chapter 8

The Effect of Excitation on Stability

8.1 Introduction
Considerable attention has been given in the literature to the excitation system and
its role in improving power system stability. Early investigators realized that the so-
called “steady-state” power limits of power networks could be increased by using the
then available high-gain continuous-acting voltage regulators [ I ] . It was also recognized
that the voltage regulator gain requirement was different at no-load conditions from
that needed for good performance under load. In the early 1950s engineers became
aware of the instabilities introduced by the (then) modern voltage regulators, and stabi-
lizing feedback circuits came into common use (21. In the 1960s large interconnected
systems experienced growing oscillations that disrupted parallel operation of large sys-
tems [3-121. It was discovered that the inherently weak natural damping of large and
weakly coupled systems was the main cause and that situations of negative damping
were further aggravated by the regulator gain [ 13). Engineers learned that the system
damping could be enhanced by artificial signals introduced through the excitation sys-
tem. This scheme has been very successful in combating growing oscillation problems
experienced in the power systems of North America.
The success of excitation control in improving power system dynamic performance
in certain situations has led to greater expectations among power system engineers
as to the capability of such control Because of the small effective time constants in
the excitation system control loop, it was assumed that a large control effort could be
expended through excitation control with a relatively small input of control energy.
While basically sound, this control is limited in its effectiveness. A part of the engi-
neer’s job, then, is to determine this limit, i.e., to find the exciter design and control
parameters that can provide good performance at reasonable cost [ 141.
The subject of excitation control is further complicated by a conflict in control
requirements in the period following the initiation of a transient. In the first few cycles
these requirements may be significantly different from those needed over a few seconds.
Furthermore, it has been shown that the best control effort in the shorter period may
tend to cause instability later. This suggests the separation of the excitation control
studies into two distinct problems, the transient (short-term) problem and the dynamic
(long-term) problem. It should be noted that this terminology is not universally used.
Some authors call the dynamic stability problem by the ambiguous name of “steady-
state stability.” Other variations are found in the literature, but usually the two prob-
lems are treated separately as noted.
309
310 Chapter 8

8.1.1 Transient stability and dynamic stability considerations

In transient stability the machine is subjected to a large impact, usually a fault,
which is maintained for a short time and causes a significant reduction in the machine
terminal voltage and the ability to transfer synchronizing power. If we consider the
one machine-infinite bus problem, the usual approximation for the power transfer is
given by

P = (V,V,/x)sinb (8.1)

where V, is the machine terminal voltage and V , is the infinite bus voltage. Note
that if V, is reduced, P is reduced by a corresponding amount. Prevention of this
reduction in P requires very fast action by the excitation system in forcing the field to
ceiling and thereby holding V , at a reasonable value. Indeed, the most beneficial
attributes the voltage regulator can have for this situation is speed and a high ceiling
voltage, thus improving the chances of holding V , at the needed level. Also, when
the fault is removed and the reactance x of (8.1) is increased due to switching, another
fast change in excitation is required. These violent changes affect the machine’s ability
to release the power it is receiving from the turbine. These changes are effectively
controlled by very fast excitation changes.
The dynamic stability problem is different from the transient problem in several
ways, and the requirements on the excitation system are also different. By dynamic
stability we mean the ability of all machines in the system to adjust to small load
changes or impacts. Consider a multimachine system feeding a constant load (a con-
dition never met in practice). Let us assume that at a given instant the load is
changed by a small amount, say by the energizing of a very large motor somewhere in
the system. Assume further that this change in load is just large enough to be recog-
nized as such by a certain group of machines we will call the control group. The
machines nearest the load electrically will see the largest change, and those farther
away will experience smaller and smaller changes until the change is not perceptible
at all beyond the boundary of the control group.
Now how will this load change manifest itself at the several machines in the control
group? Since it is a load increase, there is an immediate increase in the output power
requirements from each of the machines. Since step changes in power to turbines
are not possible, this increased power requirement will come first from stored energy in
the control group of machines. Thus energy stored in the magnetic field of the machines
is released, then somewhat later, rotating energy [( 1/2)mu2]is used to supply the
load requirements until the governors have a chance to adjust the power input to the
various generators. Let us examine the behavior of the machines in the time interval
prior to the governor action. This interval may be on the order of 1 s. In this
time period the changes in machine voltages, currents, and speeds will be different for
each machine in the control group because of differences in unit size, design, and elec-
trical location with respect to the load. Thus each unit responds by contributing its
share of the load increase, with its share being dictated by the impedance it sees at its
terminals (its Thevenin impedance) and the size of the unit. Each unit has its own
natural frequency of response and will oscillate for a time until damping forces can
decay these oscillations. Thus the one change in load, a step change, sets up all
kinds of oscillatory responses and the system “rings” for a time with many frequencies
present, these induced changes causing their own interaction with neighboring machines
(see Section 3.6).
 Effect of Excitation o n Stability 31 1

Now visualize the excitation system in this situation. In the older electromechani-
cal systems there was a substantial deadband in the voltage regulator, and unless the
generator was relatively close to the load change, the excitation of these machines would
remain unchanged. The machines closer to the load change would recognize a need for
increased excitation and this would be accomplished, although somewhat slowly.
Newer excitation systems present a different kind of problem. These systems recognize
the change in load immediately, either as a perceptible change in terminal voltage,
terminal current, or both. Thus each oscillation of the unit causes the excitation sys-
tem to t r y to correct accordingly, since as the speed voltage changes, the terminal
voltage also changes. Moreover, the oscillating control group machines react with one
another, and each action or reaction is accompanied by an excitation change.
The excitation system has one major handicap to overcome in following these
system oscillations: this is the effective time constant of the main exciter field which
is on the order of a few seconds or so. Thus from the time of recognition of a desired
excitation change until its partial fulfillment, there is an unavoidable delay. During
this delay time the state of the oscillating system will change, causing a new excitation
adjustment to be made. This system lag then is a detriment to stable operation, and
several investigators have shown examples wherein systems are less oscillatory with the
voltage regulators turned off than with them operating [7, 121.
Our approach to this problem must obviously depend upon the type of impact
under consideration. For the large impact, such as a fault, we are concerned with
maximum forcing of the field, and we examine the response in building up from normal
excitation to ceiling excitation. This is a nonlinear problem, as we have seen, and the
shape of the magnetization curve cannot be neglected. The small impact or dynamic
stability problem is different. Here we are concerned with small excursions from nor-
mal operation, and linearization about this normal or “quiescent” point is possible and
desirable. Having done this, we may study the response using the tools of linear sys-
tems analysis; in this way not only can we analyze but possibly compensate the system
for better damping and perhaps faster response.

