1

Detailed off-line parameter identification of

Synchronous generator based on frequency
response tests
Topic number: Educational

 Zd(s) d-axis operational impedance

Zq(s) q-axis operational impedance
Abstract— This work clearly shows a general methodology, G(s) Armature to field transfer function
based on Standstill Frequency Response Tests indicated on IEEE Zafo(s) Armature to field transfer impedance
Std 115-A. It allows calculating the direct and quadrature axis
operational impedances of a salient pole synchronous generator
I. INTRODUCTION
from a chosen circuit topology. The novel and educational
approach, is based on the fact that the order of the circuit can be
whatever, even a customized circuit topology since the
methodology does not rely on a set of a given formulation, but in
N OWADAYS there is a trend to move toward “more electric
aircraft”, that is, replace hydraulic and pneumatic sources
of power to relay on electrical power, such as actuators,
a general and reproducible comprehensible approach. Nowadays deicing, engine –start, cabin air-conditioning, among others.
many methods are presented in the literature to calculate such a
As a consequence the electrical power requirement is expected
parameters, but with the dedicated data processing and analysis
methodology, time constants and reactances identification is to easily overpass 1 MW. Typical aircraft generators are rated
carried out, resulting in an inexpensive accurate and at 120 kVA, so new electrical system architectures are
harmlessness methodology. expected [1], but the core of the system is still the
synchronous generator. On the other hand, regarding the
Index Terms— Synchronous genetaror, standstill tests, machine’s medialization, the Park’s model has survived until
equivalent circuit, identification parameters. now, by means of adding resistor-inductance branches, more
effects can be modelled, such a second or third order effects.
NOMENCLATURE In the literature second order model if synchronous machine is
ed d-axis armature voltage [V] referred to as SSFR2 and the third order as SSFR3. By means
eq q-axis armature voltage [V] of these models different issues can be studied, for example
efd Field voltage [V] calculation of voltage regulation, fault identification and
varm voltage between two armature terminals during test [V]
id d-axis armature current [A]
analysis, small and large disturbance analysis, exciters
iq q-axis armature current [A] designs, high performance controllers development [2]. In this
ifd Field current [A] context the need of a reliable and feasible set of circuit
iarm Armature current during test [A] parameters is mandatory, this has been and is still a
Ra Armature resistance [Ω]
Rfd Field resistance referred to the armature [Ω]
challenging investigation line.
Rkd d-axis damper winding resistance, k =1,2…n [Ω] Parameter identification is classified into two methods, i.e.
Rkq q-axis damper winding resistance, k =1,2…n [Ω] off-line and on-line identification respectively. In off-line
Ll Armature leakage inductance [H] methods the machine is out of service, among them one can
Lad d-axis armature to rotor mutual inductance [H]
Laq q-axis armature to rotor mutual inductance [H]
find open circuit frequency response, dc-excitation methods
Lfd Field winding leakage inductance [H] and standstill frequency response [3], which is used into IEEE
Lkd d-axis damper winding leakage inductance, k =1,2…n [H] Std. 115-A, and in the methodology of this work .
Lkq q-axis damper winding leakage inductance, k =1,2…n [H] However, the determination of the subtransient reactances, in
Ld(s) d-axis operational inductance
Lq(s) q-axis operational inductance
both axes, under saturation conditions is crucial for the
T’d0 d-axis open circuit transient time constant [s] calculation of the electrical and mechanical integrity of the
T’’d0 d-axis open circuit subtransient time constant [s] machine during a three phase or two phase short-circuit [4].
T’q0 q-axis open circuit transient time constant [s] Besides, the estimation of the machine flux linkages is
T’’q0 q-axis open circuit subtransient time constant [s]
T’d d-axis short circuit transient time constant [s]
difficult, due to the strong saturation in the operation of the
T’’d d-axis short circuit subtransient time constant [s] machine, which in turn causes the cross-saturation effect.
T’q q-axis short circuit transient time constant [s] Thus, flux-linkages are function of both d and q axis currents
T’’q q-axis short circuit subtransient time constant [s] although the Standstill frequency response modelling is used
operational impedance measured between two armature
Zarmd(s)
terminals during d-axis tests
here because the d and q axis are decoupled and this method
operational impedance measured between two armature allows an accurate determination of high order models since
Zarmq(s)
terminals during q-axis tests the sinusoidal signal can excite almost all frequencies with the
same amplitude. Besides, the sinusoidal signals are easy to
noise decorrelation [5]. This parameter estimation
 2

