Vous êtes sur la page 1sur 5

1

Effect of Synchronous Generator Loading on


Automatic Voltage Regulation
Hemanshu R. Pota
SEIT, UNSW@ADFA
Canberra ACT 2600 Australia
h.pota@adfa.edu.au

Prahlad Tilwalli
Hill Michael Consultants
Melbourne Vic Australia
ptilwalli@hmac.com.au

AbstractThis paper presents two basic results with proof.


One result backs up the industry practice of using the first-order
synchronous machine model for automatic voltage regulator
(AVR) tuning. The other result gives a condition on when the
loading of the generator results in an unstable system with the
AVR tuned for the first-order model. This analysis also implies
that when generator loading destabilises the system the standard
IEEE AVR models are inadequate to guarantee stability for the
loaded generator.
Index Termsautomatic voltage regulator, excitation system,
dynamic stability, linear synchronous machine, power systems.

I. I NTRODUCTION

HE standard industry practice is to tune AVRs using


the first-order synchronous machine model [1][5]. The
effect of loading conditions on the AVR performance has been
discussed over the years [6][8], but research continues into
the AVR tuning methods which use the first-order synchronous
machine model [9]. The first-order model captures the field
coil dynamics without including the effect of induced emf due
to stator coil currents.
To analyse the effect of generator loading on the stability
of the system with an AVR tuned for the first-order model [9],
a third-order linearised transient synchronous machine model
for a single-machine infinite-bus system is used [5]. The
governing conditions on the effect of loading derived in this
paper have an interpretation in terms of the generator loadangle thus making the results useful for a general multimachine system. These ideas have a significant bearing on
the stability and operation of power systems and owing to
this they have been discussed for decades [10]. The intention
behind the presentation of the results in this paper is to put
the results in a form which enables an easy transition to the
design of filters or simple control blocks to provide stability
to the system with an AVR in spite of the variation in the
loading of the generator.
Since we analyse the effect of varying load conditions on
the system stability, for a fixed AVR, it is useful to look at
the change in the dynamcial properties of the synchronous
machine as the loading varies. The change in the frequency
response of the third-order generator model for varying load
conditions is analysed next. We look at the change in the low
frequency gain, pole and zero locations, and the high frequency
gain, for varying load conditions.

978-1-4799-1303-9/13/$31.00 2013 IEEE

Md. Jahangir Hossain


School of Engg, Griffith University
Gold Coast Qld Australia
j.hossain@griffith.edu.au

II. S YNCHRONOUS M ACHINE M ODEL


For transient stability studies the synchronous machine
is modelled using a second-order equation representing the
mechanical dynamics and a first-order equation modelling the
RL field coil [11],

E q

= 0 0


1
Pm Eq Iq + xd xq Iq Id
=
2H

1
=
Efd Eq (xd xd ) Id

do

(1)
(2)
(3)

The algebraic equations for this system are (re = 0):


Eq V cos
xd + xe
V sin
Iq =
xq + xe

Id =

(4)
(5)

Vq = Eq xd Id

(6)

xq Iq

(7)

Vd =

q
Vt = Vq2 + Vd2
q
= (xq Iq )2 + (Eq xd Id )2

(8)

A linear model is obtained by linearising equations (1)


(3), about an equilibrium point, and the resulting model for
a single-machine infinite-bus system is commonly represented
by a block diagram as shown in Figure 1. The transfer function
for the synchronous machine model in Figure 1 (with Tm =
0) is:
K2 K3 0
(s)
=
Ef d (s)
(1 + sT3 ) (2Hs2 + KD s + K1 0 ) K2 K3 K4 0
(9)
The K-parameters in the block diagram are given here for
easy reference.

2
cos 2 0
Eq0 V cos 0 (xq xd ) Eq0 V cos 0 V
+
,
K1 =
xq + xe
(xd + xe )(xq + xe )
K2 =

V sin 0
,
xd + xe

T3 = do
K3 ,

K4 =

K3 =

xd + xe
,
xe + xd

xd xd
V sin 0
xd + xe

K4

Efd
+

Field Circuit
Eq
K3
1 + sT3

K2

+
Te

Tm
1
r
2Hs + KD

K6
Vt
Fig. 1.

