Model Development and Validation of Brushless

Exciters
Tony Bertes
DIgSILENT Pacific
Melbourne, Australia 3004
Email: [email protected]
AbstractThis paper describes the process of performing site
tests and model development of brushless exciters. Brushless
exciters are often a source for model inaccuracy when studying
synchronous machines and excitation system, which may lead to
incorrect stabiliser design and poor performance. The process described here involves on site testing of the excitation system both
with the generating unit operating open circuit and synchronised
to the network which will enable the extraction of parameters to
fit a model for the brushless exciter based on the IEEE Std 421.5.
Two case studies shall be presented to contrast the difference in
approach given the different limitations faced at each site.
KeywordsBrushless exciter, model validation, excitation system, AVR, synchronous machine, stability, field testing.

Therefore when the connection study of a new AVR is

performed, the interaction of the new AVR model with an
existing exciter model that is likely to be incorrect is interesting
and frequently yields conflicting results in practice leading to
repeat tests, repeat PSS design and downtime of the generator
(both physically and commercially).

I NTRODUCTION

Excitation systems play an important part in determining

power system stability when analysing the performance of a
power station or a meshed network. Accurate models of the
excitation system are required for power system planning,
optimal tuning of control systems such as power system
stabilisers (PSS), and stability analysis.
Block diagrams, when supplied, often are based on typical
or default values, and do not represent the behaviour of the
exciter machine which is essentially an inside-out three phase
synchronous machine.
The excitation system consists of the following elements:

Automatic voltage regulator (AVR);

PSS (often equipped with the AVR particularly in

digital systems);

Voltage and current sensing;

Exciter (static or rotating); and

Protection and monitoring.

For a number of years, manufacturers of AVRs often provided IEEE models for use in power stability studies. However
there has been a gradual movement towards detailed models for
each type of AVR that is on the market. This has largely been
assisted by the improvement of control system implementation
in digital based hardware which enables greater transparency
of how the regulator is programmed and behaves under various
conditions.
Unfortunately this is not the case with exciter machines.
The inaccuracies of exciter models are often amplified during
a refurbishment project. For example, a generator (asset owner)

II.

T HE B RUSHLESS E XCITER

By definition, a brushless exciter is an alternator-rectifier

exciter employing rotating rectifiers with a direct connection
to the synchronous machine field, thus eliminating the need
for brushes [2]. It is essentially an inside-out three-phase
synchronous generator, the field winding of which is mounted
on the stator housing, the three-phase windings being attached
to the rotor. The three phase output voltage is rectified by
diodes most commonly mounted on the rotating shaft and
applied directly to the main generator field winding. Note that
it is also possible to have an ac exciter with stationary diodes.
In this case, slip rings are rings are required and it becomes
possible to directly measure the main field quantities.
Figure 1 shows the commonly adopted simplified representation
of an Machine
ac rotatingand
exciter
as published
2.3.
Exciter
Mainsystem
Generator
Model by the IEEE

[1]. The voltage applied to the field winding of the exciter is

represented by eFD whereas eFD is the voltage applied to the
The
dynamical
simulation
for the
machine
shown
in Figure
main
generator
field model
winding
onexciter
the right
of the
model
[5].12 is also based
421.5 (part of the excitation system model AC9C).
eFD

1
sTE

vE
fEX

VE,min

iFD

v X = v E SE (v E )

fEX = f (i N )

KE

K i'
i N = C FD
vE

vX

IN

i'FD

KD

Figure 1.

e'FD
Renormalization

I.

may wish to upgrade an analogue AVR control system to a new

digital based system, but the rotating exciter remains unaltered.
The model for the AVR supplied by the manufacturer may
be well documented, but the exciter model is likely to be
inaccurate. Data for the exciter machine is often carried over
from its original installation (often 20 to 30 years old) and its
validity not questioned, nor has the data ever been verified.

AC exciter model based

IEEE Standard
Figureon
12the
Exciter
machine 421.5
and main machine model.

