Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Abstract—In this paper, a method to determine load One of the most important components in a power system
conditions under which the armature current of a salient is the synchronous generator. In countries where the elec-
pole synchronous machine is on the q-axis is described. tric power generation is based on hydraulic sources, salient
The proposed method allows the determination of a locus in
the P −Q space in which the armature current is completely pole synchronous machines are responsible for generating the
on the q-axis. The derivation of these conditions simplifies majority of the electric power. For this reason, efforts to
the determination of the machine q-axis parameters during develop tests to model these machines are fundamental to
load rejection tests, since it eliminates the necessity for the determination of accurate parameters, resulting in adequate
trial and error procedures or power angle estimation. The power system models.
method was tested in a 100 MVA, 13.8 kV, 60 Hz, 40 poles
salient pole synchronous machine installed in a hydro Many models and parameters determination tests have been
power plant. Tests results show that loads calculated with developed across the years to simulate the machines behavior
the proposed method are in fact points where the armature accurately [6]. Those tests are also part of standards such as [7]
current is on the q-axis. and [8]. Models structures are based on the physical relations
Index Terms—Hydroelectric Generators, Parameter esti- between machine voltages, currents, magnetic fluxes, speed
mation, Synchronous Machines. and torque. In such models it is common to apply the two-
axis theory [9] in order to derive circuits for the machine direct
I. I NTRODUCTION axis (d-axis) and quadrature axis (q-axis).
Once a model structure is defined, it is necessary to estimate
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
the steady state condition under which the armature current inputs the values that are subject to changes during the load
is on the q-axis. Those methods are based only on estimates rejection tests. In [21], an incremental time domain model
of the machine synchronous reactances and measurements of is derived from electrical equations presented in [22] and
its armature voltage. The first author of this paper proposed mechanical equations deduced by the author of that paper. This
analytical methods to determine the load conditions under incremental model simplifies on-line parameter identification
which the armature current is on the q-axis for round rotor since its initial states are always known and equal to zero.
machines [11] and for salient pole machines [12]. In that In [23], an incremental model slightly different from the one
papers, the methods were validated through simulations. presented in [21] is used as a linear approximation for the
This paper is an extension of [12]. The main novelty non-linear model.
introduced in both papers is an analytical method to calculate Once the most suitable model structure is chosen, it is
the load conditions that guarantee pure q-axis armature current. necessary to determine the parameters for a specific machine.
This allows the execution of load rejection tests to determine Some methods to execute that task are discussed in the
q-axis parameters without the necessity to measure δ or following section.
use more advanced parameter estimation procedures. In this
paper, a detailed derivation of the equations introduced in III. S YNCHRONOUS MACHINE PARAMETERS
[12] is presented and the method is validated with field tests DETERMINATION
performed on a 100 MVA, 13.8 kV, 60 Hz, 40 poles salient
pole synchronous machine, confirming the simulation results Once the model is defined, there are two main approaches
obtained in [12]. to determine its parameters. The first one, generally used by
The paper is structured as follows: In the following sec- manufacturers to predict parameters for a new design, is based
tion, salient pole synchronous machine models are discussed. on the calculation of the reluctances of the machine’s magnetic
Section III briefly reviews methods for synchronous machine circuits based on its geometry and materials characteristics.
parameters determination. In section IV, the derivation of These calculations may be based on analytical equations, such
the proposed method is explored in a more detailed manner as the ones developed in [24]–[26] or finite element models
than in [12]. The section V introduces an example of how [27].
to determine load conditions under which the salient pole The second approach is to determine these parameters from
synchronous machine armature current is on the q-axis. A tests. In that approach, a transient phenomenon that excites the
sensibility analysis to variations on synchronous reactances machine dynamics must occur. This transient is applied to one
estimates is also presented in that section. In section VI, the of the machine windings and responses are measured on the
tests results are shown and discussed. Finally, in section VII, other windings, allowing parameters determination through a
the conclusions are presented. classical gray box system identification procedure [28].