0
s

K1

K5

Synchronous Machine Linear Model

Claim 1: The damping of the complex pole of the system


in (9) is proportinal to K2 K3 K4 and it approaches zero for
generator with zero load-angle.
Proof: The poles of the transfer function (9) are the roots
of the following polynomial:
s3 +

0 (K1 K2 K3 K4 )
1 2 K 1 0
s +
= 0.
s+
T3
2H
2HT3

A. Terminal Voltage as Feedback Output


The transfer function between the input field voltage and
the terminal voltage output for the block diagram in Figure 1
can be written as:


K3 K6 2Hs2 + KD s + K1 0 K2 K5 0
Vt (s)
=
Ef d (s)
(1 + sT3 ) (2Hs2 + KD s + K1 0 ) K2 K3 K4 0
(10)

Let the roots of the above polynomial be: p , p p .


Then we can write from the coefficients of the polynomial:

Small changes in the terminal voltage Vt in (8) are expressed


as a linear function of (s) and Eq :


1
0
p + 2p =
, p 2p + p2 =
(K1 K2 K3 K4 ) ,
T3
2HT3

Vt (s) = K5 (s) + K6 Eq (s).

and

2p p + p2 =

K 1 0
.
2H

From the above relations it can be seen that the damping term
2p = T13 p , i.e., the distance of the real root from T13
is an indicator of the system damping. This distance can be
estimated by substituting s = T13 in the above polynomial
and checking how far the result is from zero. With this
substitution we get:
 3  3
0 (K1 K2 K3 K4 )
K 1 0
1
1
+
+

T3
T3
2HT3
2HT3
K2 K3 K4 0
=
.
2HT3
The above relation shows that the larger the term K2 K3 K4 ,
the more is the distance between the real root p and
T13 , thus the larger is the damping. Of course the necessary
condition K1 K2 K3 K4 > 0 should always be satisfied for
dynamic stability. The parameter K3 doesnt change with the
loading of the generator, parameters K2 and K4 increase in
proportion to the loading, but the parameter K1 decreases with
loading. Finally since K2 and K4 are zero for zero load-angle,
there is no damping in the system for that operating condition.
In short it can be concluded that the natural damping of
the generator is the least when it is lightly loaded and
electrical loading provides damping to the electromechanical
oscillations.

(11)

where


xe xq + xd + 2xq xd sin 2 0
2
2
2 xq + xe (xd + xe )
!

xe Eq0 xe + xd V cos 0
Eq0 xd sin 0
, K6 =

2
2
(xd + xe )
Vt0 (xd + xe )
xq xd

V xe
K5 =
Vt0

III. S YNCHRONOUS M ACHINE M ODEL FOR AVR T UNING


The block diagram of a synchronous machine with an AVR
is shown in Figure 2.
Vref
+

Fig. 2.

AVR

Efd

Generator

Vt

Block Diagram of a Synchronous Machine with an AVR

For AVR design synchronous generator is modelled as


a first-order system [9]. For an unloaded generator Eq is
equal to the terminal voltage and thus equation (3) (with
Vt = Eq ) describes the dynamics between the field and
terminal voltages. Most of the exciter design is done for
unloaded generator. Setting Id = 0 in (3), we get the following
transfer function for the synchronous machine:
1
Vt (s)
=

Efd (s)
1 + sTd0

(12)

Although the first-order machine model (12) is used for AVR


tuning, it proves to be quite robust even when the machine

is loaded. In the previous section it has been shown that the


natural damping of the generator is the least when the loadangle 0 is equal to zero.
In comparison with the first-order model in (12), the general
transfer function between the reference voltage and terminal
voltage, for the block diagram in Figure 1 is:


K3 K6 2Hs2 + K1 0 K2 K5 0
Vt (s)
.
=
Ef d (s)
(1 + sT3 ) (2Hs2 + K1 0 ) K2 K3 K4 0
(13)
Figures 3 and 4 show frequency responses of a generator,
using the transfer function given in (13), under two different
load conditions. For the sake of comparison the frequency
response of the first-order the model in (12), for the same
generators, is also shown in Figures 3 and 4. A few points need
to be carefully observed about the two frequency responses in
Figures 3 and 4.
Bode Diagram

Magnitude (dB)

100

200

300

Phase (deg)

400
0

90

180

270
3

10

Fig. 3.

