You are on page 1of 6

Structural Identifiability Analysis of Steady-State

Induction Machine Models

Ahmed M. Alturas, Shady Gadoue Bashar Zahawi A. S. Abdel-Khalik


and Mohammed A. Elgendy Electrical & Computer Engineering Electrical Engineering Department,
School of Electrical and Electronic Khalifa University of Science Alexandria University, Alexandria,
Engineering, Newcastle University Technology & Research (KUSTAR) Egypt
Newcastle upon Tyne, UK Abu Dhabi, UAE
Ahmed.alturas@ncl.ac.uk bashar.zahawi@kustar.ac.ae

Abstract-Many mathematical models have been developed to If the used model was unidentifiable, one of the infinite
describe the dynamic behaviour of induction machines and have mathematical solutions will be obtained depending on the used
been utilized in induction machines parameter identification. In algorithm [6]. Although each parameter of an 1M has an
some cases, model parameters may not be uniquely estimated, importance in the modelling, some parameters may not reliably
regardless of the used algorithm and the quality and quantity of estimated based on a given measurements by many reasons
the used measurements. This non-identifiability is related to the
including poor algorithm and model structure. If the model
structure of the model itself. In this paper, the structural
parameters are unidentifiable, there will always be
identifiability of three commonly used steady-state induction
combinations of these parameters. Therefore, it is important to
machine models (the standard T-model, the inverse r-model and
find the unidentifiable parameters of the model so they can be
the r-model) is investigated. Such analysis deals with the
excluded from the parameter vector by eliminating them or
uniqueness of the solution for the unknown model parameters
and is, therefore a prerequisite for induction machine parameter
fixing their values to proper ones.
identification. Two structural identifiability techniques, the In general, identifiability analysis can be done either
transfer function and bond graph, are reviewed and applied for structurally or practically. Approaches that analytically study
testing the identifiability of the three models. The results show
the model equations, no attention is paid for any restrictions
the importance of identifiability analysis before performing
related to the model operation, are called structural (a priori)
parameter identification. Structural identifiability investigation
methods. A structural non-identifiability arises when there are
confirms the non-identifiability of the T-model and, on the other
redundant parameters in the model structure, and can be
hand, the global identifiability of both the inverse r- and r­
removed analytically by introducing new restrictions, e.g. an
models.
identifiable combination of some unidentifiable parameters [7].
Keywords-component; Induction machine; Equivalent circuit, In practical (posteriori) identifiability analysis, identifiability is
Structural identifiability; tested by fmding out if the measured information are enough to
estimate the parameter reliably or not using simulated or
measured data [8]. The analysis of structural identifiability of
I. INTRODUCTION
mathematical models has received major attention in recent
Structural identifiability analysis should be considered years [9]. Various approaches exist to detect structural non­
when designing a model for induction machine parameter identifiability by a priori analyzing of the system model [10,
identification. A structural identifiability analysis determines 11]. More specifically two techniques, the transfer function and
whether the induction machine unknown parameters can be the bond-graph, have been applied to 1M identifiability
uniquely determined from the input/output measurements. In analysis. The analysis was applied only to the T-model and
the literature, considerable progress has been made towards results showed that the conventional machine T-model
deriving different models of induction motors (lMs) and many parameters are not uniquely identifiable [4, 12]. However,
approaches have been introduced for parameter identification these approaches have not been employed to assess the
[1, 2]. identifiability of other machine models.
A standard approach for IMs parameter identification is to In this paper, structural identifiability of three induction
estimate all parameters of the IMs based only on the external motor equivalent circuits namely T-model, inverse r-model
measurements of voltage, current, speed and/or torque [3-5]. In and r -model is investigated. Two techniques based on the
such approaches, it is essential to ascertain whether the transfer function and the bond-graph representing structural (a
unknown parameters are uniquely identifiable before collecting priori) identifiability analysis are utilized. These techniques are
measurement data. Any divergence between the true and the first reviewed and applied for the identifiability analysis of the
estimated model parameter values may result in a wrong machine T-model. The same techniques are utilized for the first
assessment of the machine performance and lead to time to assess the identifiability of the other 1M models
deterioration in its behaviour. (inverse r - and r -models).