One of the first proposed approaches to cause a transient is
to impose a sudden short circuit on armature terminals while
II. S ALIENT POLE SYNCHRONOUS MACHINE MODELS
the machine is excited and running at rated speed [29]. This
The behavior of synchronous machines changes depending corresponds to a negative step in the armature voltage. After
on the windings that are under flux variation. In the steady the short circuit, the transient armature current is measured,
state condition, the most significant flux variations are on and the machine d-axis parameters can be determined by
the armature windings. In salient pole synchronous machines, fitting exponential functions to the current waveform envelope.
due to the rotor salience, the reluctance of the magnetic A disadvantage is that the sudden short circuit test may
circuit seen by the armature fluxes changes depending on the introduce hazardous currents and torques to the machine,
rotor position. To model this effect, the two-axis theory was which may damage some of its structural components. Another
developed during the second quarter of the last century [9], disadvantage is that it is not possible to determine the q-axis
[13]. The application of this theory to the steady state analysis parameters from the data obtained in sudden short circuits.
of salient pole synchronous machines can be found in standard Another transient phenomenon that can be used to determine
textbooks such as [14]–[16]. machine parameters was proposed in [21]. In that paper, a step
In transient conditions, the field and the amortisseur win- in the field voltage is caused by introducing an inductance
dings are under significant flux variations and, to obtain an connected in parallel with a switch downstream the excitation
accurate mathematical model, it is necessary to consider the transformer. Switch state modifications introduce or bypass the
resistances and the inductances of those windings. The resul- inductance in the circuit, changing the machine field voltage.
tant model can then be analyzed in the frequency domain with Currents and voltages from the machine windings are mea-
transfer functions [17] or in the time domain, with state space sured during switch changes and then a least-squares algorithm
models [18], [19]. Those models are parametric, meaning that is used to fit the model parameters to the experimental data.
their structure is equal for machines with the same constructive In [23], the transient is caused by superposing a white noise
characteristics. Each machine is then differentiated by the to the machine field voltage. Then terminal voltage and active
numerical values of its parameters. power are measured, and two transfer functions are estimated.
Transient multivariable time domain models are flexible and Parameters from these transfer functions are then related to
can be reorganized to simplify the parameter determination certain classical mechanical and electrical parameters of the
procedure. In [20], a hybrid model is proposed to select as machine, completing the identification procedure.
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Ef (pu)
0
1st root of Pp(|Ef|)=0
final solution |Ef | for a given δ in the interval described in the
−0.2 2nd root of P (|E |)=0
equation 22 is defined as the root that satisfies both equations. p f
1st root of Pq(|Ef|)=0
−0.4
The calculation of the roots for each δ in the interval defines 2nd root of Pq(|Ef|)=0
a discrete curve in the |Ef | − δ space for which the armature −0.6
desired |Ef |.
With the determination of the pair (δ, |Ef |) that satisfies Fig. 1. Solutions of the equations 11 and 18 for Vt = 1 pu and the
the armature current on the q-axis, all other variables, such reactances presented in the table I.
as P and Q, can easily be determined using the equations
1 to 4. This allows for the creation of a locus on the P − Q Quadrature axis current loci
space that defines the points for which the pure q-axis armature 1.5
Data−sheet values
currents are expected. Then, speed and voltage regulators can 80% of data−sheet values
120% of data−sheet values
be adjusted to make the machine run under a load on the
desired locus, and then a load rejection can be performed,
allowing the validation of the q-axis synchronous reactance 1
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
TABLE II locus. When this condition was reached, |Vt | was checked
S IGNALS ACQUIRED DURING TESTS AND EMPLOYED ACQUISITION again. If no significant change had happened, the armature
MODULES
breakers were opened and the load rejection data was recorded.
Signal Measurement Element Acquisition
module
Armature voltages Voltage transformer D. Method validation
Phases R, S, T 14400 V/ 120 V NI 9244 To validate the method, thirteen load rejection tests were
Armature currents Current transformer
Phases R, S, T 6000 A/ 5 A NI 9227 performed, including some loads close to the locus of pure
Field current Shunt Transducer q-axis armature current and others far from this locus. To
2000 A / 100 mV NI 9238 verify if the armature current before load rejection was on
Field voltage Transducer
0 V-300 V / 0 mA -10 mA NI 9203 the q-axis, the variation of field current during the rejection
Active Power Transducer was calculated. Since the field winding is on the d-axis, it is
SINEAX DME442 NI 9203 expected that q-axis transient phenomena do not interfere with
4 mA -20 mA /
0 MW -150 MW
voltages and currents from this winding. So, if the armature
Reactive Power Transducer current is on the q-axis before the load rejection, no significant
SINEAX DME442 NI 9203 field current variation is expected [10].