10

Frequency Response

10
10
Frequency (rad/sec)

Vt (s)
Vref (s)

10

10

(1 pu Load and No-load)

Bode Diagram

Magnitude (dB)

50

100

150

Let the poles of the above transfer functions be at: p ,


p p . Then we can write from the coefficients of the
denominator of the transfer function:

1
0
p + 2p =
, p 2p + p2 =
(K1 K2 K3 K4 ) ,
T3
2HT3

200
90

Phase (deg)

45
0
45
90

and

135
3

10

Fig. 4.

a higher gain at low frequencies thus predicting a smaller


steady-state error with the automatic voltage regulation.
2) From (13), at higher frequencies the loaded machine
3 K6
response can be approximated by KsT
and that of the
3
K3
unloaded machine as sT3 . In most cases K6 is likely
to be less than one thus, for the same AVR, a loaded
generator should have a higher gain-margin than an
unloaded generator.
3) The most important difference between the frequency
responses of the unloaded and loaded generators is the appearance of a complex pole-zero pair. It can be seen that
owing to the pair, the phase and magnitude contribution
due to the pair appears only near the electromechanical
frequency of oscillation. To enable an independent design
of the excitation system and the power system stabiliser
the compensation due to AVR should not alter the frequency response in the range of this complex pole-zero
pair.
Moreover, the complex pole-zero pair in the loaded synchronous machines, under different loading conditions, as
can be seen from frequency responses in Figures 3 and 4
appears in two different ways. In Figure 3 a complex
pole appears at a frequency lower than the complex zero
frequency, and the other way round in Figure 4. When
a complex zero precedes the complex pole, the phase
goes from 90 to 90 and then back to 90 . In this
case an AVR tuned for an unloaded generator will work
well for the loaded synchronous generator. But when the
complex pole precedes the complex zero then the phase
goes from 90 to 270 and then back to 90 . For
this situation an AVR tuned for no-load conditions will
make the system unstable when the generator is loaded.
The next claim gives an easy condition to check if the complex
zero precedes the complex pole in the transfer function (13).
Claim 2: If K5 > 0 then z < p , where z and p are
the complex zero and pole of the transfer function in (13),
respectively.
Proof: It can easily be seen from (13) that the complex
zero of the loaded generator transfer function is at:
r
0
z =
(K1 K6 K2 K5 ).
2HK6

10

Frequency Response

10
10
Frequency (rad/sec)

Vt (s)
Vref (s)

10

10

(0.25 pu Load and No-load)

1) From the transfer function in (13), the low frequency


3 (K1 K6 K2 K5 )
gain of the the loaded machine is K
K6 (K1 K2 K3 K4 ) , and
the unloaded machine has a gain of one. For most load
conditions with K5 > 0 the unloaded generator will have

K 1 0
.
2H
The damping p is normally small and it can be seen that
p is close to T13 which is also a small number thus we can
assume that
r
K 1 0
.
p
2H
We can now see that z < p if K2 K5 > 0. Normally K2
is greater than zero so we come to the conclusion that the
2p p + p2 =

AVR tuned for an unloaded generator will not cause unstable


closed-loop operation for a loaded generator if K5 > 0.

Closedloop Loaded Synchronous Machine


1.4

1.2

The variation of K5 as a function of the load-angle for a


typical generator is shown in Figure 5 (assuming a fixed Eq0 ).
From Figure 5 it can be seen that for load-angle variation from
0 to 30 the value of K5 is increases and then as the loadangle increases the value of K5 decreases. For load-angles
ablove 53 K5 is negative and thus for load-angle above 53 ,
IEEE recommended AVR blocks may not be able to stabilise
the system.

Amplitude

0.8

0.6

0.4

0.2

Fig. 6.