8.2 Effect of Excitation on Generator Power limits

We begin with a simple example, the purpose of which is to show that the excitation
system can have an effect upon stability.

Example 8.1
Consider the two-machine system of Figure 8.1, where we consider one machine
against an infinite bus. (This problem was introduced and analyzed by Concordia [ 11.)
The power output of the machine is given by
P = [ E I E z / ( X I + X2)]sin 6
6 = 61 + 6 2

Fig. 8.1 One machine-infinite bus system.

31 2 Chapter 8

Fig. 8.2 Phasor diagram for Example 8.1.

This equation applies whether or not there is a voltage regulator. Determine the effect
of excitation on this equation.

Solution
We now establish the boundary conditions for the problem. First we assume that
XI = X 2 = 1.0 pu and that V, = 1 .O pu. Then for any given load the voltages E , and
E2must assume a certain value to hold at 1.0 pu. If the power factor is unity, E,
and E2 have the same magnitude as shown in the phasor diagram of Figure 8.2. If
E, and E2 are held constant at these values, the power transferred to the infinite
bus varies sinusoidally according to (8.2) and has a maximum when 6 is 90".
Now assume that E, and E2are both subject to perfect regulator action and that
the key to this action is that V, is to be held at 1.0 pu and the power factor is to be
held at unity. We write in phasor notation

E, = 1 + jf = dmej*/z E2 = I - jf = dme-j6/2

Adding these equations we have

E, + E2 = 2 = 2 r n C O S 6 / 2

I
I
I
I
I
I
I
I
I
I
I
I
Angle 6, degrees

Fig. 8.3 Comparison of power transferred at unity power factor with and without excitation control.
 Effect of Excitation on Stability 313

or

= E2 =
I
-
El (8.3)
cos 612
Substituting (8.3) into (8.2) and simplifying, we have for the perfect regulator, at unity
power factor,

P = tan612 (8.4)
The result is plotted in Figure 8.3 along with the same result for the case of constant
(unregulated) E l and E 2 .
In deriving (8.4), we have tacitly assumed that the regulators acting upon E l and
E 2 do so instantaneously and continuously. The result is interesting for several reasons.
First, we observe that with this ideal regulation there is no stability limit. Second, it is
indicated that operation in the region where d > 90" is possible. We should comment
that the assumed physical system is not realizable since there is always a lag in the
excitation response even if the voltage regulator is ideal. Also, excitation control of the
infinite bus voltage is not a practical consideration, as this remote bus is probably not
infinite and may not be closely regulated.

Example 8.2
Consider the more practical problem of holding the voltage E2 constant at I .O pu
and letting the power factor vary, other things being the same.

Solution
Under this condition we have the phasor diagram of Figure 8.4 where we note
that the locus of E2 is the dashed circular arc of radius 1.0. Note that the power factor
is constrained by the relation

e, = a2/2 (8.5)

where 8, = IT - 8 and 6 = 6, +
6,.
Writing phasor equations for the voltages, we have

Fig. 8.4 Phasor diagram for Example 8.2.

314 Chapter 8

Toque Angle, 4, degrees

Fig. 8.5 System parameters as a function of 6 2 .

El = I + jT = I - [sine + j l c o s e = ~ , e ' "

E2 = 1 - j r = I + Isin6 - jfc o s e = E2e-jb2 (8.6)
where 6, el, A I , and a2 are all measured positive as counterclockwise. Noting that E2 =
1, we can establish that
I = 2sin0, E , sin6 = 2sinJ2
sin 6 = 2sin6, tan6, = sinb2/(2 - cos&)
03-71
Thus once we establish J2. we also fix 0, I, 6, and 6 , , although the relationships
among these variables are nonlinear. These results are plotted in Figure 8.5 where
equations (8.7) are used to determine the plotted values. We also note that
P = V,~COS~ (8.8)
but from the second of equations (8.6) we can establish that I cos 8 = sin a2 or
P = sin& (8.9)
so 62 also establishes P. Thus P does have a maximum in this case, and this occurs
when 62 = 90" (E' pointing straight down in Figure 8.4). In this case we have at
maximum power
E, = 2 + jl = 2.235/26.6" I = 1.414
e = -450 6 = 116.6"
The important thing to note is that P is again limited, but we see that 6 may go
 Effect of Excitation o n Stability 315

0 90 180
Torque Angle b, degrees
Fig. 8.6 Variation of P with 6.

beyond 90" to achieve maximum power and that this requires over 2 pu E , . The
variation of P with 6 is shown in Figure 8.6.

These simple examples show the effect of excitation under certain ideal situations.
Obviously, these ideal conditions will not be realized in practice. However, they provide
limiting values of the effect of excitation on changing the effective systey parameters.
A power system is nearly a constant voltage system and is made so because of system
component design and close voltage control. This means that the Thevenin impedance
seen looking into the source is very small. Fast excitation helps keep this impedance
small during disturbances and contributes to system stability by allowing the required
transfer of power even during disturbances. Finally, it should be stated that while the
ability of exciters to accomplish this task is limited, other considerations make it
undesirable to achieve perfect control and zero Thevenin impedance. Among these is
the fault-interrupting capability.
8.3 Effect of the Excitation System on Transient Stability
In the transient stability problem the performance of the power system when sub-
jected to severe impacts is studied. The concern is whether the system is able to main-
tain synchronism during and following these disturbances. The period of interest is
relatively short (at most a few seconds), with the first swing being of primary impor-
tance. In this period the generator is suddenly subjected to an appreciable change in
its output power causing its rotor to accelerate (or decelerate) at a rate large enough
to threaten loss of synchronism. The important factors influencing the outcome are the
machine behavior and the power network dynamic relations. For the sake of this dis-
cussion it is assumed that the power supplied by the prime movers does not change in
the period of interest. Therefore the effect of excitation control on this type of transient
depends upon its ability to help the generator maintain its output power in the period
of interest.
To place the problem in the proper perspective, we should review the main factors
that affect the performance during severe transients. These are:
1. The disturbing influence of the impact. This includes the type of disturbance, its
location, and its duration.
2. The ability of the transmission system to maintain strong synchronizing forces during
the transient initiated by a disturbance.
3. The turbine-generator parameters.
The above have traditionally been the main factors affecting the so-called first-swing
transients. The system parameters influencing these factors are:
316 Chapter 8

1. The synchronous machine parameters. Of these the most important are: (a) the
inertia constant, (b) the direct axis transient reactance, (c) the direct axis open cir-
cuit time constant, and (d) the ability of the excitation system to hold the flux
level of the synchronous machine and increase the output power during the transient.
2. The transmission system impedances under normal, faulted, and postfault condi-
tions. Here the flexibility of switching out faulted sections is important so that large
transfer admittances between synchronous machines are maintained when the fault is
isolated.
3. The protective relaying scheme and equipment. The objective is to detect faults
and isolate faulted sections of the transmission network very quickly with minimum
disruption.