Transfer functions of the exciter machine model:

GEM ,11(s ) =

e'FD (s )
KVE
=
,
eFD (s ) K E + KVx + sTE

KVx =

v x
,
v E

KVE =

i N fEX (fEX v E )
v E i N fEX

Exciter models are typically assumed to be represented by

a first order model. This philosophy is adopted from the IEEE
Standard 421.5 [1]. This is an initial assumption made in this
paper but is tested later in both the time and frequency domain.
The assumption that a first order model is adequate as opposed
to second order or higher models has been tested in many
studies. Industry experience shows that in the vast majority
of cases the compromise between complexity and accuracy
provides acceptable results.
As with the synchronous generator, the exciter is also
affected by saturation. As the exciter has an output voltage
designed to allow field forcing, the effects of saturation at
normal loading (steady state) are quite modest. The saturation
effect is represented from the output voltage of the exciter,
VE, and the saturation function SE(eFD ).

Parameter KC is set appropriately to represent this behaviour.

The IEEE recommended transfer function for FEX is defined
in Figure 2.

Figure 2.

Characteristics of the non linear transfer function FEX

A. Exciter Field Current

C. Comparison of Exciters

The exciter model provided in Figure 1 does not provide

an accurate signal for exciter field current. In most cases, the
signal iFD (also called Vfe) which is the negative input to the
summing junction before the exciter integrator, is assumed to
be exciter field current. This signal tends to show a slower
response as it includes the armature reaction demagnetizing
effect (KD) [4].

The two most common types of excitation methods are (a)

the brushless exciter (also referred to as a rotating exciter),
and (b) the static exciter. The two excitation methods are
shown graphically in Figure 3.

Dynamically, the exciter field current is affected by the time

constant of the exciter, TE, and the saturation effect from the
path (SE + KE). In steady state, this result allows for exciter
field voltage and current to be at the same level in per unit. The
saturation can affect the overall exciter time constant, as when
saturation occurs the inductance in the exciter reduces, and
the time constant abides by the L/R relationship. Therefore,
an accurate model of saturation is required for this method.
For the purposes of this paper, an alternative derivation
was assumed. The signal required for exciter field current was
exciter field voltage (denoted eFD ), and with this signal, a
parallel transfer function is created as follows:

iFD = eFD

1
TE
1 + s. KE+SE

(1)

~
SM

Rotating exciter

B. Rectifier Model
The rectifier conduction introduces non-linear effects, depending on the field current (iFD ) and the applied voltage, VE.
These non-linear effects have been calculated and represented
in a three step linearised function, FEX . This function is a
simplification of a non-linear response but experience with the
widely used approximation has been such that it is seldom
modelled to a higher level of complexity. It is also not
uncommon to have KC set to zero, which introduces further
approximations into the model.
All ac sources that supply rectifier circuits have an internal
impedance that is predominately inductive. This impedance has
an effect in the commutation process and produces a non linear
characteristic as the rectifier load current and voltage varies.

SM

Static excitation

Figure 3. Comparison of the two main exciter types (a) the brushless exciter,
and (b) the static exciter

Table I provides a brief comparison of the two main

excitation system types. Performance aside, the choice of
excitation system type often depends on economic factors
(maintenance and cost of equipment).
Table I.

Where iFD is the exciter field current (iFD is the main

machine field current).

C OMPARISON OF E XCITATION S YSTEM T YPES

Brushless Excitation

Static Excitation

Positive field forcing capability only

Positive and negative field forcing capability

Fast excitation response time

Excitation response limited by exciter

time constant TE
No slip rings required (less maintenance)
Direct measurement of rotor quantities
not possible (unless special instrumentation provided
Excitation of 1 to 200A
Additional modelling required

Slip rings required (maintenance required)

Direct measurement of rotor quantities
possible
Excitation of 100 to 10,000A
Cost of converter bridge quite high

From a power system stability viewpoint, the brushless

exciter is often considered more complicated for control and
PSS design. For example, when analysing the transfer function
from speed to torque through the PSS control path, the exciter
adds considerable amount of phase lag that needs to be
compensated for. This transfer function, shown in Figure 4,
reflects the phase and gain relationship between the component

of torque produced by the action of the PSS and the generator

speed oscillations, with no PSS phase compensation.
0
Brushless Exciter
Static Exciter

-40

-60

C ASE S TUDY A - G AS T URBINE G ENERATOR

The first case study involves a 79.4 MVA aero derivative

gas turbine. Initially, DIgSILENT performed routine testing of
the excitation system in 2011 to evaluate compliance with the
National Electricity Rules (NER) and the Generators Performance Standard (GPS). Tests showed non-compliance with the
unsynchronised and synchronised settling time requirement,
and the registered model did not adequately represent the
actual plant response.

