount of control is required. Hence, This is in essence a PA controller, and has the
even if the control is bound limited, it will still same advantages. It follows similar steps as PA
provide better damping. For example, for the system controller.
given by eqn. (4), if T is chosen as If the open-loop poles are shifted radially
towards the origin such that T becomes
T = 1-1. 247q +0. 777q (23) -1 n
T = A (afq ) =
i+cta1q + x)0anq s* (24)
the system response to an impulse when under the
control of a PA controller is shown in Fig. 5. where af, called alpha factor, is close to, but less
Comparing Fig. 5 with Fig. 4 it can be seen that the than, one, then d in eqn. (22) becomes
PA controller provides much better damping than the MV
dT [a2
controller when both controllers are working under d
= l (C -1)a
f 2
l-1 )...
f
.a
n
(an_-1)0
f
... 0] (25)
control limits. It can be seen from eqn. (1 6) that
the control is not directly computed from inverse of With af close to one, elements of d are close to zero,
polynomial B as in the MV case. Thus the possibility the values of fi and gi are small, and a stable
of achieving a stable cntroller is higher for a zero controller is achieved. So, for a stable open-loop
mean or non-minimum phase system. However, there is identified system, computing T on-line from eqn. (24)
no specific guideline on the choice of T, and in this will always result in a stable controller.
aspect this algorithm is similar:.to MRAC. Reduction of magnitude of coefficients of A to
f orm T, such that T becomes
1.0
AI
T =
1+sign a1 la-aflq + +sign anlan-OLf1q
is not a valid solution to the problem. Although it
is easy to implement, it might cause an unstable T,
a)
and hence an unstable closed loop system. As can be
noticed, the PS controller has the 4th property
4c:0 I. ...J mentioned at the start of this section, ie., a single
30 50 parameter, ctf trade-off to determine best -control
effort.
Samples
5. Steady State Vs. Transient Control:
1.0 V
All the algorithms discussed in the paper are for
Fig. 5. System response with band limitted single input, single output systems. For the desired
P.A. Controller. control using excitation system, control under normal
1987
operating conditions is based on voltage feedback order, it is identified as if na = nb = 3. It can be
through an automatic voltage regulator (AVR). It is seen from Fig. 8 that during the steady state there is
well known that the damping of transients under not much variation in the parameters. After the
disturbance conditions requires additional stabilizing disturbance, the parameters move to the new steady
signals such as speed or power. Although the voltage state values.
does reflect system oscillations it is a less direct
and useful measure of the electrical power deviations. Next a short circuit of 0. 4 s duration at the
One could perform adaptive control using an middle of one of the two parallel lines is applied
output signal which combines steady state and with a successful reclosure. Load angle responses for
transient effects by adding the voltage and speed or this case can be seen in Fig. 9 for three different
power error signals (61. This approach gives some stabilizers, where the controllers bring the machine
feedback control of voltage, but the gain that can be back to original operating point after successful
used is very limited (10-20) as compared to that with reclosure. Lastly, a 0.4 s duration short-circuit
pure AVRs (100-400). followed by a loss of line is applied on one of the
Noting that conventional AVR performs the task of two lines, and responses are shown in Fig. 10. It can
steady state control adequately, it is preferable to be seen from Figs. 7, 9 and 10 that although LQ
add an adaptive stabilizer signal to improve the stabilizer is a little inferior in settling time than
damping of the transients. This adaptive stabilizer PS stabilizer, both of them have much faster settling
approach is shown in Fig. 6. time. Also, in all cases, except one, the overshoot
is less than that with the conventional stabilizer.

68.0 I\

)I II %I /
, -,,
/ N% e
I I I -

I') I
57.01 t kI
I
10

.-1 46.0 P.S. Stabilizer


0 ---
L.Q. Stabilizer
Conventional
35.0 Stabilizer
0.00 1.20 2.40 3.60 4.00 6.00
TIME s
Fig . 7. A 0.25 p.u. step change in load.