8.3.1 The role of the excitation system in classical model studies

In the classical model it is assumed that the flux linking the main field winding
remains constant during the transient. If the transient is initiated by a fault, the arma-
ture reaction tends to decrease this flux linkage [15]. This is particularly true for the
generators electrically close to the location of the fault. The voltage regulator tends
to force the excitation system to boost the flux level. Thus while the fault is on, the
effect of the armature reaction and the action of the voltage regulator tend to counter-
act each other. These effects, along with the relatively long effective time constant of
the main field winding, result in an almost constant flux linkage during the first swing of
1 s or less. (For the examples in Chapter 6 this time constant K37j0is about 2.0 s.)
It is important to recognize what the above reasoning implies. First, it implies the
presence of a voltage regulator that tends to hold the flux linkage level constant. Sec-
ond, it is significant to note that the armature reaction effects are particularly pro-
nounced during a fault since the reactive power output of the generator is large. There-
fore the duration of the fault is important in determining whether a particular type of
voltage regulator would be adequate to maintain constant flux linkage.
A study reported by Crary [2] and discussed by Young [ 151 illustrates the above.
The system studied consists of one machine connected to a larger system through a 200-
mile double circuit transmission line. The excitation system for the generator is Type 1
(see Chapter 7) with provision to change the parameters such that the response ratio
(RR)varies from 0.10 to 3.0 pu. The former corresponds to a nearly constant field
voltage condition. The latter would approximate the response of a modern fast excita-
tion system. Data of the system used in the study are shown in Figure 8.7. A transient
stability study was made for a three-phase fault near the generator. The sending end
power limits versus the fault clearing time are shown in Figure 8.8 for different exciter
responses (curves 1-5) and for the classical model (curve 6).
From Figure 8.8 it appears that the classical model corresponds to a very slow and
weak excitation system for very short fault clearing times, while for longer clearing
times it approximates a rather fast excitation system. If the nature of the stability
study is such that the fault clearing time is large, as in “stuck breaker” studies [IS],
the actual power limits may be lower than those indicated when using the classical
model.
In another study of excitation system representation [ 161 the authors report (in a
certain stability study they conducted) that a classical representation showed a certain
generator to be stable, while detailed representation of the generator indicated that loss
of synchronism resulted. The authors conclude that the dominant factor affecting loss
 Effect of Excitation on Stability 317

Exciter

Fault

Generator: Regulating system:

xd = 0.63 pu Pz = 20
x = 0.42 P U P,,= 4
Xd7 = 0.21 pu r, = 0.47 s
H = 5.0 s E,,, = 2.25 PU
=
T ~ O5.0 s E,i, = -0.30 PU
x, = 0.10 pu
Line: System damping:
x = 0.8 Il/mi/line Fault
r = 0.12 Q/mi/line Fault on cleared
y = 5.2 x mho/mi/line
rdll I 4
System: Td12 3
x, = 0.2 pu Td21 0 3
H = 50.0 s rd22 15 18
Fig. 8.7 Two-machine system with 200-mile transmission lines.

of synchronism is the inability of the excitation system of that generator, with response
ratio of 0.5, to offset the effects of armature reaction.

8.3.2 Increased reliance on excitation control to improve stability

Trends in the design of power system components have resulted in lower stability
margins. Contributing to this trend are the following:
I . Increased rating of generating units with lower inertia constants and higher pu re-
actances.

::I\
2. Large interconnected system operating practices with increased dependence on the
transmission system to carry greater loading.
These trends have led to the increased reliance on the use of excitation control as a

i*
.-
-0

a
1.05 Curve re? RR
2 I 0.042 s 3.0
L? 1.00 2 0.17 s 2.0
a 3 0.68 s I .o
-59.0 4 2.70 s 0.25
0 0.02 0.04 0.06 0.08 0.10 5 11.0 s 0.10
Fault Clearing Time, I 6 Classical model
Fig. 8.8 Sending-end power versus fault clearing time for different excitation system responses.
318 Chapter 8

0.0 1.0 2.0 3.0

Time, I

6)

lime, s

(C 1
Fig. 8.9 Results of excitation system studies on a western U.S. system: (a) One-line diagram with fault lo-
cation, (b) frequency deviation comparison for a four-cycle fault, (c) frequency deviation compari-
son for a 9.6-cycle fault: A = 2.0 ANSI conventional excitation system; B = low time constant ex-
citation system with rate feedback; C = low time constant excitation system without rate feedback.
(@ IEEE. Reprinted from IEEE Trans.. vol. PAS-90, Sept./Oct. 1971.)

means of improving stability [ 17). This has prompted significant technological ad-
vances in excitation systems.
As an aid to transient stability, the desirable excitation system characteristics are
a fast speed of response and a high ceiling voltage. With the help of fast transient
forcing of excitation and the boost of internal machine flux, the electrical output of the
machine may be increased during the first swing compared to the results obtainable
with a slow exciter. This reduces the accelerating power and results in improved
transient performance.
 Effect of Excitation on Stability 319