-80

In order to rectify the non-compliance issues, the AVR and

PSS would need to be re-designed to improve the performance
of the excitation system. However, to achieve a new design, an
accurate model of the excitation system would be required. For
this reason, DIgSILENT performed Parameter Identification
tests to optimise the existing exciter model that accurately
represents the plant.

-100

-120

-140

-160

-180

-200
-1
10

Figure 6 shows an example of a voltage step response, with

the actual plant response compared to the model prior to the
model validation.

10
Frequency (Hz)

10

DIgSILENT

Uncompensated Phase Response (Degrees)

-20

III.

1.022

Figure 4. Uncompensated phase response of the transfer function Te/ of

a generating unit with (a) brushless excitation system, and (b) static excitation
system

We can see that from the phase response shown in Figure 4,

the difference in the area of interest (1-3Hz) is over 100
degrees, primarily due to the inductance of the exciter machine
(TE). The objective of selecting the PSS phase compensation
is to introduce the necessary phase shift in the PSS control path
to compensate this transfer function to have a phase of nearly
zero degrees throughout the range of frequencies of interest
(0.1 to 3Hz) [3].

DIgSILENT

In addition, the time domain response is also significantly

slower when a rotating exciter is present, meaning it is often
difficult to achieve settling times compliant with the Rules
of the local Grid Authority. Figure 5 shows the time domain
response of terminal voltage to an unsynchronised voltage step
comparing a generator with (a) brushless excitation and (b) a
static excitation.
1.060

1.008

1.002

0.995

0.988
10.00

14.40
Measured: Measured Terminal Voltage [p.u.]
G6: Terminal Voltage in p.u.
Measured: Upper Band
Measured: Lower Band

18.80

23.20

27.60

[s]

32.00

Figure 6. Synchronised terminal voltage response to a 2.5% step applied to

the AVR of gas turbine, overlayed with the initial registered model

A. Field Testing
At this particular installation, the excitation system comprised of:

1.045

1.030

1.015

1.000

0.985
0.000

1.015

3.200
sym_30451_1: Static - Terminal Voltage
sym_30451_1: Brushless - Terminal Voltage

6.400

9.600

12.80

[s]

16.00

Figure 5. Unsynchronised voltage step response of a generating unit with

(a) brushless excitation system, and (b) static excitation system

A dual channel ABB UNITROL F digital AVR;

A dual input channel PSS conforming to the IEEE

PSS2B type;

Potential source rectifier (shunt supply);

A brushless ac exciter supplied by Brush Electric.

Access to the main generator field was not possible, and as

such signals such as main generator field voltage and current,
and rotor angle were not able to be recorded during on site
tests. Due to the limited access of these signals, testing was
limited to:

Voltage step responses (existing data from 2011 reused);

Transfer function measurement of the AVR; and

C. PowerFactory Parameter Identification

A load rejection to derive the generators d-axis parameters.

As tests were limited due to lack of available primary

signals for validating the exciter, a series of Parameter Identification simulations were performed using the existing data
recorded during the compliance tests. Model Parameter Identification is a tool in PowerFactory that optimises parameters
of power system elements when compared to input and output
measurement signals. Figure 9 shows the process of identification.

As the interactions between the AVR, exciter and generator

are meshed and complicated, there was a need to isolate the
source of error between the AVR and the exciter and generator.
Note that because the output of the exciter could not be
recorded, the error of any potential exciter model developed
is coupled with that of the generator. This is highlighted in
Figure 7.

Output

Input
Set Point

Error
+
-

Vr

Controller
(Regulator)

Efd

Terminal Voltage

Plant
(Generator)

Power Amplifier
(Exciter)

Terminal Voltage

Feedback Element

Figure 7.

