IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO.

1, MARCH 2002 85

Estimating the Parameters of Induction Machines at

Standstill
Cursino Brandão Jacobina, Senior Member, IEEE, João Edgar Chaves Filho, and
Antonio Marcus Nogueira Lima, Member, IEEE

Abstract—This paper presents a method to determine the con-

tinuous-time parameters of induction machines which are neces-
sary to implement the field-oriented control strategy. This method
does not require the use of any mechanical devices or speed mea-
surement sensor and it is conceived to be used during the auto-com-
missioning of the drive-system. The estimation of the stator resis-
tance is based on the dynamic model of the machine considering
that the variables are dc quantities. The estimation of the stator
transient inductance, the stator inductance, and the rotor time con-
stant, are based on the dynamic model of the machine. The exper-
imental results which are presented demonstrate the feasibility of
the proposed method.
Index Terms—Induction machine, parameter estimation.

I. INTRODUCTION

T HE USE of system identification techniques in the char-

acterization of an induction machine has been reported
by many research teams [1]–[8]. By executing the no-load and Fig. 1. AC drive system.
locked rotor tests [9], all the parameters can be determined.
However, this implies some kind of approximations and requires easily estimated. The parameter estimation procedure is based
special procedures that are relatively difficult to be automati- on a linear least-squares technique. The linear model is derived
cally implemented in a standard ac drive-system. from the dynamic model of the machine for the standstill op-
Some previous works have discussed the estimation of the erating condition. Particular attention is given to the definition
machine parameters in standstill [6]–[8]. Willis et al. [6] pro- of the excitation voltage waveform that is used in the estima-
posed a frequency domain test to determine the machine pa- tion procedure. This excitation waveform supply all the machine
rameters. Seok et al. [7] proposed a parameter identification phases with symmetry. The method does not require the
procedure to determine the rotor resistance , the rotor time use of any mechanical apparatus or speed measurement sensor,
constant , the stator inductance , and the stator transient in- so it is indicated to be used in the autocommissioning of the
ductance by using several tests implemented with a pulse drive-system.
width modulated inverter controlled by the indirect rotor flux
oriented control technique. To start-up the procedure an initial II. AC DRIVE SYSTEM
estimate of some parameters is calculated from the nameplate In the characterization of the electrical machine as discussed
data and the estimation of the rotor time constant is determined in this paper it is assumed that the induction motor is part of
by using a nonlinear technique. Ribeiro et al. [8] proposed a a microcomputer ac drive system, as shown in Fig. 1. The
linear least-squares approach to estimate , , the stator in- drive system is composed of a standard voltage source-inverter,
ductance, , and . a three-phase induction motor and a microcomputer, whose
This paper is based on a previous work of the authors [8]. software controls the overall functioning of the drive.
However, in this work a new signal is used to maintain the ma-
chine in standstill. Besides, a sensitivity study of the models is III. MACHINE DYNAMIC MODEL
accomplished to determine the parameters which can be more
For the purposes of the present investigation the induction
Manuscript received November 19, 1999. This work was supported in part by machine is assumed to be described by a stator reference frame
the Conselho National de Desenvolvimento Científico e Tecnológico (CNPq)
and the Fundação Coordenação de Aperfeiçoamento de Pessoal de Nível Supe- model given by
rior (CAPES).
C. B. Jacobina and A. M. N. Lima are with Laboratório de Eletrônica Indus- (1)
trial e Acionamento de Máquinas, Departamento de Engenharia Elétrica, Uni-
versidade Federal da Paraíba, Paraíba, Brasil (e-mail: jacobina@dee.ufpb.br). (2)
J. E. Chaves Filho is with the Departamento de Eletricidade, Universidade (3)
Federal do Amazonas, Amazonas, Brasil.
Publisher Item Identifier S 0885-8969(02)01499-7. (4)
0885–8969/02$17.00 © 2002 IEEE
86 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 1, MARCH 2002

The variables and parameters are defined as follows.

(13)
stator voltage vector;
stator current vector;
stator flux vector; (14)
rotor current vector referred to the stator;
rotor flux vector referred to the stator;
shaft speed;
stator inductance; V. EXPERIMENT DESIGN
magnetizing inductance;
rotor inductance referred to the stator; In the present paper two types of exciting signals have been
stator resistance; employed in the parameter estimation tests. In the case of Model
rotor resistance referred to the stator; a simple dc voltage waveform is used to supply the machine.
operator ; In the case of Model , a more persistently exciting signal is
imaginary unit. required. It is possible, for example, to supply the machine with
the six-steps voltage waveform [Fig. 2(b)], but in this case it
IV. ESTIMATION MODELS is necessary to use mechanical devices to keep the machine at
standstill. This paper discusses a special voltage waveform that
The use of linear least-squares estimation techniques requires keeps the machine at standstill without any mechanical appa-
the model of the system under study to be written as a regression ratus. This waveform is synthesized by a specific stator voltage
equation like vector sequence that is employed for each cycle .
The voltage waveform for one of the machine phases is shown in
(5)
Fig. 2(a). Fig. 2(b) presents the six-step waveform for the same
where , and are the prediction vector, the regression phase to help the understanding of how that special sequence
matrix and the estimated parameter vector, respectively. has been synthesized.
From (1)–(4) it is possible to derive linear regression models This waveform can be generated either by using only active
like (5). The basic equations of the least-squares algorithm used and null vectors or by using some kind of pulse width modu-
to compute may be found in [10]. lation technique. In the case of the active and null vectors the
voltage waveform of the Fig. 2(a) is given by
A. Model : Estimation of
From (1)–(4) with dc supply the following regression
model may be written to estimate