1.25 a2
Controller Y
.640.
ldaptive stabilizerj
Fig. 6. Proposed excitation control scheme. _: a3
6. Simulation Studies: - .58
Digital computer simulation studies performed on
an HP 21MX minicomputer are described in this -1. 2
section. The system under consideration is a a1
synchronous machine tied to an infinite bus through a
long double circuit transmission line. The machine is -1 . E
modelled by a set of non-linear differential equations 0.00 1.20 2.40 3.00 4.00 6.00
based on Park 's equations [15]. The system model is TIME s
completed by incorporating an AVR and an adaptive
stabilizer as shown in Fig. 6. Fig. 8. Parameter variations during
The electrical power output was chosen as step change in load.
stabilizing signal for the adaptive stabilizer.
Although recently accelerating power has received more
attention as the stabilizing signal, electrical power
was chosen because of simplicity. The electrical 54.0.
power is then processed through a washout circuit to
a)
eliminate the d.c. component, and is sampled every 100
ms. The characteristic of the washout circuit is t 36.0
given in Ref. [161. The system identification is
performed by superimposing a sequence of pseudorandom :
noise with the control input, u. . ______ P.S. Stabilize:Ir
W 18.0 ______ L.Q. Stabilize:
With the system operating at steady state at 0.5
p.u. load, the load is changed suddenly to 0.75 p.u. . Co___Conventional
Load angle variation under such a condition is shown 0.00 I
Stabilizer
in Fig. 7, where PS and LQ stabilizers are compared 0.00 1.20 2.40 3.60 4.80 6.00
with conventional stabilizer deriving its input from TIME s
accelerating power [17] . The behaviour of ai
parameters are given in Fig. 8. It should be noted Fig. 9. Short circuit test, successful
that although the actual system model is of 10th reclosure.
1988

7. K. J. Astrom: Introduction to stochastic control


theory. Academic Press, 1970.
; 68.0
8. H. R. Outhred and P. J. Evans: A model reference
adaptive controller for turbo-alternators in large
57.0 power systems. Power System Computation
Conference Proceedings, Grenoble, 1972, Paper
,
#3. 1/9.
H , P.S. Stabilizer
0- 46.0 -' PL.Q. Stabilizer Dynamic system
9. G.C. Goodwin and R.L. Payne:
Conventional identification: Experiment design and data
35.0
Stabi lizer analysis. Academic Press, 1977.
0.00 1.20 2.40 3.60 4.80 6.00
10. K.J. Astrom and B. Wittenmark: On self-tuning
TIME s
regulators. Automatica, Vol. 9, pp. 185-199,
1973.
Fig. 10. Short circuit and line loss test.
11. P.E. Wellstead, D. Prager and P. Zanker: Pole
assigned self-tuning regulator. Proc. IEE, Vol.
7. Conclusions: 126, pp. 781-787, 1979.
An overall view of applicability, advantages and 12. B.D.O. Anderson and J.B. Moore: Linear optimal
disadvantages of adaptive controllers for power control, Prentice-Hall, 1971.
systems have been examined in this paper. With
certain engineering constraints, benefits can be 13. T. Kailath: Linear Systems, Prentice-Hall, 1980.
obtained by using adaptive controllers. Because they
are designed primarily as transient controllers, 14. L.E. Dickson: First course in the theory of
special attention must be paid towards a good steady equations. Wiley, 1972.
state control. This has been illustrated for an
excitation control scheme where an AVR maintains 15. B. Adkins and H.G. Harley: The general theory of
steady-state performance, and the adaptive controller alternating current machines. Chapman and Hall,
works as a transient gain stabilizer. 1 975.
Among the adaptive controllers discussed,
explicitly identified controllers are preferred for 16. D.C. Lee, R.E. Beaulieu and J.R.R. Service: A
power system applications. Minimum variance power system stabilizer using speed and electrical
controllers can only be used for small disturbance power inputs- design and field experience. IEEE
cases, provided the system is minimum phase. Pole Trans. Power Apparatus and Systems, Vol. PAS-100,
assigned controllers can only be used in the cases pp. 4151-4157, 1981.
where the system is well known, although it can damp
out large disturbances efficiently. Linear quadratic 17. F.P. DeMello, L.N. Hannett, J.M. Undrill:
and pole shifting controllers can be most beneficial Practical approaches to supplementary stabilizing
when the system is not well known, but has stable from accelerating power. IEEE Trans. on Power
open-loop poles. Apparatus and Systems, Vol. PAS-97, pp. 1515-1522,
1 978.
References:

1. W. Borisson and R. Syding: Self tuning control of


an ore crusher. Automatica, Vol. 12, pp. 1-7,
1976.