Overview of the model for the excitation system for Case A

The frequency response of the voltage regulator alone

at standstill (summing junction input to AVR output) was
measured to confirm the model of the AVR. Figure 8 shows the
measured transfer function of the regulator compared with the
model. The alignment is such that it demonstrates the source
of error is within the exciter and/or synchronous machine.

Figure 9.

The PowerFactory Parameter Identification process

550

Figure 10 shows the composite model used for the Parameter Identification simulations. A window function was used
to ensure the Parameter Identification calculation commenced
only when a transient event occurs.

500
450

Gain

400
350
300
250
200
150
100
1

10

Frequency (Hz)
Simulated AVR Gain

Ch1 Measured AVR Gain

DIgSILENT

0.1

Ident: Parameter Identification Frame

Ch2 Measured AVR Gain

40
35
Compare Block

30
25
Measured Data

Phase

20
15

in1mea

Sig
*

10

in2mea

0
-5
-10
0.1

in1sim

sym Slot
ElmSym*

10

Compare
ElmComp*

Frequency (Hz)
Simulated Data

Simulated AVR Phase Angle

Ch1 Measured AVR Phase Angle

in2sim

Ch2 Measured AVR Phase Angle

avr Slot
ElmAvr*,ElmVc..
4

qzpf

Figure 8. Measured transfer function of the voltage regulator compared with

the model

window
*
Window Function

B. Exciter saturation
The exciter saturation function SE(eFD ) is defined as
a multiplier of pu exciter output voltage to represent the
increase in exciter excitation requirements due to saturation.
To accurately derive the exciter saturation, the exciter output
voltage needs to be measured and plotted against exciter field
current. Unfortunately, the exciter output voltage could not be
measured and the exciter saturation could not be derived from
site testing. Typical data has been assumed from similar sized
Brush machines that allows for a small amount of saturation
when the exciter is loaded.

Figure 10.

Composite frame for Parameter Identification simulations

This identification is principally performed in the following

way:
1)
2)

A measurement file object is created which maps the

raw measured data onto one or more signals.
The measurement signals are used as inputs by the
signals which are used as inputs by the models of

3)

4)

5)

the power system elements for which one or more

parameters have to be identified.
The output signals of the power system elements
are fed into a comparator, just as the corresponding
measured signals. The comparator is thus given the
measured response on the excitation and the simulated response of the element models.
The comparator calculates an objective function,
which is the weighted sum of the differences between
the measured and the simulated response, raised to a
power.
The identification command will collect all objective
functions from all comparator objects in the currently
active study case and will minimize the resulting
overall objective function.

Five primary scenarios were considered for Parameter

Identification simulations. These were:
1)
2)
3)
4)
5)

Full load, under excited;

Full load, over excited;
Medium load, unity power factor;
Low load, under excited; and
Low load, over excited.

By considering the chosen range of operating scenarios,

the Parameter Identification simulation optimises the model
to reflect the variation in generator dynamics. The variation
in dynamics is affected by loading and temperature. The
operating point at medium load and unity power factor was
chosen with the exciter in mind. When the machine is operated
at its MCR, an excitation current is present which is called
the nominal excitation current Ifn. If a vector is drawn from
this point to the -Ug2 /Xd point expressed in per unit on the
reactive power axis (assume terminal voltage is at 1 per unit),
the magnitude of this vector is proportional to field current.
The load of the exciter is increased or decreased with operation
on this vector. Therefore, points 3, 4, and 5 cover the loading
of the exciter.

DIgSILENT

Figure 11 shows the measured terminal voltage response to

a 2.5% step applied to the AVR, overlayed with the validated
model response.
1.040

1.033

C ASE S TUDY B - H YDRO GENERATING UNIT

The second case study involves a 100 MVA hydroelectric

generating unit. The asset owner is intending to upgrade the
excitation system controller. This will trigger a requirement
under the NER to provide a validated model of the excitation
system, including the exciter. The determination of settings
for the excitation controller and PSS is dependent on the
characteristics of the rotating exciter. DIgSILENT and the
asset owner considered it prudent to first develop and validate
the exciter model before undertaking the analysis to set the
controller.
As opposed to Case A, there was no existing model of
the exciter and therefore DIgSILENTs role was to propose a
model for the rotating exciter.
The excitation system comprised of:

A single channel ABB UNITROL M analogue AVR;

A dual input channel PSS;

Potential source rectifier (shunt supply);

An ASEA brushless ac exciter.