Modern excitation systems can be effective in two ways: in reducing the severity
of machine swings when subjected to large impacts by reducing the magnitude of the
first swing and by ensuring that the subsequent swings are smaller than the first. The
latter is an important consideration in present-day large interconnected power systems.
Situations may be encountered where various modes of oscillations reinforce each other
during later swings, which along with the inherent weak system damping can cause
transient instability after the first swing. With proper compensation a modern excita-
tion system can be very effective in correcting this type of problem. However, except
for transient stability studies involving faults with long clearing times (or stuck
breakers), the effect of the excitation system on the severity of the first swing is rela-
tively small. That is, a very fast, high-response excitation system will usually reduce
the first swing by only a few degrees or will increase the generator transient stability
power limit (for a given fault) by a few percent.
In a study reported by Perry et al. [I81 on part of the Pacific Gas and Electric
Company system in northern California, the effect of the excitation system response on
the system frequency deviation is studied when a three-phase fault occurs in the network
(at the Diablo Canyon site on the Midway circuit adjacent to a 500-kV bus). Some
of the results of that study are shown in Figure 8.9. A one-line diagram of the network
is shown in Figure 8.9(a). The frequency deviations for 4-cycle and 9.6-cycle faults
are shown in Figures 8.9(b) and 8.9(c) respectively. The comparison is made between
a 2.0 response ratio excitation system (curve A ) , a modern, low time constant excita-
tion with rate feedback (curve B) and without rate feedback (curve C). The results
of this study support the points made above.

8.3.3 Parametric study

Two recent studies [ 17,191 show the effect of the excitation system on “first-swing’’
transients. Figure 8.10 shows the system studied where one machine is connected to
an infinite bus through a transformer and a transmission network. The synchronous
machine data is given in Table 8. I .
The transmission network has an equivalent transfer reactance A’, as shown in

Table 8.1. Machine Data for the Studies

of Reference [ 191
xd = 1.72 pu T;O = 6.3 s
X; = 0.45 PU T;O = 0.033 s
XE = 0.33 PU 7;o = 0.43 S
2
X:
=
=
=
1.68 PU
0.59 pu
0.33 PU
7p
H
=
=
0.033s
4.0 s

-
Fig. 8.10 System representation used in a parametric study of the effect of excitation on transient stability.
(e IEEE. Reprinted from IEEE Trans.. vol. PAS-89, July/Aug. 1970.)
320 Chapter 8

Figure 8.10. A transient is initiated by a three-phase fault on the high-voltage side

of the transformer. The fault is cleared in a specified time. After the fault is cleared,
the transfer reactance X , is increased from x , b (the value before the fault) to X,, (its
value after the fault is cleared). The machine initial operating conditions are sum-
marized in Table 8.2.

Table 8.2. Prefault Operating Conditions,

All Values in p u

0.2 1.0 0.94 0.90 0.39

0.4 1.0 0.90 0.90 0.45
0.6 1.0 0.91 0.90 0.44
0.8 1.0 0.97 0.90 0.44

With the machine operating at approximately rated load and power factor, a three-
phase fault is applied at the high-voltage side of the step-up transformer for a given
length of time. When the fault is cleared, the transmission system reactance is changed
to the postfault reactance X,, and the simulation is run until it can be determined if the
run is stable or unstable. This is repeated for different values of X , until the maxi-
mum value of X,,, is found where the system is marginally stable.
Two different excitation system representations were used in the study:
1. A 0.5 pu response alternator-fed diode system shown in Figure 8.1 1.
2. A 3.0 pu response alternator-fed SCR system with high initial response shown in
Figure 8.12. This system has a steady-state gain of 200 pu and a transient gain of
20 pu. An external stabilizer using a signal V , derived from the shaft speed is also
used (see Section 8.7).

“REF
I
1
-0.0445 + 0 . 5 II -
‘FD

I U

-
1
0.16s
+ I

Fig. 8.1 I Excitation block diagram for a 0.5 R R alternator-fed diode system. (c IEEE. Reprinted from
IEEE Trans.,VOI. PAS-89, July/Aug. 1970.)

From the data presented in [ 191, the effect of excitation on the “first-swing” tran-
sients is shown in Figure 8.13, where the critical clearing time is plotted against the
transmission line reactance for the case where X, = X , b and for the two different
types of excitation system used. The critical clearing time is used as a measure of
relative stability for the system under the impact of the given fault. Figure 8.13 shows
that for the conditions considered in this study a change in exciter response ratio from
0.5 to 3.0 resulted in a gain of approximately one cycle in critical clearing time.
 Effect of Excitation on Stability 321

"REF t4.9 pu
I

Fig. 8.12 Excitation block diagram for a 3.0 RR alternator-fed SCR excitation system. (@ IEEE. Re-
printed from IEEE Trans., vol. PAS-89, July/Aug. 1970.)

8.3.4 Reactive power demand during system emergencies

A situation frequently encountered during system emergencies is a high reactive
power demand. The capability of modern generators to meet this demand is reduced
by the tendency toward the use of higher generator reactances. Modern exciters with
high ceiling voltage improve the generator capability to meet this demand. It should
be recognized that excitation systems are not usually designed for continuous operation
at ceiling voltage and are usually limited to a few seconds of operation at that level.
Concordia and Brown [ I71 recommend that the reactive-power requirement during sys-
tem emergencies should be determined for a time of from a few minutes to a quarter-
or half-hour and that these requirements should be met by the proper selection of the
generator rating.

8.4 Effect of Excitation on Dynamic Stability

Modern fast excitation systems are usually acknowledged to be beneficial to tran-
sient stability following large impacts by driving the field to ceiling without delay.
However, these fast excitation changes are not necessarily beneficial in damping the
oscillations that follow the first swing, and they sometimes contribute growing oscilla-
tions several seconds after the occurrence of a large disturbance. With proper design
and compensation, however, a fast exciter can be an effective means of enhancing
stability in the dynamic range as well as in the first few cycles after a disturbance.
Since dynamic stability involves the system response to small disturbances, analysis
as a linear system is possible, using the linear generator model derived previously [ 1 I].
For simplicity we analyze the problem of one machine connected to an infinite bus