methodology consists on extracting the time constants The procedure for calculating the synchronous machine
applying curve-fitting and then the equivalent circuit parameters requires curve-fitting techniques which is done in
parameters are obtained by solving a set of non-linear an easy and intuitive way by using the Matlab software.
equations. In the literature these non-linear equations are
solved through numerical optimization and this is a process
full of numerical difficulties [6-8].
It should be noted however that on-line methods are receiving
much attention because of their minimal impact and
interference in the operation of the machine, which is in
service [9-12] such that no interference occurs in the operation
of the system. Mainly they belong to time domain tests by
using a small perturbation on the machine terminals, standstill
time domain test, and load rejection test, but the standstill
frequency response test is still used because its reliability its Fig.2. d- axis equivalent circuits
harmlessness to the machine with a relatively simple
measurement set-up. The d- and q-axis operational impedances are respectively
Furthermore, on-line identification methods can be classified Z d ( s )  Ra  s  Ld ( s ) (1)
into three categories, white-box, black-box and grey-box
Z q ( s )  Ra  s  Lq ( s ) (2)
modelling. The white box modelling, knows exactly the
topology of the system the parameters are to be calculated or and both can be evaluted through small changes in armature
estimated, that is exists a theoretical model because the voltage and current when the field winding is short-circuited,
physical knowledge is known a priori, among them extended by positioning the rotor, respectively, for d- and q-axis tests:
Kalman filter [13], conjugate gradient method, neural e ( s ) (3)
Z ( s)   d
networks, maximum likelihood, evolutionary programming, d
id ( s ) e fd  0
orthogonal series functions, and Hartley series can be found in
eq ( s ) (4)
the literature [14]. The black box however, is based on the Z q ( s)  
input-output data because no physical knowledge is known, in iq ( s ) e fd  0
such a way that is known as a data driven model. On the other Additional measures must be taking into account depending
hand the grey box is in between, is a hybrid model wherein the on the order d- and q-axis equivalent circuits required to
analytical equations and input-output data relations are used to simulate accurately the transient behaviour of the synchronous
obtain the parameters of the model. In this context a white machine. These measurable parameters are the armature to
model is used here because the topology of the circuit is field transfer function, G(s), and the armature to field transfer
known and the order is assumed. impedance, Zafo(s). The armature to field transfer function can
But from an educational perspective, a general methodology be evaluted through small changes in field and armature
is presented, in such a way that from a customized chosen currents when the field winding is short-circuited:
circuit topology, it is able to obtain the circuit parameters, i ( s ) (5)
based on Matlab software. sG ( s )  fd
id ( s ) e fd  0
The work is divided into five sections. In Section II is
While the armature to field transfer impedance can be
presented the required tests, based on the IEEE Std. 115-A, on
evaluated through small changes in field voltage and armature
d-q axis for data extraction. Section III develops the general
current when the field winding is in open-circuit:
methodology for machine parameters determination. Section
e ( s ) (6)
IV presents the obtained results and its validation. Finally, Z ( s )   fd
id ( s ) i fd  0
afo
Section V presents the conclusions.

II. STANDSTILL FREQUENCY RESPONSE TESTS In order to show how the general methodology based on
Standstill Frequency Response works, the second order d- and
In this work a method based on IEEE Std 115A-1987 is
q-axis equivalent circuits are chosen, but this methodology can
described for obtaining synchronous machine parameters by
be applied to other equivalent circuits that can be found in
performing frequency response tests with the machine at
IEEE Std 115-A. If secon order equivalent circuits are chosen
standstill. Usually the second order d- and q-axis equivalent
then only two measurable parameters are required over a range
circuits, Fig. 1and Fig. 2, are used to analyze the synchronous
of frequencies, these are d- and q-axis operational impedances.
machine stability.
The first step is to position the rotor for d-axis tests by doing
the connections shown in Fig. 3. After that the amplifer is set
to approximately 100 Hz. The next step is to turn the rotor
slowly until the observed field voltage is nulled. When this
condition is accomplished the magnetic axis of the field
winding is aligned with the series connection of phase a and b.