A. Field Testing and Model Development

Unlike in Case A, access to the rotor was available, and
therefore the main generator field voltage was able to be
measured. Direct measurement of the generator field current
was not possible due to the rotating rectifiers, and a signal
for rotor angle was not available. The test methodology was
therefore extensive, and included:

Resistance measurements of the generator stator, rotor

and exciter field;

Transfer function measurement of the exciter and

synchronous generator;

Transfer function measurement of the AVR;

Measurement of the generators open circuit and short

circuit characteristics;

Voltage step responses;

Load rejection tests for d-axis parameters and units

inertia.

Due to the extensive testing, the process of model development did not rely on Parameter Identification, but rather on
deriving values for the exciter parameters from first principles.
The model was therefore developed and verified in the time
and frequency domain using the results from site.

1.026

1.019

1.012

1.005
12.00

IV.

14.00
sym_30451_1: Terminal Voltage in p.u.
Signal Model: Measured Terminal Votage
Signal Model: UgLower
Signal Model: UgUpper

16.00

18.00

20.00

[s]

22.00

Figure 11. Synchronised terminal voltage response to a 2.5% step applied

to the AVR of gas turbine, overlayed with validated exciter model

With the availability of the mentioned signals, each component of the excitation system could be theoretically be tested
and verified independently. In this case, the AVR is planned
to be replaced by a new digital based system and could be
considered as a black box as a model for the existing AVR
is not required to be submitted as part of the connection of
the planned alteration. However, for closed loop simulations a
reasonably accurate model is required. To verify the adequacy
of the model for the regulator, a transfer function measurement
was made. Figure 12 shows the frequency response of the AVR

10
5
0
-5
Gain (dB)

model compared with the measured transfer function, prior to

the gain introduced by the converter. The agreement is suitable
to proceed with the model development of the exciter.
14

Gain (dB)

-15
-20

Simulated
Measured

13

-25

12

-30

11

-35
-2
10

-1

10

10

10

10
9
20

8
7

Measured
Simulated

6
-1
10

10

Phase Angle

10

-20
-40
-60

-5

Phase Angle

-10

-80

-10

-100
-2
10

-1

10

10

10

Frequency (Hz)

-15
-20

Figure 13. Measured transfer function of the exciter compared with the model

-25
-30
-1
10

Figure 12.
the model

10
Frequency (Hz)

10

Measured transfer function of the analogue AVR compared with

The frequency response of the model shows good agreement with the measured transfer function, which suggests that
the developed model is suitable for use in large scale stability
studies, including the design of a stabiliser.

B. Resistance measurements
The resistance measurements of the generator stator, rotor
and exciter field winding provide valuable information to be
able to assess the steady state performance of the model. As
the measurements were taken at standstill, the temperature of
the winding under test must be considered to be able to correct
to rated temperature.
For all three tests, the resistance was measured three
times, both in the forward and reverse direction. For the rotor
resistance, RFD the per unitisation occurs as follows:
ZF Dbase =

RF D =

3phV Abase
i02
F Dbase

M easuredResistance
ZF Dbase

(2)

(3)

The rotor resistance is of particular importance particularly when deriving steady state generator field current from
measurements of field voltage (from Ohms law), and also
understanding the affect it has on the generators (and exciters)
time constant. It is also of significance when determining the
saturation characteristic of the exciter and the synchronous
generator.
C. Frequency response
The frequency response of the exciter was measured during
unsynchronised operation. The unit was operated below 0.8 pu
terminal voltage to minimize the effects of saturation in the
exciter or in the main machine during the test. Figure 13 show
the measured and corresponding simulated frequency response
of the ac exciter. The simulated response was calculated using
the validated settings, and corrected to test temperature.