E6 2L
0 0.2 0.4 0.6 0.8

xeo = Xeb, PU

Fig. 8.13 Transient stability studies resulting from studies of [19]: A = 0.5 RR diode excitation system;
E = 3.0 pu RR SCR excitation system. (Q IEEE. Reprinted from IEEE Trans., vol. PAS-89,
July/Aug. 1970.)
322 Chapter 8

through a transmission line. The synchronous machine equations, for small perturba-
tions about a quiescent operating condition, are given by (the subscript A is omitted
for convenience)
T, = Kl6 + K2E; (8.10)
E; = [K,/(I - [KiK4/(1
K~T;oS)]EFD K~T;oS)]~ (8.1 1)
V, = K56 + K6E; (8.12)
T ~ W S= T,,, - T, (8.13)
where is the direct axis open circuit time constant and the constants K , through K6
depend on the system parameters and on the initial operating condition as defined in
Chapter 6. In previous chapters it was pointed out that this model is a substantial
improvement over the classical model since it accounts for the demagnetizing effects
of the armature reaction through the change in E; due to change in 6.
We now add to the generator model a regulator-excitation system that is repre-
sented as a first-order lag. Thus the change in EFD is related to the change in V,
(again the subscript A is dropped) by
E F D / V ,= -Kc/(I + 7,s) (8.14)
where K , is the regulator gain and T, is the exciter-regulator time constant.
8.4.1 Examination of dynamic stability by Routh’s criterion
To obtain the characteristic equation for the system described by (8. l0)-(8.14),
a procedure similar to that used in Section 3.5 is followed. First, we obtain
r -

From (8.13) for T,,, = 0,

s26 = - ( W ; / T , ) T , = - ( w R / ~ H ) T , (8.16)
By combining (8.15) and (8.16) and rearranging, the following characteristic equa-
tion is obtained:
s4 + as’ + ps2 + ys + 7 = 0 (8.17)
where CY = I/?, + l/K3rAO
p = [(I + K3K6Kc)/K3T;OTr] + KI(WR/2H)
= wR (Ki/Tr
2H
+ Ki/&T;O - K2&/Th)

.=-[ WR
2H
KI(I + K3K6Kr)
K3 4 0 Tf
K2K4

Applying Routh’s criterion to the above system, we establish the array

 Effect of Excitation on Stability 323

where

b2=O c ~ = u ~ = v (8.18)
According to Routh's criterion for stability, the number of changes in sign in the first
column ( I , a, al , b l , and c I ) corresponds to the number of roots of (8.17) with positive
real parts. Therefore, for stability the terms a,a l ,b l , and cI must all be greater than
zero. Thus the following conditions must be satisfied.
1. a = 1/7, + l/K37;0 > 0, and since 7, and ri0 are positive,
dO/T, > - 1/K3 (8.19)
K3 is an impedance factor that is not likely to be negative unless there is an exces-
sive series capacitance in the transmission network. Even then 72,)/T, is usually large
enough to satisfy the above criterion.
2. 01 = p - y/a > 0
,0
+ K3K6Kt K~ z)- K3 4
2 k1( K ~ T ; o T ,) - q]
(1 + 0 1, Tt + K3dO K2K4
K3 T;O 7 , 2H K3T;O + T,

or

(8.20)

This inequality is easily satisfied for all values of constants normally encountered in
power system operation. Note that negative K , is not considered feasible. From
(8.20) K , is limited to values greater than some negative number, a constraint that
is always satisfied in the physical system.

We now recognize the first expression in parentheses in the last term of (8.21) to be
the positive constant CY defined in (8.17). Making this substitution and rearranging
324 Chapter 8

to isolate K , terms, we have

(8.22)

The expressions in parentheses are positive for any load condition. Equation (8.22)
places a maximum value on the gain K, for stable operation.
4. c , = q > o

Since KIK6 - K2K5 > 0 for all physical situations, we have

This condition puts a lower limit on the value of K,.

Example 8.3
For the machine loading of Examples 5.1 and 5.2 and for the values of the con-
stants K, through K6 calculated in Examples 6.6 and 6.7, compute the limitations on the
gain constant K , , using the inequality expressions developed above. Do this for an ex-
citer with time constant 7, = 0.5 s.

Solution
In Table 8.3 the values of the constants K, through K6 are given together with the
maximum value of K, from (8.22)and the minimum value of K, from (8.23). The regu-
lator time constant 7, used is O S s , 7 j 0 = 5.9s, and H = 2.37s. Case I is discussed
in Examples 5 . I and 6.6; Case 2, in Examples 5.2 and 6.5.
From Table 8.3 it is apparent that the generator operating point plays a significant

Table 8.3. Computed Constants for the

Linear Regulated Machine
Constants Case I (Ex. 5.1) Case 2 (Ex. 5.2)

Kl 1.076 I .448
K2 1.258 t.317
K3 0.307 0.307
K4 1.712 1.805
KS -0.041 0.029
K6 0.497 0.526
a 2.552 2.552
K2 K3K47t 0.33 1 0.365
K372cl + Tt 2.313 2.313
K37207t, 0.906 0.906
K2K41aTdo 0.143 0. I58
K45 0.85 1 0.949
a KS7d0 -0.616 0.442
K47237, 5.051 5.325
1/7, 4.000 4.000
Kt
K. <
’ - 2.3
269.0
-3.2
1120.2
 Effect of Excitation o n Stability 325

role in system performance. The loading seems to influence the values of K, and K ,
more than the other constants. At heavier loads the values of these constants change
such that in (8.22) the left side tends to decrease while the right side tends to increase.
This change is in the direction to lower the permissible maximum value of exciter-
regulator gain K,. For the problem under study, the heavier load condition of Case 1
allows a lower limit for K, than that for the less severe Case 2.

Routh’s criterion is a feasible tool to use to find the limits of stable operation in a
physical system. As shown in Example 8.3, the results are dependent upon both the sys-
tem parameters and the initial operating point. The analysis here has been simplified to
omit the rate feedback loop that is normally ar! integral part of excitation systems. Rate
feedback could be included in this analysis, but the resulting equations become compli-
cated to the point that one is almost forced to find an alternate method of analysis.
Computer based methods are available to determine the behavior of such systems and
are recommended for the more complex cases [20, 211.
One special case of the foregoing analysis has been extensively studied [ I I]. This
analysis assumes high regulator gain (K,K,K, >> I ) and low exciter time constant
( 7 , << K 3 ~ j O )I.n this special case certain simplifications are possible. See Problem 8.4.

8.4.2 Furtfrer conridemtionr of the regulator gain and time constant

At no load the angle 6 is zero, and the 6 dependence of (8.10)-(8.23) does not apply.
For this condition we can easily show that the machine terminal voltage V, is the same
as the voltage E:. Changes in this latter voltage follow the changes in EFD with a
time lag equal to 7A0. A block diagram representing the machine terminal voltage
at no load is shown in Figure 8.14. From that figure the transfer function for V,/VREF
can be obtained by inspection.