978-1-4673-9130-6/15/$31.00 ©2015

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on December 01,2022 at 12:09:37 UTC from IEEE Xplore. Restrictions apply.
II. INDUCTION MOTOR STEADy-STATE MODELS

Due to their simplicity and shorter computation time,


steady-state 1M models have gained more acceptances in 1M
parameter identification [9]. The equivalent T-model, Fig. 5, is
the most 1M model that can be found in the literature and is
becoming increasingly used in different applications including
1M parameter identification [13]. This model consists of five
electrical parameters: R" R" II" Ilr and Ln" where Rs is the stator
resistance, Rr is the rotor resistance (referred to the stator), lis Fig. I. Induction motor T-equivalent circuit.
and Ilr are the leakage inductances for stator and rotor
respectively, Lm is the magnetising inductance and s is the slip
given by (ns-nr)/n" where ns the stator field speed and nr is the
rotor mechanical speed. For T-model the relationship between
the flux linkages (lJf/, lJf/), and the winding currents (i/, i/) is:

R,./s
(I)

A simple change to obtain two mathematically equivalent


Fig. 2. Induction motor Inverse f- equivalent circuit.
circuits with only two inductances that give the same
performance as the T-equivalent circuit have been proposed
[6]. These equivalent circuits are known as inverse f-model
(Fig. 2) and f-model (Fig. 3) of the 1M. As long as the input
voltage is the same, the two models have the same input
impedance and produce the same torque as the T-model.
The relationship between the flux linkages and the winding
currents of inverse f-model is given by:

(2) Fig. 3. Induction motor f- equivalent circuit.

The relations between the parameters of the T-model and


III. STRUCTURAL IDENTIFIABILITY ANALYSIS
the inverse f-model are as follow [6]:
ApPROACHES
a = Lm / Lf' Models describing dynamic properties of 1M are usually
I;s = Ls -aLm' defined as a system of differential equations and can be written
(3) in the form of state-space. The dynamic model of IMs
Lm =aLm, depending on a parameter vector PE R"' is described by:
,
2
Rr =a Rr
x{t) A{P)X{t) + B{P)V{t)
=
(6)
From Fig. 3, the relationship between the flux linkages and
Y{t) C{p)X{t)
=

the winding currents of f-model is given by:


where A, Band C are matrices of proper dimensions and each
(4) of them consisting of some or all of unknown parameters. The
vector P is said to be globally identifiable if [14]:

The relations between the parameters of the T-model and


the f- model is given by [6]: (7)

a = Ls / Lm,
"
2
Ilr =a Lr -aLm, Implies that P = P 'if P, PE �n
(5)
L�, =aLm' The structural identifiability term means whether it is
"
2
possible to recover the parameter P vector uniquely from
Rr =a Rr input/output measurements, Vet), Yet), tE [O,T], or not. In other
words, the parameter vector P is identifiable if and only if any
where Ls and L r are the self-inductances of the stator and rotor, change in the values of the parameters results in a change in the
respectively. measured quantities Yet).

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on December 01,2022 at 12:09:37 UTC from IEEE Xplore. Restrictions apply.
Based on the structural identifiability analysis results, In this approach, the system can be represented by lines and
models can be categorized into three different classes; globally symbols identifying the power flow paths of the system by a
identifiable where all the unknown parameters in the model can combination of efforts and flows. Any interface between two
be uniquely determined, locally identifiable when one or more elements in the bond graph is known as a port. In electrical
of the parameters in the parameter vector may take more than domain, port variables of the bond graph are the voltage over
one of a [mite number of values, and unidentifiable where there the element port and the current through the element port.
is one or more unidentifiable parameters in the parameter
The bond graph has five types of elements; two active and
vector P. In order to uniquely identify all parameter for
three passive. The two active elements are the bond graph
unidentified models, a solution should be made including
sources i.e. effort source SE (voltage) and flow source SF
model reduction, reparameterization, and measurements
(current). The remaining three are represented by an R, I and C
increasing [15].
for resistive, inductive and capacitive elements, respectively.
Recently, some work has been carried out to obtain general Each of these elements has a single power bond attached (line
criteria for a priori identifiability. Various methods have been with the element at the end of the bond) showing the exchange
proposed in the literature to analyze the identifiability of of the power at one location.
system models [16, 17]. Among structural identifiability
Power bond may join in one of two types of junctions
techniques, two approaches based on the transfer function [18]
called I-junction and O-junction. I-junction represents
and the bond-graph [12, 19] are introduced.
locations in the circuit with common current flow, where 0-
In this section, a brief discerption is given for the two junction represents nodes of the circuit where voltages are
techniques (the transfer function and bond-graph). same. Consequently, series connected elements are connected
at I-junction while parallel branches are connect at O-junction.
A. The Transfer Function Approach A power flow diagram is created by connecting the elements in
an energy conserving mode by means of junctions and nodes.
A simple approach can be used if the transfer function of
Fig. 3 shows how to generate the bond-graph from a physical
the model can be computed as a function of its parameter
model.
vector P. The complete dynamic model can be described as,
recalling Equation (5):