D. Steady state performance

The basis of the per unit system of the exciter is to have
the required excitation to produce 1 pu terminal voltage on the
air gap line. This per unit system is universally used in power
system stability studies as it offers considerable simplicity [4].
This is achieved by ensuring that parameters KE and KD are
carefully selected to achieve this condition.
To evaluate the steady state performance of the exciter,
it is of interest to identify the steady state error of the AC
exciter alone. This has been done by injecting the measured
generator field voltage to the steady state AC exciter model
and finding the calculated value of exciter field current. The
calculated exciter field current is then compared with the
measured value (in amperes) and an error assigned to the value.
This methodology also confirms the exciter field current base
with the least error. The outcome is shown in Figure 14.
E. Exciter saturation
The saturation function in the developed PowerFactory
model uses an exponential function to model the saturation
characteristic. DIgSILENT have reviewed the ASEA exciter
datasheet provided by the generator and have designed the
saturation characteristic of the exciter model accordingly. The
exciter saturation function has been modelled as an exponential
function:
S(Ex ) = A eBEx

(4)

The simulated open-circuit air gap line and the saturation

characteristic of the revised exciter model are shown in Figure 15.

TransGrid recommended that a scaling factor to the normally calculated value of Fex may be required to
meet the steady state accuracy of exciter field current. An outcome of determining exciter field current
base was to apply a scaling factor of 0.94 to the Fex value.
Based on the data presented, the base exciter field current is determined as 8.84 A that is 8.84 A exciter
field current for 1.0 p.u generator terminal voltage on the air gap line.

field voltage according to the d-axis time constants (Tdo and

Tdo) and the exciter time constant. Note that the unit was
operating at 0.90 p.u. terminal voltage to minimize the effects
of saturation.

10

The response of the rotor voltage to this sudden deexcitation and discharge is shown in Figure 16.

2
1.00

X = 13.370 s

DIgSILENT

% error in calculated exciter field current

X = 13.858 s

-2
0.75

-4

-6
0.50

-8
0.341
0.337

Y = 0.337

-10
0

50

100

150
200
Measured rotor voltage (V)

250

0.25

300

Figure 4: Percentage error in steady exciter field current values against measured rotor voltage

DIgSILENT

Figure 14. Percentage error in steady exciter field current values against
measured generator field voltage
1000.0

[-]

800.00

0.00

-0.25
13.00

13.40
Signals: Measured Rotor Voltage in p.u.
G3: Simulated Rotor Voltage in p.u.

13.80

14.20

14.60

[s]

15.00

Figure 16. De-excitation test of the field to determine the open loop time
constant of the exciter

600.00

The time taken for the rotor voltage to fall by 63.2% is

0.488 seconds. This is the time it takes for the induced dc
current (excitation current) maintaining the flux linkages to
decay to zero. It is not the exciter open loop transient time
constant as the exciter is connected to the rotor winding and
there is a resulting feedback path through the demagnetising
gain, KD. It is thus expected that the observed time constant
in this test will be lower than the required time constant TE.

400.00

200.00

6
0.0000
0.0000

8.0000
16.000
avr_T1G3: Exciter Field Current (A) / Exciter Voltage (V) Air Gap
avr_T1G3: Exciter Field Current (A) / Exciter Voltage (V) with Saturation

Figure 15.

24.000

32.000

[-]

40.000

PowerFactory model exciter saturation characteristic

F. Exciter time constant

The exciter time constant is represented by a simple fixed
time constant TE with a minimum limit limiting the state
variable and all negative values, as the ac exciter can not
provide negative field current due to the rotating diodes.
The exciter time constant may be determined by a simple
time domain test whereby the discharge resistor on the excitation system is bypassed and a diode is put in place such that
the field of the exciter is allowed to discharge through itself.
With the diode in place the excitation is suddenly removed
and the field allowed to discharge. The time constant of the
decaying waveform is then determined. This is taken to be the
time constant of the exciter machine.
As the test is performed in the unsynchronised condition,
the active component is the d-axis stator flux and flux in the
exciter (as affected by the field winding and demagnetisation
caused by the rotor current through KD) and hence terminal
voltage of the synchronous machine responds to a change in

The decay of the rotor voltage is largely dependent on TE

and KE and for simplification, KD was excluded from the
calculation. The time constant TE is calculated to be 0.655
seconds. Using this time constant in the closed loop model
produces a close match between the measured results and the
simulated results, as shown in Figure 16.