4 mA-20 mA /
-100 MVAr -100 MVAr
To calculate the field current variation, the average value
Angular velocity Encoder 120 slots of the moving average of that signal, calculated between
Speed monitor - Transducer NI 9203 0.55s and 0.05s before rejection, was determined and defined
0 Hz -400 Hz/
0 V-10 V/
as the field current before test (If b ). Then, maximum and
4 mA-20 mA minimum values of the moving average after load rejection
were obtained and defined as If amax and If amin , respectively.
Given those values, the field current variation ∆If was defined
as follows:
comprised of a Toshiba analog voltage regulator, which is
responsible for voltage control when the unit is synchronized
∆If = max(|If amax − If b |, |If amin − If b |) (23)
with the power system. The controller has the following
limiting functions: OEL (Overexcitation limiter), UEL (Sub- In figures 3 and 4, results of two of the thirteen tests are
excitation limiter) and PSS (Power system stabilizer). The presented. On the left, in both figures, the q-axis load locus in
speed regulation system is from Mitsubishi, model EA-5, ana- the P − Q space, obtained with the method described in this
log type. This system offers the function of primary frequency paper, is plotted together with the average P and Q measured
regulation. Additionally, it has a PID regulation characteristic, between 0.55s and 0.05s before the load rejection. On the
which performs the primary frequency regulation and imple- right, in both figures, the field current measured during load
ments the statism by re-feeding the actuator position. This rejection, and a moving average of this signal, are shown. The
loop provides an actuator position setpoint that is sent to the moving average was adopted to filter signal noise.
position control loop. The position control of the actuator is As can be seen in the left graphic in figure 3, the load
carried out through a P loop. This regulation system allows just before that test was in the locus where pure q-axis
the control of P and Q and is able to operate in a constant armature current is expected. The graphic in the right shows
field voltage mode. that the field current variation was negligible during that test,
as expected from the results of this paper. Under that condition,
B. Machine instrumentation the value of ∆If calculated according to equation 23 is 0.8%.
Data in the left side of figure 4 shows that, before the
The data acquisition system used during the validation of the
load rejection, the load was far from the locus where pure
proposed method was the National Instruments CompactDAQ
q-axis current is expected. The curve on the right shows
9188. Details about the acquisition modules used and the ac-
that field current variation was significant. For that condition,
quired signals are presented in table II. The acquisition system
∆If =15.95%. This result is coherent with the deduction of
is adjusted through the LabVIEW SignalExpress platform,
the locus described in this paper.
with a 5000 Hz sampling rate.
To summarize the results of all thirteen tests, the minimum
Euclidean distance between the point defined by the average
C. Test procedure P and Q, measured between 0.55s and 0.05s before the load
During the tests, the average |Vt | was measured and the rejection, and the pure q-axis locus on P − Q space, was
method described in the section IV was executed to determine calculated for each one of the tests. Field current variation
the curve in the |Ef | − δ space for which the armature current was also calculated with the equation 23, following the same
would be on q-axis considering the synchronous reactances procedure used for the first two tests. The results are plotted
shown in table I. Then, P and Q were calculated for each point in figure 5. From that figure it is possible to notice that loads
in the |Ef |−δ curve and a locus in the P −Q space, satisfying with small distances from the pure q-axis locus in the P − Q
the q-axis condition, was plotted. In sequence, voltage and space imply small field current variations when rejected. This
speed regulators were adjusted in such a way that the machine indicates that these loads are in fact q-axis loads, validating the
active and reactive powers defined a point close to the desired procedure derived in this paper. On the other hand, loads far
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
1 720
Field current
Moving average with 50 samples
0.8 710
700
P (pu)
0.6
If (A)
0.4 690
Calculated locus
Pre−rejection condition 680
0.2
Fig. 3. Load rejection under a condition close to the P − Q locus for pure q-axis armature current. On the left, the calculated locus for the terminal
voltage immediately before the rejection and the load condition immediately before the rejection. On the right, field current variation measured
during load rejection and its moving average.