Fig.1. q-axis equivalent circuit

 3

La Oscillator

Lb
Frequency
Response
Power La
Lc Varm I arm Analyzer Amplifier
Lfd
Shunt

Lfd
Lb Lc
Fig. 3. Positioning the rotor for d-axis tests

After the d-axis is positioned the d-axis tests can be started by Fig. 6. Quadrature axis impedance transfer function test
doing the connections shown in Fig. 4., where a and b
armature phases are supplied by using a frequency variable The leakage inductance is needed, be it taken from the zero
power amplfier, and c phase is in open-circuit. power factor test or taken from the manufacturer.
The field circuit will be short circuited with a non-inductive
A. d-axis parameters from tests
metering shunt. Connect the voltage and current
measurements of the stator winding to the measurement From the d-axis impedance transfer function test the
instrument. Perform the measurement over a bandwidth from armature voltage and current are measured depending on the
0,001Hz to 1000Hz. frequency, so the operational impedance measured between
two armature terminals during d-axis tests, Zarmd (s), can be
calculated from its relation. In this test there is the series
connection of the windings then d-axis operational impedance
La is:
Varm I arm Z
Z  arm (7)
d
2
The result is shown in Fig. 7
Lfd
d-axis impedance
Lb Lc 1e1

80

1e0

Fig. 4. d- axis impedance transfer function test 60

Zd Magnitude ( )

Zd Phase ( º)
1e-1
Similarly, once the q-axis is positioned the q-axis tests can be
started by doing the connections shown in Fig. 5. The field 40
circuit will be short circuited as aforementioned and the same 1e-2

bandwidth is applied, as presented in Fig.6.

20
1e-3

La 1e-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
0
3

Lb Lc Frequency (Hz)
Fig. 7. d-axis impedance

Lfd The armature resistance, Ra, can be obtained from the real
component of Zd (s) extrapolated at zero frequency.
The operational inductance is d-axis is
Z s   Ra (8)
L s   d
Fig. 5. Wiring of armature and field winding for rotor positioning for d
s
quadrature axis test The d-axis inductance at zero frequency Ld(0) can be
calculated extrapolating Ld(s) to zero frequency as shown Fig.
8.
III. METHODOLOGY
The direct-axis armature to rotor mutual inductance is
The design methodology proposed in this work starts with obtained from the difference between the d-axis inductance at
the definition of the topology of d-q axis circuits, as aforesaid zero frequency and the leakage inductance as follows:
mentioned in this work a SSFR2 is used, but it must be noted
that this methodology can be properly applied to whatever
circuit representation.
 4

d-axis inductance
1e1 0

q-axis inductance

1e0
-10 0
Ld Magnitude ( mH)

1e1

Ld Phase ( º)
1e-1

Lq Magnitude ( mH)
-20 -10
1e0

Lq Phase ( º)
1e-2
1e-1
-30 -20

1e-3 1e-2

-40 -30
1e-3
1e-4
-3 -2 -1 0 1 2 3
10 10 10 10 10 10 10
Frequency (Hz) 1e-4 -40
-3 -2 -1 0 1 2 3

Fig. 8. d-axis inductance 10 10 10 10

Frequency (Hz)
10 10 10

Fig. 10. q-axis inductance

Lad  Ld 0   Ll (9)
The q-axis inductance at zero frequency Lq(0) can be
B. d-axis equations set calculated extrapolating Lq(s) to zero frequency. And similar
The first step is to define the symbolic parameters that are to equation (9), from this value subtracting the leakage
represented in the equivalent circuit. The following steps inductance the quadrature inductance is calculated
consist on the association of the components in order to obtain
the symbolic transfer function. The time constants presented in Laq  Lq 0  Ll (13)
equation (#) , which depends on the circuit order, are fitted
with the experimental data shown in Fig. 8. D. q-axis equations set
By means of the same procedure the rest of the model
Ld 0  1  s  T 'd   1  s  T ' 'd  (10) parameters for q-axis equivalent circuit L1q, R1q, L2q, R2q, are
Ld s  
1  s  T 'd 0   1  s  T ' 'd 0  determined by curve-fitting technique from the operational
Ld(0) is the direct axis inductance in the low frequency Lq(s).
bandwidth.
Lq 0  1  s  T 'q  1  s  T ' 'q  (14)
Lq s  
The fitted coefficients are equal to the corresponding 1  s  T '  1  s  T ' ' 
q0 q0
constant time in the transfer function. In this way, an equation
system can be obtained in order to find the parameters of the Lq(0) is the direct axis inductance in the low frequency
circuit. It must be solved in order to obtain the rest of the bandwidth.
model parameters in Fig. 2, that is, Lfd, L1d, R1d, Lfd .
C. q-axis parameters The algorithm applied on both axes, for this purpose is
exposed below.
As, in d-axis, the armature impedance measurement for
each sampling frequency will be obtained from the quadrature
1: Set the d-q circuits SSFRi order, or another customized
axis impedance transfer function test, from which the
circuit topology.
quadrature-axis operational impedance is obtained, as shown
2: Obtain module and argument of d-axis operation impedance
in equation
3: Obtain resistance of armature
Z
Z  arm (11) 4: Obtain module and argument of d-axis operation inductance
q
2
6: Determination time constants
q-axis impedance
Curve fitting: invfreqs command from Matlab care must
1e1
be taken when choosing the weighting factors.
1e0
80
5: Obtain set of equations by means of symbolic toolbox
children and collect Matlab commands to properly
60
equation manipulation
Zq Magnitude ( )