V,/~RW = K,/l(1 + K , ) + ~ ( 7 +, d o ) + d o 7 , S 2 ] (8.24)

Equation (8.24) can be put in the standard form for second-order systems as

v /VREF = K/(.V* + 2{W,S + W,’) (8.25)

where K = K , / 7 ; 0 7 , , W: = ( 1 + K,)/T&T,, 2 { w , = ( I / T * + 1/7i0).

For good dynamic performance, i.e., for good damping characteristics, a reasonable
value of { is I/a. For typical values of the gains and time constants in fast exciters
we usually have T A ~>> T , and K , >> 1. We can show then that for good performance
K, T ; ~ / ~ T , This
. is usually lower than the value of gain required for steady-state
performance. In [ 1 I ] de Mello and Concordia point out that the same dynamic per-
formance can be obtained with higher values of K , by introducing a lead-lag network
with the proper choice of transfer function. This is left as an exercise (see Problem 8.5).

Fig. 8.14 Block diagram representing the machine terminal voltage at no load.
326 Chapter 8

8.4.3 Effect on the electrical torque

The electrical torque for the linearized system under discussion was developed in
Chapter 3. With use of the linear model, the electrical torque in pu is numerically
equal to the three-phase electrical power in pu. Equation (3.13) gives the change in the
electrical torque for the unregulated machine as a function of the angle 6. The same
relation for the regulated machine is given by (3.40). From (3.13) we compute the
torque as a function of angular frequency to be
TJ6 = K , - [K2K3K4/(I+ w 2 K : ~ 2 ) ] ( -
1 JwK3~i0) (8.26)
The real component in (8.26) is the synchronizing torque component, which is reduced
by the demagnetizing effect of the armature reaction. A t very low frequencies the
synchronizing torque T, is given by
TsZZ Kl - K2KJK4 (8.27)
In the unregulated machine there is positive damping introduced by the armature reaction,
which is given by the imaginary part of (8.26). This corresponds to the coefficient of the
first power of s and is therefore a damping term.
I n the regulated machine we may show the effect of the regulator on the electrical
torque as follows. From (3.40) the change of the electrical torque with respect to the
change in angle is given by

_
Te -
- K 9 - -K2 K4 s + (]/re + K5K/K4Te)

I t can be shown that the effect of the terms K 2 K 4 (1 + 7 , s ) in the numerator is very
small compared to the term K 2 K S K , . This point is discussed in greater detail in [ I I].
Using this simplification, we write the expression for Tc/6as

which at a frequency w can be separated into a real component that gives the synchro-
nizing torque T, and into an imaginary component that gives the damping torque Td.
These components are given by

(8.31)

Note that the damping torque Td will have the same sign as K S . This latter quantity
can be negative at some operating conditions (see Example 6.6). In this case the regula-
tor reduces the inherent system damping.
At very low frequencies (8.30) is approximately given by
T, EZ K I - K 2 K s / K 6 (8.32)
which is higher than the value obtained for the unregulated machine given by (8.27).
 Effect of Excitation on Stability 327

I l c r r “1

Fig. 8.15 Block diagram of a linearized excitation system model.

Therefore, whereas the regulator improves the synchronizing forces in the machine at
low frequencies of oscillation, it reduces the inherent system damping when K 5 is nega-
tive, a common condition for synchronous machines operated near rated load.

8.5 Root-locus Analysis of a Regulated Machine Connected to an Infinite Bus

We have used linear system analysis techniques to study the dynamic response of
one regulated synchronous machine. In Section 7.8, while the exciter is represented
in detail, a very simple model of the generator is used. In Section 8.4 the exciter
model used is a very simple one. In this section a more detailed representation of the
exciter is adopted, along with the simplified linear model of the synchronous machine
that takes into account the field effects. The excitation system model used here is
similar to that in Figure 7.54 except for the omission of the limiter and the saturation
function S E . This model is shown in Figure 8.15. In this figure the function G&) is the
rate feedback signal. The signal V , is the stabilizing signal that can be derived from any
convenient signal and processed through a power system stabilizer network to obtain
the desired phase relations (see Section 8.7).
The system to be studied is that of one machine connected to an infinite bus through
a transmission line. This model used for the synchronous machine is essentially that
given in Figure 6.3 and is based on the linearized equations (8.10)-(8.13). To simulate
the damping effect of the damper windings and other damping torques, a damping
torque component - D w is added to the model as shown in Figure 8.16.
The combined block diagram of the synchronous machine and the exciter is given
in Figure 8.17 (with the subscript A omitted for convenience).

Fig. 8.16 Block diagram of the simplified linear model of a synchronous machine connected to an infinite
bus with damping added.
 -
KR
l + T S
R

Fig. 8.17 Combined block diagram of a linear synchronous machine and exciter.
 Effect of Excitation on Stability 329

KA &
% vt
N(4
]1+T
A
I
+ ‘e) -

I KR
1 + 7 s
R

To study the effect of the different feedback loops, we manipulate the block diagram
so that all the feedback loops “originate” at the same takeoff point. This is done by
standard techniques used in feedback control systems [22]. The common takeoff point
desired is the terminal voltage V , , and feedback loops to be studied are the regulator
and the rate feedback GF(s). The resulting block diagram is shown in Figure 8.18.
I n that figure the transfer function N ( s ) is given by
K3K6(2HS2 + DS + Kim) - mKzK3K5
N(s) = (8.33)
(1 + ~ + Kim) - q K z K &
K ~ T A ~ s ) ( WfS DS
Note that the expression for N ( s ) can be simplified if the damping D is neglected or if
the term containing K, is omitted ( K , is usually very small at heavy load conditions).
The system of Figure 8. I8 is solved by linear system analysis techniques, using the
digital computer. A number of computer programs are available that are capable of
solving very complex linear systems and of displaying the results graphically in several
convenient ways or in tabular forms [20, 211. For a given operating point we can
obtain the loci of roots of the open loop system and the frequency response to a sinu-
soidal input as well as the time response to a small step change in input.
The results of the linear computer analysis are best illustrated by some examples.
In the analysis given in this section, the machine discussed in the examples of Chap-
ters 4,5, and 6 is analyzed for the loading condition of Example 6.7. The exciter data
are K, = 400, = 0.05, KE = -0.17, = 0.95, KR = 1.0 and T~ = 0. The machine
constants are 2H = 4.74 s, D = 2.0 pu and 7A0 = 5.9 s. The constants K I through K6
in pu for the operating point to be analyzed are
KI =: 1.4479 KS = 0.3072 KS = 0.0294
K2 = 1.3174 K4 = 1.8052 K6 = 0.5257

Example 8.4
Use a linear systems analysis program to determine the dynamic response of the
system of Figure 8.18 with and without the rate feedback. The following graphical
solutions are to be obtained for the above operating conditions:
1. Root-locus plot.
2. Time response of VA to a step change in VREF.
3. Bode diagram of the closed loop transfer function.
4. Bode diagram of the open loop transfer function.
Fig. 8.19 Root locus of the system of Figure 8.17: (a) without rate feedback, (b) with rate feedback.