X(t) A (P ) X(t) +B(P ) U(t)


=
(8)
Y(t) C(p ) X(t) +D (P ) U (t)
=

where X is the state vector, U is the input vector, Y is the Detennine the system physical domain and identify basic elements
output vector, A is the state matrix, B is the input matrix, C is like (e.g. resistors (R), inductors (I), ...).
the output matrix, and D is the feed forward matrix.
Any dynamical system can be completely characterized by
its transfer matrix G. The transfer function for this model, (8), Give a unique name of the identified elements in the previous step.

is given by:
Identify reference effort (voltage in electrical domain).
(9)

In this approach, the transfer functions are written in a Identify all other efforts with different names.
canonical form, common factors in numerator and denominator
are cancelled and the transfer function is simplified so that the
Identify all effort/flows differences needed to connect the ports of the
coefficients of the higher power of s in the denominator is elements.
always one. After this simplification, all the transfer
coefficients are often referred to as moment invariants. The t
identifiable parameters are the parameters that can be uniquely Construct the effort differences using a I -junction and flow difference
deduced from the coefficients of the transfer function matrix with O -junction
[20]. If it is not possible to uniquely determine the parameters
from the transfer function coefficients, it is essential to re­
Connect the ports of all elements obtained in step with 0 and 1-
arrange or to use another model [16, 18].
junctions for corresponding efforts and flows differences.

B. Bond-Graph Approach
Bond graph maps the power flow through a system and
describe the relationships between the conjugate variables
I Simply the resulted graph.
I
(voltage and current in electrical domain) in each branch of the [ End I
system. The graphical nature of the method enables the
characteristics of the model to be easily visualized and to Fig.3 How to generate a bond-graph model from a physical model.

determine whether or not the model is appropriate for the


parameter identification process [21].

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on December 01,2022 at 12:09:37 UTC from IEEE Xplore. Restrictions apply.
Bond graphs have a concept called causality that indicates TABLE I. Numerical example of different parameter sets of the IM
the direction of the effort and flow for each bond of the power
Parameters Set 1 Set 2 Set3
flow diagram and identifies the causal relationships between all
variables. For each bond, causality is identified by the causal R, 7.987959 n 8.999628 n 4.998228 n
stroke (end-bar) which is independent of the power flow
R, 7.193455 n 2.999839 n 2.091855 n
direction. The effort information moves toward the causal
stroke and the flow infonnation moves away from it. At 0- I" 0.011419 H 0.137927 H 0.149678 H
junction, one of the bonds sets the effort for the others.
I" 0.048195 H 3.2IE-03 H 1.10E-02 H
Consequently, only one causal stroke is on the O-junction,
while the others are away from it. At I-junction, one bond sets L/II 0.355336 H 0.230314 H 0.221085 H
the flow; indicates that only one bond has the causal stroke
IZ,ql 115.63 n 115.82 n 115.22 n
away from the I-junction, while the others are on it.
LZeq 80.870 80.390 80.840