Zq Phase ( º)

1e-1

6: Solve set of equations, fsolve command from Matlab

7: If with the obtained parameters, the dynamic simulated
40
1e-2

behavior of the machine does not agree with experimental data

then go to 1 else
20
1e-3

8: end
1e-4 0
-4 -3 -2 -1 0 1 2 3
10 10 10 10 10 10 10 10
Frequency (Hz)
Algorithm 1. Solving procedure to obtain the machine parameters
Fig. 9. q-axis impedance

The quadrature operational inductance is For the sake of better understanding the scheme of the
procedure is shown in Fig. 11.
Z q s   R a
Lq  s  
s (12)
 5

Fig. 1. Algorithm flow chart

Laq 0.007155 [H]

Lfd 0.000985 [H]
IV. RESULTS AND VALIDATION L1d 0.000617 [H]
The proposed methodology approach is applied to the L1q 0.000522 [H]
L2q 0.0109 [H]
machine used in the Std. in order to validate the obtained
T’d0 3.89 [s]
results, for the sake of clarity here indicated, i.e. 192.8 MVA, T’’d0 0.0156 [s]
18 kV, 60 Hz. T’q0 1.83 [s]
The obtained results are shown in Fig. 12 and Fig. 13, where T’’q0 0.3251 [s]
T’d 0.8018 [s]
are presented the data obtained by means of the tests indicated
T’’d 0.011 [s]
in Section II, with the calculated according to the presented T’q 0.999 [s]
methodology in Section III. T’’q 0.094 [s]

q-axis inductance

d-axis inductance 1e1

1e1 0 0

1e0
Lq Magnitude ( mH)

1e0
Lq Phase ( º)

measured measured
Ld Magnitude ( mH)

1e-1
calculated calculated
Ld Phase ( º)

1e-1

measured 1e-2
calculated measured
1e-2
calculated
1e-3

1e-3

1e-4
-3 -2 -1 0 1 2 3
10 10 10 10 10 10 10
1e-4 Frequency (Hz)
-3 -2 -1 0 1 2 3

Fig. 2. Experimental and calculated q-axis operational inductance

10 10 10 10 10 10 10
Frequency (Hz)

Fig. 12. Experimental and calculated d-axis operational inductance

V. CONCLUSIONS
Ra 0.0016[Ω]
Rfd 0.0021 [Ω] This works has presented a methodology for synchronous
R1d 0.0934 [Ω] generators d-q equivalent circuit calculation parameters based
R1q 0.0126 [Ω] on IEEE Std 115-A.
R2q 0.0116 [Ω] The obtained values by the presented methodology are in good
Ll 0.000795 [H]
Lad 0.007155 [H] agreement with than that of reported experimentally. The
 6