Fig. 8.20 Time response to a step change in V R E F :(a) GF(s)= 0, Ib) GF(s)# 0.

Fig. 8.21 Bode plots of the closed loop transfer function: (a) GF = 0. (b) GF z 0.
 Effect of Excitation on Stability 33 1

Fig. 8.22 Bode plots of the open loop transfer function: (a) GI; = 0.(b) G F + 0.

Compute these graphical displays for two conditions:

(a) GI;(S) = 0
(b) GAS) = sK,/(l + T#), with KF = 0.04, and T~ = 1.0 s
Solution
The results of the computer analysis are shown in Figures 8.19-8.22 for the different
plots. I n each figure, part (a) is for the result without the rate feedback and part (b)
is with the rate feedback.
Figures 8.19--8.20 show clearly that the system is unstable for this value of gain
without the rate feedback. Note the basic problem discussed in Example 7.7. With
G&) = 0, the system dynamic response is dominated by two pairs of complex roots
near the imaginary axis. The pair that causes instability is determined by the field

Table 8.4. Root-Locus Poles and Zeros of Example 8.4

Condition Zeros Poles

(a) K F = 0 -0.21097 + j10.45130 - 0.27324

-0 21097 - j10.45130 -20.00000
- 0. I7894
-0.35020 + j10.72620
-0.35020 - j10.72620
(b) K F = 0.04 - 1.19724 + j0.83244 - 20.00000
- 1.19724 - j0.83244 -0.17894
-0.40337 + j10.69170 -0.27324
-0.40337 - j10.69170 -0.35021 + j10.72620
-0.35021 - j10.72620
- I .ooooo
332 Chapter 8

winding and exciter parameters. The effect of the pair caused by the torque angle loop
is noticeable in the Bode plots of Figures 8.2 1-8.22. These roots occur near the natural
frequency w, = (1.4479 x 377/4.74)'12 = 10.73 rad/s. The rate feedback modifies the
root-locus plot in such a way as to make the system stable even with high amplifier
gains. The poles and zeros obtained from the computer results are given in Table 8.4.

Example 8.5
Repeat part (b) of Example 8.4 with (a) D = 0 and (b) K 5 = 0.
Solution
(a) For the case of D = 0 it is found (from the computer output) that the poles
and zeros affected are only those determined by the torque angle loop. These poles
now become -0.13910 + j10.72550 (instead of -0.35021 i j10.72620). The net effect
is to move the branch of the root locus determined by these poles and zeros to just
slightly away from the imaginary axis.
(b) It has been shown that K 5 is numerically small. Except for the situations where
K 5 becomes negative, its main effect is to change 0, to the value
w2n = ( w R / ~ H ) (-
K ,K z K s / K ~ )
The computer output for K 5 = 0 is essentially the same as that of Example 8.4.
The root-locus plot and the time response to a step change in VREFfor the cases
of D = 0 and K 5 = 0 are displayed in Figures 8.23-8.24.

The examples given i n this section substantiate the conclusions reached in Sec-
tion 7.7 concerning the importance of the rate feedback for a stable operation at high
values of gain. A very significant point to note about the two pairs of complex roots
that dominate the system dynamic response is the nature of the damping associated
with them. The damping coefficient D primarily affects the roots caused by the torque
angle loop at a frequency near the natural frequency w , . The second pair of roots,
determined by the field circuit and exciter parameters, gives a somewhat lower fre-

Fig. 8.23 Root locus of the system of Example 8.5: (a) D = 0, (b) K S = 0.
 Effect of Excitation o n Stability 333

Fig. 8.24 Time response to a step change in VREFfor the system of Example 8.5: (a) D = 0, (b) K S = 0.

quency and its damping is inherently poor. This is an important consideration in the
study of power system stabilizers.
8.6 Approximate System Representation
I n the previous section it is shown that the dynamic system performance is domi-
nated by two pairs of complex roots that are particularly significant at low frequencies.
In this frequency range the system damping is inherently low, and stabilizing signals
are often needed to improve the system damping (Section 8.7). Here we develop an ap-
proximate model for the excitation system that is valid for low frequencies.

it can be neglected at low frequencies (s = j w -

We recognize that the effect of the rate feedback G&) in Figure 8.17 is such that
0) or near steady state (f
We have already pointed out that K S is usually very small and is omitted in this
a). -
approximate model. The feedback path through K4 provides a small positive damping
component that is usually considered negligible [ 1 I]. The resulting reduced system is
composed of two subsystems: one representing the exciter-field effects and the other repre-
senting the inertial effects. These effects contribute the electrical torque components
designated T,, and T,, respectively.
8.6.1 Approximate excitation system representation
The approximate system to be analyzed is shown in Figure 8.25 where the exciter
and the generator have been approximated by simple first-order lags [ 1 I]. A straight-
forward analysis of this system gives

4 Gx 0) -
Fig. 8.25 Approximate representation of the excitation system.
334 Chapter 8

(8.35)

where W , is the undamped natural frequency and {, is the damping ratio:

(8.36)
We are particularly concerned about the system frequency of oscillation as compared
to w , . The damping f, is usually small and the system is poorly damped.
The function G,(s) must be determined either by calculation or by measurement
on the physical system. A proven technique for measurement of the parameters of
G,(s) is to monitor the terminal voltage while injecting a sinusoidal input signal at
the voltage regulator summing junction [8, I 2 , 2 3 , 2 4 , 2 5 ] . The resulting amplitude and
phase (Bode) plot can be used to identify G,(s) in (8.35).
Lacking field test data, we must estimate the parameters of G,(s) by calculations
derived from a given operating condition. It should be emphasized that this procedure
has some serious drawbacks. First, the gains and time constants may not be precisely
known, and the use of estimated values may give results that are suspect [IO, 12,241.
Second, the theoretical model based on the constants K I through K6 is not only load
dependent but is also based on a one machine-infinite bus system. The use of these
constants, then, requires that assumptions be made concerning the proximity of the
machine under study with respect to the rest of the system. A procedure based on
deriving an equivalent infinite bus, connected to the machine under study by a series
impedance, is given in Section 8.6.2.