2. C.G. Kallstrom, K.J. Astrom, N.E. Thorell, J.


Erikson and L. Sten: Adaptive autopilots for
large transfers. Proceedings IFAC World Congress,
Helsinki, 1978, paper #9B.5.

3. E. Irving, J.P. Basset, C. Charcossey and J.P.


Nonville: A method to improve the power network
'steady state stability' and to reduce the unit
stress. The multivariable adaptive control of
generators. Proceedings, IFAC Symposium on
Automatic Control and Protection of Electric Power
Systems, Melbourne, 1977, pp. 60-69.

4. O.P. Malik, G.S. Hope and R. Ramanujam: Real-time


model reference adaptive control of synchronous
machine excitation. IEEE Power Engg. Society
Winter Meeting, New York, 1978, Paper #A78 297-4.

5. G. Ledwich: Adaptive excitation control. Proc.


IEE, Vol. 126, pp. 249-253, 1979.

6. M.A.H. Sheirah, O.P. Malik and G.S. Hope: A self-


tuning automatic voltage regulator. Electric
Power System Research, Vol. 2, pp. 199-213, 1979.
1989
Discussion A. Ghosh, G. Ledwich, 0. P. Malik, and G. S. Hope: the authors wish
W. Watson (Ontario Hydro, Toronto, ON, Canada): This paper to thank the discusser for his comments.
demonstrates some practical possibilities in the application of adaptive The authors have demonstrated before [A] the application of a similar
control to steady state stability problems in power systems. I am aware adaptive technique to the transient stability problem. In theory it is quite
of power systems where configurations may change to such an extent possible to identify the system during the transient state. However, it
that adaptive control of stabilizer parameters would be of value. Such is essential that the speed at which the identification is performed be com-
is not the case at Ontario Hydro, however, where we have optimized patible with the speed of the transient, i.e. the rate at which the system
stabilizer parameters to ensure the maximum steady state stability limit states are changing. If the transient is too fast for the identification routine
with a system operating close to that limit. to be able to track it, large errors in identification will result. This will
However, although it may fall outside the scope of the paper, I would tend to make the control very large and the system performance will
appreciate the comments of the authors on the possible application of deteriorate.
adaptive techniques to the transient stability problem The computational speed of microprocessors is still not fast enough
In a 1975 paperll] a control strategy was described in which, follow- for the identification routines to be able to track the power system tran-
ing a transient fault the control algorithm of the exciter and stabilizer sients in real-time. Because of this, the authors have proposed in the paper
was modified to introduce an additional signal proportional to machine that the identification be turned-off during the fastest part of the tran-
angle deviation for specified levels of the variables. This was shown to sient. this gives a satisfactory performance without the control becom-
be effective in increasing transient stability limits for certain types of ing very large at any time.
disturbance. We are now studying the implementation of this type of
control on a wider system basis. The authors may care to comment on
the application of system identification techniques to this situation. REFERENCE
REFERENCE [Al M. A. H. Sheirah, 0. P. Malik and G. S. Hope, "Self-tuning voltage
[1] J. P. Bayne, P. Kundur, W. Watson, "Static Exciter Control to Im regulator-implementation and test results", Paper #A79 060-5, IEEE
prove Transient Stability, IEEE Transactions on Power, Apparatus Power Engg. Society 1979 Winter Meeting.
and Systems, vol PAS 94 no 4 pp. 1141-1146, July/August, 1975.
Manuscript received February 13, 1984. Manuscript received April 9, 1984.

Vous aimerez peut-être aussi