G. Dynamic performance
The dynamic performance of the model in time domain
was verified by performing unsynchronised and synchronised
voltage step response simulations and comparing them to the
actual plant response. Figure 17 compares the responses of
the model and the actual plant response to a synchronised
voltage step response. Good agreement is obtained between
the simulated and measured terminal voltage.
In addition to the step response tests, the exciter model
was further validated by open loop simulations. This involved
injecting measured excitation voltage (treating the AVR as a
black box) to the field of the exciter model and comparing
the simulated rotor voltage with the measured data. Figure 18
compares the open loop response of the exciter model and the
actual plant response to a synchronised voltage step response.

DIgSILENT

1.095

1.087

the time domain and frequency domain due to access to the

main field. The simulated responses agree well with the actual
plant response in both domains.
The results in both cases are quite good and show high
level of agreement with the actual plant response. The validated
models for both case studies confirms that the simplified IEEE
ac exciter model is suitable for power system stability studies.

1.079

1.071

ACKNOWLEDGMENT
The author would like to thank the staff at DIgSILENT
Pacific, in particularly Tim George and David Browne for their
assistance in the model validation work and support, and also
Ron Oosterwijk for his assistance in field testing and support.

1.063

1.055
12.00

14.00
Signals: Measured Terminal Voltage in p.u.
G3: Terminal Voltage in p.u.
Signals: UgLower
Signals: UgUpper

16.00

18.00

20.00

[s]

22.00

R EFERENCES

DIgSILENT

Figure 17. Synchronised voltage step response to a 2.5% step applied to the
AVR of Hydro, overlayed with actual plant response
1.40

[1]

[2]

1.20

[3]
1.00

[4]

0.80

0.60

[5]
0.40
10.00

12.80
15.60
iexc file: Measured Excitation Current in p.u.
avr_T1G3: Simulated Excitation Current (Vfe) in p.u.

18.40

21.20

[s]

24.00

12.80
Signals: Measured Rotor Voltage in p.u.
sym_20843_3: Simulated Rotor Voltage in p.u.

18.40

21.20

[s]

24.00

2.10

1.70

1.30

0.90

0.50

0.10
10.00

15.60

Figure 18. Open loop response to a 2.5% step applied to the AVR of Hydro,
overlayed with actual plant response

V.

C ONCLUSION

DIgSILENT has approached two case studies differently

based on the availability of key signals at each site. The
simplified IEEE representation for the brushless ac exciter was
used to represent the exciter at both sites, however the approach
to develop and validate the model was quite different.
In case study A, minimum time was spent on site perform
field tests. However, a considerable amount of time was
required to perform Parameter Identification simulations to
optimise the model to fit the actual plant response. As no
access was available to the main field, the exciter model can
not be verified in isolation, however the overall simulated
response of the generating unit agrees well with the actual
plant response.
In contrast, case study B required time on site and access to
the generator to perform all the measurements at standstill, in
the unsynchronised condition, and synchronised to the grid.
The extensive tests and data collected reduced the time to
derive model parameters and validate the model for the exciter.
The exciter model was able to be verified in isolation both in

IEEE Power Engineering Society, IEEE Recommended Practice for

Excitation System Models for Power System Stability Studies, Standard
421.5-2005, April 2006.
IEEE Power Engineering Society, IEEE Standard Definitions for Excitation Systems for Synchronous Machines, Standard 421.1-2007, July 2007.
IEEE Power Engineering Society, IEEE Tutorial Course Power System
Stabilisation via Excitation Control, June 2007.
Prabha Kundur, Power System Stability and Control, McGraw-Hill, Inc,
1994.
S.M.L.Kabirn, R.Shuttleworth, Brushless exciter model, IEE, January
1994.