IV. IDENTIFIABILITY ANALYSIS OF 1M T-MODEL


In this section, the identifiability of the T-model of the 1M B. Bond-Graph Approach
is investigated using the two aforementioned discussed Fig. 4 shows the bond-graph of the T-model. The half
approaches. arrows show the direction of power flow. The causal strokes
are set in accordance with the procedure discussed above. For
A. The Transfer Function Approach example, at I-junction, the effort (vs) of source SE moves
towards its causal stroke while the flow ( is ) moves away from
Mathematically, the transfer function of the T-model shown
its causal stroke. The inductor element imposes flow, hence
in Fig. 1 is given by the admittance (G(s) = J(s)/V(s) = lIZeq)'
can be modeled as a source with a causal stroke at the element
The input impedance of this model given by:
side. For a resistive element, causal stroke can go on both
2 directions in such a way to satisfy the junction at the other end
p sa + PS b + !X + d of the bond. For a proper causal completion, the causal strokes
(1 0)
for Rs and Rr are set on the near of I-junctions.
Zeq ( p, s)
(psb / d + 1)
=

It is essential that one bond imposes an effort on each 0-


where p=jws , and a, b, c and d are functions of the five
junction (i.e. one causal stroke is on the O-junction). This has
electrical parameters that can be obtained from the following
expressions: not been realized in the bond graph of the T-model shown in
Fig. 4. With such a causality conflict, it is not possible to
construct a proper bond-graph for the circuit that can be used in
a 'Ir Ols + Lm ) + '15 Lm ' b
Rs Olr + Lm ) parameter identification.
Rr Rr
= =

(11)
c =
'Is + Lm, d Rs = V. IDENTIFIABILITY ANALYSIS OF 1M INVERSE r-MODEL

Similar to the T-model, the identifiability of the inverse r­


Taking Laplace transfonns:
model is tested using the two techniques (the transfer function
and bong-graph).
G(s)
_1_ (b / ad)s + (1/ sa) (1 2)
2
+ (sb + c ) / sa + d / sa
= =

Ze q S A. The Transfer Function Approach


The input impedance of the inverse r -model in Fig. 2 as a
Although the T-model has five physical parameters, only
function of the slip s is given by:
four different coefficients (a, b, c and d) can be uniquely
2
detennined from the input/output measurements defined by
Z �q ( p, s )
p s + psa] / b] + !X] + aId] (13)
G(s). Different combination for a, b, c, and d will give the =

same input-output relation for any slip and supply frequency. psi bl +dl

where p=jw" and ai, bl, CI and dl are functions of the four
Any parameter can be expressed by other parameters is not
identifiable and vice versa. Thus, an infinite number of
equivalent circuits of the T-model which are theoretically electrical parameters of the model and they can be obtained
equivalent if the external measurements are used exist. from the following expressions:

R : Rs R,.
R:--
Table I provides a numerical example of this situation
where different parameter sets of T-model give the same total
i s

L
Causality
impedance (Zeq). The table is obtained using a nonlinear curve­ is c��!�ict. i,.
v is '. i,.
.

fitting technique in conjunction with measured steady-state

isl ; {' li,.


SE: Vs �l� O � l
"

machine current, voltage and speed based on a 1.1 kW three­


phase induction 1M. Although the parameters are totally is

different, the table shows a very good agreement between the


total impedance of the three sets at a machine speed of 1491 I: lis I: Lm I: It,.
rpm (slip of 0.006). Fig. 4. T-model causal bond graph.

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on December 01,2022 at 12:09:37 UTC from IEEE Xplore. Restrictions apply.
al =
Rs' bl I;s,
=
R:Rs
(14)
c, =
R;. (l;s +Lm) /l;sLm' d, =
R� /1;sLm

The transfer function of this model is given by:


Ss /bl +dl
G(s) 1/Z�q 2 (15)
S +Ssal /bl +SCI +al dl
= =

S
I://s' I:Lm'
With inverse f-model, there are four coefficients of G(s) Fig. 5. Bond-graph of IM Inverse r-equivalent circuit.
that can be uniquely determined if the external measurements
are used. Model parameters can then be uniquely identified
from these coefficients by using (14).