8.6.2 Estimate of G,(s)

The purpose of this section is to develop an approximate method for estimating
K Ithrough K6 that can be applied to any machine in the system. These constants
can be used in (8.36) to calculate the approximate parameters for G,(s).
The one machine-infinite bus system assumes that the generator under study is
connected to an equivalent infinite bus of voltage Vm/cr through a transmission line
of impedance z, = Re + jX,. This equivalent impedance is assumed to be the The-
venin equivalent impedance as “seen” at the generator terminals. Therefore, if the
driving-point short circuit admittance E.i at the generator terminal node i is known,
we assume that

z, = I / F i (8.37)
The equivalent infinite bus voltage vmis calculated by subtracting the drop &Ze
from the generator terminal voltage E,, where & is the generator current. The pro-
cedure is illustrated by an example.
 Effect of Excitation o n Stability 335

Example 8.6
Compute the constants K , through K6 for generator 2 of Example 2.6, using the
equivalent infinite bus method outlined above. Note that the three-machine system is
certainly not considered to have an infinite bus, and the results might be expected to
differ from those obtained by a more detailed simulation.
Solution
From Example 2.6 the following data for the machine are known (in pu and s).
Xd2 = 0.8958 Xq2 = 0.8645 X42 = 0.0521 H2 = 6.4
= 0.1 198 xi2 = 0.1969 ~ i o 2= 6.0
We can establish the terminal conditions from the load-flow study of Figure 2.19:
1 2 k & = 1,2 + J1,2 = (6- j Q z ) / h
= (1.630 - j0.066)/1.025 = 1.592/-2.339" pu
From Figure 5.6
t a n ( ~ 5-~P~2 ) = X , ~ ~ , ~ - ) 1.272
/ ( xVq 2~ f X 2=
620 - p2 = 51.818"

But from the load flow P2 = 9.280",

620 = 51.818 + 9.280 = 61.098"
Then - P2 + d2 = 54.156"and
-
V2 = V2/& - 820 = 1.025 /-51.818" = V,,
- + jV, = 0.634 - j0.806 pu
12 = f 2 / - ( & 0 - P2 + &) = 1.592 /-54.156" = f 4 2 + jZ, = 0.932 - j1.290 pu
Neglecting the armature resistance, r = 0,
Eqoo = v42 - ~ ~ 2 1 4=2 1.749 PU
E20 = 52 - Xd21dz = 1.789 PU
From Table 2.6 the driving-point admittance at the internal node of generator 2
is given by
-
Y2, = 0.420 - j2.724 pu
The terminal voltage node of generator 2 had been eliminated in the reduction process.
However, since it is connected to the internal node by xi2, z, can be obtained
by using the approximate relation Z , = l/Fz2- jxi2. The exact reduction process
gives

z, = 0.0550 + j0.2388 = 0.2450/77.029" pu

Then we compute from (6.56)

KI = +
1/[RZ (x, + X,)(X; +
X , ) ] = 1/0.39925 = 2.5084
1/K3 = 1 + K , ( X d - X ; ) ( X , + X , ) = 3.1476
K3 = 0.3177
We can compute the infinite bus voltage
 Next Page
336 Chapter 8
- -
v, = V,/a = v2 - ZJ,
= 1.02519.280" - (0.2450 /77.029")( I .592 /6.941")
= 0.9706 - j0.2226 = 0.9958 I- 12.914"
The angle required in the computations to follow is
y = 620- CY = 61.098 - (-12.914) = 74.012"
K, = K,Vm(Eqao[R,siny+ (xi + X,)cosy] + I,o(x, - x;)[(x, + X,)siny - R,cosyll
= 2.4750
Kz = K,{R,E,,o + I,,[RZ + (x, + X,)']l = 3.0941
K4 = VmK,(xd- x;)[(x, + X,)siny - R,cosy] = 2.0265
KS = (K,Vm/V,O)Ix;~qo[ReCOSy - (xq + Xe)sinyI
- x q VdO[(x; + X,)cosy + R,siny]l = 0.0640
K6 = (V,O/KO)[I- K,x;(xq + XP)I - (VO/V,o)K,xqRe = 0.5070
Summary:
K1 = 2.475 KJ = 0.318 KS = 0.064
K2 = 3.094 K4 = 2.027 K6 = 0.507
Note that these constants are in pu on 100-MVA base whereas the machine is a 192-
MVA generator. The constants K, and K2 should be divided by 1.92 to convert to
the machine base.

Example 8.7
The exciter for generator 2 of the three-machine system has the constants K, = 400
and r , = 0.95 s. Compute the parameters of G,(s). For the system natural fre-
quency (see Example 3.4) calculate the excitation control system phase lag. (Here again
we emphasize the need for actual measurement of the system parameters. Lacking
such measurement, a judgment is made as to which parameters should be used. We use
the regulator gain and the exciter time constant. It is judged that the latter is impor-
tant at the low frequencies of interest. This point is a source of some confusion in the
literature. It is sometimes assumed, erroneously, that the regulator time constant is
to be used when the excitation system is represented by one time constant. This is not
valid for low frequencies.)
Solution
From (8.36) we have
w, = d(0.507 x 400)/(6.0 x 0.93) = 5.967 rad/s
(0.95 + 0.318 x 6.0)/(2 x 5.967 x 0.318 x 6.0 x 0.95) = 0.132
5; =

and the excitation system is poorly damped.

From Example 3.4 the dominant frequency of oscillation is approximately 1.4 Hz or
w,,, z 8.8 rad/s. At any frequency the characteristic equation of G,(s) is obtained by
substituting s = j w in the denominator of the first expression in (8.35):
d( jw) = I - 0.028 Iw2 + j0.0443~
At the frequency of interest ( w = 8.8 rad/s) we have
d( jwosc) = - I . I761 + j0.3898
$,ag = tan-' (0.3898/- I . I76 I ) = I6 1.661