(6)
In the above expression a term like means that
(7)
and are employed in this sequence for . Note that all
(8) the active vectors are used as well as the null vector. Fig. 2(c)
shows the voltage vectors generated by the inverter as well as
their sequencing (1 to 18) which correspond to the numbering
B. Model : Estimation of , and of Fig. 2(a).
Solving the model (1)–(4) for , assuming that the ma- With this waveform, the average stator voltage is zero at each
chine is at standstill and that is known (e.g., ob- interval, the average electromagnetic torque is close to zero
tained by Model ) one may derive the following regression and consequently the machine stays at standstill. In fact, in every
model: interval the average voltage applied is zero or pulsating,
i.e., the negative sequence equals the positive sequence. More-
(9) over, this voltage waveform has enough frequency harmonics to
persistently excite the machine. Note that the waveforms of the
other phases and are similar but phase-shifted of
(10)
rads to provide three-phase symmetry.
Fig. 3 presents the frequency spectrum of the voltage wave-
(11) form shown in Fig. 2(a). It is considered that the fundamental
frequency of the special waveform is 5 Hz. Observe that the
where is the leakage factor which is given by spectrum is concentrated around the fundamental frequency of
, and . the waveform.
From the parameter vector given in (11) one may determine The signals , , , , and are obtained by
, and by measuring the three-phase quantities. The signals are filtered
by second order analog filters and converted to digital words
(12) through analog–digital converters. The derivative terms ,
, , , , , and are obtained by digital
JACOBINA et al.: ESTIMATING THE PARAMETERS OF INDUCTION MACHINES AT STANDSTILL 87

Fig. 4. Sensitivity of the Model B . Point 0 . Solid 0 . Dashdot 0 .

TABLE I
MACHINE PARAMETERS

Equation (15) represents an analog third-order low-pass filter

where is the cut-off frequency of the filter.

VI. PARAMETER SENSITIVITY

The analysis of the sensitivity of the estimation models is
very important for determining how easy is to estimate a given
parameter.
The output variable of the proposed estimation models
is the weighted sum of the parameters. The weights are the
Fig. 2. Voltage waveforms and dq voltage plane. components of the regression vector like shown, for instance,
in (5). Then, , that represents the sensitivities with
respect to the parameters, is directly determined from these
weights. In this section, the sensitivity of Model with respect
to , and is investigated. Fig. 4 shows

and

which weights the components of the parameter vector . As it

can be seen from Fig. 4, the sensitivity of the model with respect
to increases with the frequency and it is always higher than
the sensitivities with respect to and . This indicates that
for high frequencies it is easier to estimate than and .
However, for low frequencies the sensitivities with respect to
Fig. 3. Frequency spectrum of the voltage waveforms proposed. , and are almost the same.

VII. EXPERIMENT RESULTS

filtering. These filters are designed by obtaining the dis-
crete-time equivalent of The experimental set-up is shown in Fig. 1. The drive system
in controlled through a PC-Pentium equipped with dedicated
plug-in boards to measure currents and voltages. The drive
(15)
system has five analog channels which permits to acquire
88 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 17, NO. 1, MARCH 2002

three phase currents and two of the three phase voltages. The
estimation algorithm employed to process the experimental
data was the recursive least-squares with the sampling time
equal to 50 s.
Table I present the complete set of parameters of the machine
used in the experimental tests and in the previous parameter
sensitivity study. These values were determined by executing
the standard test procedures, i.e., the no-load and locked rotor
tests.
The experimental tests have been carried out in the following
way. First of all, the is determined with Model . Secondly,
the machine is supplied by the special waveform voltage pre-
sented in Fig. 2(a) and Model is employed to determine the
other parameters.
Fig. 5 shows the waveforms of the stator currents as ob-
served in the test with special voltage waveform. As it can be
seen from the phase plane , the current vector changes
from one direction to another passing by zero, like the voltage
vector showed in Fig. 2(c).
The parameters estimated with Model , Model and those
determined by the standard tests are presented in Table II. Note
that the estimated parameter are consistent with the standard
parameters.