B. Bond-Graph Approach
To avoid the causality conflict occurred in the bond graph
of the T-model, the stator leakage inductance is combined
together with that of the rotor as suggested in [12]. Fig. 5
shows the bond graph of the inverse f-model where the
parameter redundancy has been removed and a proper bond
graph is obtained. I:Lm" I:ltr"
Fig.6. Bond-graph of IM r-equivalent circuit.
VI. IDENTIFIABILITY ANALYSIS OF 1M f-MODEL
In this section, the structural identifiability of the f-model VII. DISCUSSIONS
is tested using the two aforementioned techniques.
When the induction motor has been already manufactured,
its parameters are determined based on the data measured from
A. The Transfer Function Approach
the motor. Standard IEEE tests including No-load, DC, and
The input impedance of the f-model in Fig. 3 as a function locked rotor are made and the measurements of the current,
of the slip s is given by: voltage, and power at the stator terminals are collected [22].
2 However, it should be pointed out that it is impossible to
Z"eq ( p, s) =
p s +PSa2 +b2c2
(16) determine I,s and I'r separately from the measurements at the
rvl 2 +C2 stator terminals. One way to determine them separately is to
f-'U
assume that to be equal, which is not always the true (e.g.
where p=jw" and a2, b], C2 and d2 are functions of the four faulty motor) and, thus based on their values other parameter
electrical parameters of the model and they can be obtained values might change in order to produce the same performance
from the following expressions: (e.g. total impedance). On the other hand, for the inverse f­
and f-models, circuit parameters can be uniquely calculated
_ RsL�, +R; Rs +L�, R; /s b
a2 - , 2 - Rs' based on the standard IEEE tests without any assumptions.
L"ml'r" (17)
As can be seen from T-model, Equation (1), the three
c2 =
R; /l;�L�" d2 (L�, +l;� )/l;�L�,
= dependent inductances (/,,, I'r and Lm) can be described only
by two independent inductances and, thus the 1M can be
modelled with an infmite of different parameter sets. For
The transfer function of this model is given by:
inverse f- and f-models, however, the inductances can be
Sd2 +c2 independently modelled as these models have only two
G(s) 11 Z�q = =
2 (18) dependent inductances, Equations (2) and (4).
S s +Ssa2 +b2c2
In this paper, structural identifiability of T-, inverse f- and
With f-model, there are four coefficients of G(s) that can f-models of the 1M is investigated using the transfer function
be determined if the external measurements are used. Model and bond graph. The transfer function of the T-model has only
parameters can then be uniquely identified from these four coefficients which are functions of the five unknown
coefficients by using (17). electrical parameters of the model. Although the values of the
four coefficients can be estimated from the experimental
B. Bond-Graph Approach measurements, these four coefficients are not sufficient to
In this model, the stator leakage inductance is combined mathematically calculate the model parameters (five
together with that of the rotor as suggested to eliminate the unknowns). In contrast, the transfer functions of both the
conflict in the T-model in [6]. Fig. 6 shows the bond graph of inverse f- and f-models have four coefficients which are equal
the f-model where the parameter redundancy has been to the number of the unknown parameters. Model parameters
removed and a proper bond graph is obtained. can then be uniquely identified from these coefficients by using
(14) and (17) respectively. This means that both inverse f- and