10
Time (sec)

12

14

16

18

20

Step Response with AVR IEEE-AC7B (0.25 pu Load)

Closedloop Loaded Synchronous Machine

12

x 10
1.5

0.06
0.05

0.04
0.5
Amplitude

0.03

0.02

0.01
0.5

0
0.01

0.02
1.5

0.03

10

20

30
40
Load Angle (deg)

50

60

500

1000

1500

2000
2500
Time (sec)

3000

3500

4000

4500

Fig. 7.
Fig. 5.

Step Response with AVR IEEE-AC7B (1 pu Load)

Variation of K5 with load angle

a) Simulation Parameters: The simulation parameters


and a matlab script for the tuning of the IEEE-AC7B AVR
can be requested from h.pota@adfa.edu.au.
IV. C ONCLUSION
For the loaded generator with frequency response shown
in Figure 3, K5 = 0.1126 pu, and for Figure 4, K5 =
0.1114 pu. An IEEE-AC7B [1] AVR is tuned using the unloaded model first. There are six parameters, KIR , KPR , KPA ,
KIA , KDR , and TDR , which need tuning for the IEEE-AC7B
AVR [1]. In the design for this paper, these six parameters
are tuned to provide a crossover frequency of 5 Hz and a
phase-margin of 65 [12]. The AVR so tuned for an unloaded
generator is used to obtain the step responses for two load
conditions. For one load condition K5 = 0.1114 pu and for
another it is K5 = 0.1126 pu; the responses for the two
load conditions with a positive and a negative value of K5 are
shown in Figures 6 and 7, respectively. It can be seen from the
figures that an AVR designed using the first-order synchronous
machine model leads to an unstable closed-loop system with a
generator loading of 1 pu (K5 = 0.1126 pu) but the closedloop system is stable for 0.25 pu (K5 = 0.1114 pu) loading.

There is a good basis for using the first-order synchronous


machine model for AVR tuning and an evaluation of the
parameter K5 sets the conditions under which novel AVRs
are needed to stabilise the closed-loop system.
R EFERENCES
[1] IEEE Std 421.5-2005 (Revision of IEEE Std 421.5-1992). IEEE
Recommended Practice for Excitation System Models for Power System
Stability Studies, 2006.
[2] IEEE Std 421.1-2007 (Revision of IEEE Std 421.1-1986). IEEE
Standard Definitions for Excitation Systems for Synchronous Machines,
15 2007.
[3] IEEE Std 421.2-1990. IEEE Guide for Identification, Testing, and
Evaluation of the Dynamic Performance of Excitation Control Systems,
1990.
[4] IEEE Commitee Report. Excitation system dynamic characteristics.
Power Apparatus and Systems, IEEE Transactions on, PAS-92(1):64
75, Jan. 1973.
[5] P. Kundur. Power system stability and control. McGraw-Hill, Inc., New
York, 1994.

[6] C. Raczkowski. Complex root compensator a new concept for dynamic


stability improvement. Power Apparatus and Systems, IEEE Transactions on, PAS-93(6):18421848, Nov. 1974.
[7] K. Bollinger and R. Lalonde. Tuning synchronous generator voltage
regulators using on-line generator models. Power Apparatus and
Systems, IEEE Transactions on, 96(1):3237, Jan. 1977.
[8] A. Barakat, S. Tnani, G. Champenois, and E. Mouni. Monovariable
and multivariable voltage regulator design for a synchronous generator
modeled with fixed and variable loads. Energy Conversion, IEEE
Transactions on, 26(3):811 821, Sept. 2011.
[9] Kiyong Kim, P. Rao, and J.A. Burnworth. Self-tuning of the PID
controller for a digital excitation control system. Industry Applications,
IEEE Transactions on, 46(4):1518 1524, July-Aug. 2010.
[10] Francisco P. Demello and Charles Concordia. Concepts of synchronous
machine stability as affected by excitation system. IEEE Trans. on Power
Apparatus and Systems, PAS-88(4):316329, April 1969.
[11] M. A. Pai, D. P. Sen Gupta, and K. R. Padiyar. Small Signal Analysis
of Power Systems. Narosa Publishing House, New Delhi, 2004. ISBN
81-7319-594-3.
[12] Rodolfo J. Koessler. Techniques for tuning excitation system parameters.
IEEE Transactions on Energy Conversion, 3(4):785791, December
1988.

Vous aimerez peut-être aussi