If (A)
0.4 500
0.2
450
−1.5 −1 −0.5 0 31 31.5 32 32.5
Q (pu) t(s)
Fig. 4. Load rejection under a condition far from the P − Q locus for pure q-axis armature current. On the left, the calculated locus for the terminal
voltage immediately before the rejection and the load condition immediately before the rejection. On the right, field current variation measured
during load rejection and its moving average.
from the locus imply greater field current variations, indicating time constants can be found with a simple determination of pa-
that a d-axis component is present. This is also expected from rameters from an exponential equation. Although the method
the results derived in this paper. An interesting fact that is depends on initial estimates of the synchronous reactances,
also shown in figure 5 is that there is a reasonable linear fit those estimates do not need to be accurate if the load is
that relates the field current variation to the distance from pure chosen from an appropriate region of the locus derived with
q-axis locus. the method.
The method was validated in a 100 MVA, 13.8 kV, 60
VII. C ONCLUSION Hz, 40 poles salient pole synchronous machine installed in
In this paper, a method to determine load conditions under the Salvajina hydro power plant. Test results show that loads
which the armature current of a salient pole synchronous calculated with the proposed method in fact resulted in points
machine is on the q-axis was described, and the equations where the armature currents are on the q-axis. It was also
to derive that method were detailed. The method allows the shown that loads far from the locus derived in this paper
determination of a locus on the P − Q space in which the present a significant d-axis component, which is evidenced by
armature currents are completely on the q-axis. The derivation the significant field current variation during load rejections.
of that locus simplifies the determination of the machine q-axis
R EFERENCES
parameters, since it eliminates the necessity for trial and error
procedures or the measurement or estimation of the load angle. [1] B. L. Agrawal and J. A. Demcko, “Wscc testing experience [generating
units],” in IEEE Power Engineering Society. 1999 Winter Meeting (Cat.
If any load in that locus is rejected, the synchronous machine No.99CH36233), vol. 1, DOI 10.1109/PESW.1999.747443, pp. 160–161
synchronous, transient and sub-transient q-axis reactances and vol.1, Jan. 1999.
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Salvajina [17] B. Adkins, The general theory of electrical machines. Chapman and
Hall, 1957.
Measured data
45 Linear interpolation [18] O. Elgerd, Electric energy systems theory: an introduction, ser. McGraw-
Hill Series in Electrical Engineering. McGraw-Hill, 1982.
40
[19] P. C. Krause, O. Wasynczuk, and S. D. Sudhoff, Analysis of electric
35 machinery and drive systems, 2nd ed. Wiley-IEEE Press, 2002.
Field current variation [%]
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIE.2018.2866096, IEEE
Transactions on Industrial Electronics
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Mateus Giesbrecht was born in Campinas, SP, Luis Alfredo Esteves Meneses Luis Alfredo
Brazil. He received the B.Sc., M.Sc. and Ph.D. Esteves was born in Cali, Valle, Colombia.
degrees in Electrical Engineering from Univer- He received the B.Sc. and M.Sc. degrees in
sity of Campinas (UNICAMP), Brazil, in 2006, Electrical Engineering from University of Valle
2007 and 2013 respectively. (UNIVALLE), Colombia, in 2011 and 2013
He is a full time professor at the School of respectively.
Electrical and Computing Engineering of the He is a leader of electrical maintenance at
University of Campinas, Brazil, after nine years hydroelectric power plant Alto Anchicayá at the
working as electrical engineer at Andritz Hydro energy company of the Pacific (EPSA) since
(former GE Hydro), with large synchronous ge- 2013. He is responsible for speed and voltage
nerators design. His fields of interests include regulation systems design and commissioning,
electric machines design, electric machines parameters identification, and synchronous generators commissioning for EPSA’s hydroelectric
system identification, time series realization and evolutionary algo- power plants. His fields of interests include electric machines
rithms. parameters identification, design of speed regulation system and voltaje
Prof. Giesbrecht is a reviewer from the IEEE Transactions on Industrial regulation system, predictive maintenance and characterization of
Electronics and IET Power Electronics. synchronous machines.
0278-0046 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.