VIII. CONCLUSION
This paper has presented a method to determine the contin-
uous-time parameters of induction machines which are neces-
sary to implement the field-oriented control strategy, i.e., ,
, , and . The stator resistance is obtained from the dc ma-
chine model. The stator transient inductance, the stator induc-
tance, and the rotor time constant, are obtained by using the dy-
namic machine model and by supplying the induction machine
with a special exciting waveform. This voltage waveform pro-
vides persistent excitation while keeps the machine at standstill
without requiring the use of any mechanical apparatus. The pro-
Fig. 5. dq currents waveforms and dq current plane. posed approach does not require any device for speed measure-
ment and may be embedded within the drive software providing
TABLE II a fully automated machine characterization procedure. So, the
COMPARISON OF THE PARAMETERS DETERMINED VIA STANDARD TEST present method is indicated to be used in the drive autocommis-
PROCEDURES WITH THE ESTIMATED PARAMETERS
sioning phase. The experimental results have demonstrated the
feasibility of the proposed method.

REFERENCES
[1] M. Vélez-Reyes, K. Minami, and G. C. Verghese, “Recursive speed and
parameter estimation for induction machines,” in Proc. IAS Conf. Rec.,
1989, pp. 607–611.
[2] J. Holtz and T. Thimm, “Identification of the machine parameters in a
vector-controlled induction motor drive,” IEEE Trans. Ind. Applicat.,
vol. 27, pp. 1111–1118, Nov./Dec. 1991.
[3] J. Stephan, M. Bodson, and J. Chiasson, “Real-time estimation of the
parameters and fluxes of induction motors,” IEEE Trans. Ind. Applicat.,
vol. 30, pp. 746–759, May/June 1994.
[4] D. E. Borgard, G. Olsson, and R. D. Lorenz, “Accuracy issues for pa-
rameter estimation of field oriented induction machine drives,” in Proc.
IAS Conf. Rec., 1994, pp. 593–600.
simultaneously five analog signals with resolution of 10 bits. [5] C. B. Jacobina, L. A. de S. Ribeiro, and A. M. N. Lima, “Parameter
Each analog channel consists of a Hall effect sensor followed estimation of induction machines under sinusoidal excitation,” in Proc.
by a pre-conditioning circuitry, an anti-aliasing filter and analog IEMDC Conf. Rec., 1997, pp. TD2–5.1:TD2–5.3.
[6] J. R. Willis, G. J. Brock, and J. S. Edmonds, “Derivation of induction
to digital converter. For the presented experimental tests four motor models from standstill frequency response tests,” IEEE Trans.
analog channels have been employed to measure two of the Energy Conv., vol. 4, pp. 608–613, Dec. 1989.
JACOBINA et al.: ESTIMATING THE PARAMETERS OF INDUCTION MACHINES AT STANDSTILL 89

[7] J.-K. Seok, S.-I. Moon, and S.-K. Sul, “Induction parameter identifica- João Edgar Chaves Filho was born on March 13, 1953, in Boa Vista, Brazil. He
tion using pwm inverter at standstill,” IEEE Trans. Energy Conv., vol. received the B.Sc. degree in electrical engineering from Federal University of
12, pp. 127–132, June 1997. Amazonas, Amazonas, Brazil and the M.Sc. degree from Federal University of
[8] L. A. de S. Ribeiro, C. B. Jacobina, and A. M. N. Lima, “The influence Paraíba, Paraíba, Brazil, in 1981 and 1991, respectively. He is currently pursuing
of the slip and the speed in the parameter estimation of induction ma- the Ph.D. degree at Federal University of Paraíba.
chines,” in Proc. PESC Conf. Rec., June 1997, pp. 1068–1074. Since 1980, he has been a Faculty Member with the Electrical Engineering
[9] IEEE Standards Board, “Electric machinery,” in IEEE Standard Department at the Federal University of Amazonas. His research interests in-
Test Procedure for Polyphase Induction Motors and Genera- clude electrical drives and system identification.
tors. Piscataway, NJ: IEEE Press, 1995.
[10] L. Ljung, System Identification: Theory for the User, 1st
ed. Englewood Cliffs, NJ: Prentice-Hall, Inc., 1987.

Antonio Marcus Nogueira Lima (S’77–M’89) was born on March 15, 1958,
Cursino Brandão Jacobina (S’78–M’78–SM’98) was born on April 17, 1955, in Recife, Brazil. He received the B.Sc. and M.Sc. degrees in electrical engi-
in Correntes, Brazil. He received the B.Sc. degree in electrical engineering from neering from Federal University of Paraíba, Paraíba, Brazil, in 1982 and 1985,
Federal University of Paraíba, Paraíba, Brazil, and the Diplôme d’Etudes Ap- respectively. He received the Ph.D. degree from Institut National Polytechnique
profondies (DEA) and Ph.D. degrees from Institut National Polytechnique de de Toulouse, Toulouse, France, in 1989.
Toulouse, Toulouse, France, in 1978, 1980, and 1983, respectively. Since September 1983, he has been a Faculty Member in the Electrical Engi-
Since 1978, he has been a Faculty Member in the Electrical Engineering De- neering Department at Federal University of Paraíba. His research interests are
partment of the Federal University of Paraíba. His research interests include in the fields of electrical machines and drives, electronic instrumentation, con-
electrical drives, power electronics, control systems, and system identification. trol systems, and system identification.