novel and educational approach, is based on the fact that the [2] R. Wamkeue, C. Jolette, A. B. M. Mabwe, and I. Kamwa, “Cross-
identification of synchronous generator parameters from RTDR test time-
order of the circuit can be whatever, even a customized circuit domain analytical responses, ”IEEE Trans. Energy Convers.,vol. 26, no.
topology since the methodology does not rely on a set of a 3, pp. 776–786, Sep. 2011.
given formulation, but in a general and reproducible [3] B. Zaker, G. B. Gharehpetian, M. Karrari and N. Moaddabi,
comprehensible approach, resulting fast and easy to "Simultaneous Parameter Identification of Synchronous Generator and
Excitation System Using Online Measurements," in IEEE Transactions
implement. on Smart Grid, vol. 7, no. 3, pp. 1230-1238, May 2016
[4] F. Maurer, M. T. Xuan and J. J. Simond, "Two Full Parameter
VI. ACKNOWLEDGMENT Identification Methods for Synchronous Machine Applying DC-Decay
Tests for a Rotor in Arbitrary Position," in IEEE Transactions on Industry
This work has been partially supported by the Generalitat de Applications, vol. 53, no. 4, pp. 3505-3518, July-Aug. 2017.
Catalunya under the project SGR 2017 SGR 967 and the [5] H. Radjeai, A. Barakat, S. Tnani and G. Champenois, "Identification of
European Commission through the Clean Sky program under synchronous machine by Standstill Frequency Response (SSFR) method -
influence of the stator resistance -," The XIX International Conference on
the project 755332-AEMS-IdFit.
Electrical Machines - ICEM 2010, Rome, 2010, pp. 1-5.
[6] R. M. Saunders, "Synchronous-machine standstill frequency-response test
FUNDING data analysis," in IEEE Transactions on Energy Conversion, vol. 6, no. 3,
pp. 564-571, Sep 1991.
This project has received funding from the Clean Sky 2 [7] J. Bladh, M. Wallin, L. Saarinen and U. Lundin, "Standstill Frequency
Joint Undertaking under the European Union’s Horizon 2020 Response Test on a Synchronous Machine Extended With Damper Bar
research and innovation program under grant agreement No Measurements," in IEEE Transactions on Energy Conversion, vol. 31, no.
1, pp. 46-56, March 2016.
755332. [8] E. C. Bortoni and J. A. Jardini, "A standstill frequency response method
for large salient pole synchronous machines," in IEEE Transactions on
Energy Conversion, vol. 19, no. 4, pp. 687-691, Dec. 2004.
[9] T. L. Vandoorn, F. M. De Belie, T. J. Vyncke, J. A. Melkebeek, and P.
Lataire, “Generation of multisinusoidal test signals for the identification
of synchronous-machine parameters by using a voltage-source inverter,
”IEEE Trans. Ind. Electron., vol. 57, no. 1, pp. 430–439, Jan. 2010.
DISCLAIMER [10] S. Pillutla and A. Keyhani,“Neural network observers for on-line tracking
of synchronous generator parameters, ”IEEE Trans. Energy Conversion,
© European Union, 2020 vol. 14, pp. 23–31, Mar. 1999
The information and views set out in this publication are [11] Melgoza, R. Jesus, G. T. Heydt, and A. Keyhani,“An algebraic approach
for identifying operating point dependent parameters of synchronous
those of the authors and do not necessarily reflect the official machine using orthogonal series expansions, ”IEEE Trans. Energy
opinion of the European Union. Neither the European Union Conversion, vol. 16, pp. 92–98, Mar. 2001.
institutions and bodies nor any person acting on their behalf [12] H. B. Karayaka, A. Keyhani, G. T. Heydt, B. L. Agrawal and D. A. Selin,
may be held responsible for the use which may be made of the "Synchronous generator model identification and parameter estimation
from operating data," in IEEE Transactions on Energy Conversion, vol.
information contained therein. 18, no. 1, pp. 121-126, Mar 2003.
Reproduction is authorized provided the source is [13] B. Aubert, J. Régnier, S. Caux and D. Alejo, "Kalman-Filter-Based
acknowledged. Indicator for Online Interturn Short Circuits Detection in Permanent-
Magnet Synchronous Generators," in IEEE Transactions on Industrial
Electronics, vol. 62, no. 3, pp. 1921-1930, March 2015.
REFERENCES [14] M. Karrari and O. P. Malik, "Identification of physical parameters of a
synchronous Generator from online measurements," in IEEE
[1] A. Griffo, R. Wrobel, P. H. Mellor and J. M. Yon, "Design and Transactions on Energy Conversion, vol. 19, no. 2, pp. 407-415, June
Characterization of a Three-Phase Brushless Exciter for Aircraft 2004.
Starter/Generator," in IEEE Transactions on Industry Applications, vol.
49, no. 5, pp. 2106-2115, Sept.-Oct. 2013