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on December 01,2022 at 12:09:37 UTC from IEEE Xplore. Restrictions apply.
f-models are identifiable while the T-model is non-identifiable [8] A. Raue, C. Kreutz, T. Maiwald, 1. Bachmann, M. Schilling, U.
according to the transfer function approach. KlingmOller, and 1. Timmer, "Structural and practical identifiability
analysis of partially observed dynamical models by exploiting the
The bond graph representation of the T-model shows a profile likelihood," Bioinformatics, vol. 25, no. 15, pp. 1923-1929,
2014.
causality conflict at the O-junction. Accordingly, this model is
[9] A. M. Alturas, S. M. Gadoue, B. Zahawi, and M. A. Elgendy, "On the
not an appropriate model for parameter identification from Identifiability of Steady-State Induction Machine Models Using
external measurements. No such a causality conflict occurs in External Measurements," iEEE Trans. Energy Convers., vol. PP, no.
the bond graph representation of the inverse f- and f-models. 99,pp. 1-9,2015.
In fact, this conflict is eliminated by combining the dependent [10] S. Peilin, T. Jun, W. Shuang, and Z. Ning, "Identifiability Analysis of
storage element together in the equivalent circuit. Local Oscillator Phase Self-Calibration Based on Hybrid Cramer-Rao
Bound in MIMO Radar," iEEE Trans. Signal Process., vol. 62, no. 22,
pp. 6016-6031,2014.
VIII. CONCLUSION [11] C. Schranz, P. D. Docherty,C. Yeong Shiong, J. G. Chase, Mo, x, and
K. lIer, "Structural Identifiability and Practical Applicability of an
This paper presented a study of structural identifiability of Alveolar Recruitment Model for ARDS Patients," iEEE Trans.
three steady-state models (T-, inverse f- and f-equivalent Biomed. Eng., vol. 59,no. 12,pp. 3396-3404,2012.
circuits) of the induction motors. The identifiability of the three [12] P. Vaclavek, P. Blaha, and I. Herman, "AC Drive Observability
models has been investigated using the transfer function and Analysis," iEEE Trans. ind. Electron., vol. 60, no. 8, pp. 3047-3059,
2013.
bond graph as structural identifiability approaches. Results
[13] E. Laroche,and M. Boutayeb,"Identification of the Induction Motor in
show that the induction machine T-model is non-identifiable Sinusoidal Mode," iEEE Trans. energy Convers., vol. 25,no. I, pp. 11-
whilst both the inverse T- and T-models are. 19,2010.
[14] M. Milanese, and G. P. Molino, "Structural identifiability of
compartmental models and pathophysiological information from the
ACKNOWLEDGMENT
kinetics of drugs," Mathematical Biosciences, vol. 26,no. 1-2, pp. 175-
A. M. Alturas would like to acknowledge the Libyan 190,1975.
[15] S. Y. A. Cheung, O. Majid, J. W. T. Yates, and L. Aarons, "Structural
Ministry of Higher Education and Scientific Research and identifiability analysis and reparameterisation (parameter reduction) of
Misurata University for sponsoring his Ph.D. project. a cardiovascular feedback model," European Journal of
Pharmaceutical Sciences, vol. 46,no. 4,pp. 259-271,2012.
[16] C. Cobelli, and G. Romanin-Jacur, "Controllability, Observability and
Structural Identifiability of Multi Input and Multi Output Biological
REFERENCES
Compartmental Systems," iEEE Trans. Biomed. Eng., vol. BME-23,
[I] M. Ojaghi, and S. Nasiri, "Modeling Eccentric Squirrel-Cage Induction no. 2,pp. 93-100,1976.
Motors With Slotting Effect and Saturable Teeth Reluctances," iEEE [17] 1. 1. DrSTEFANO, "On the Relationships Between Structural
Trans. Energy Convers., vol. 29,no. 3, pp. 619-627,2014. Identifiability and the Controllability, Observability Properties," iEEE
[2] L. Monjo, H. Kojooyan-Jafari, F. Corcoles et al., "Squirrel-Cage Trans. Automat. Contr, vol. AC-22,no. 4,pp. 652,1975.
Induction Motor Parameter Estimation Using a Variable Frequency [18] M. Hajek, "A contribution to the parameter estimation of a certain class
Test," iEEE Trans. Energy Convers., vol. PP,no. 99,pp. 1-8,2014. of dynamical systems," Kybernetika, vol. 8,no. 2,pp. 165-173,1972.
[3] L. Whei-Min, S. Tzu-Jung, and W. Rong-Ching, "Parameter [19] J. F. Broenik, " Introduction to physical systems modelling with bond
Identification of Induction Machine With a Starting No-Load Low­ graphs," SiE Whitebook on Simulation methodologies. University of
Voltage Test," iEEE Trans. ind. Electron., vol. 59, no. I, pp. 352-360, Twente,1999.
2012. [20] A. Raue, C. Kreutz, T. Maiwald, Klingmu, x, U. lIer, and 1. Timmer,
[4] A. M. N. Lima, C. B. Jacobina, and E. B. de Souza Filho, "Nonlinear "Addressing parameter identifiability by model-based
parameter estimation of steady-state induction machine models," iEEE experimentation," Systems Biology, iET, vol. 5, no. 2, pp. 120-130,
Trans. ind. Electron., vol. 44,no. 3,pp. 390-397,1997. 2011.
[5] C. Moons, and B. D. Moor, "Parameter identification of induction [21] P. J. Gawthrop, and G. P. Bevan, "Bond-graph modeling," iEEE
motor drives," Automatica, vol. 31,no. 8,pp. 1137-1147, 1995. Control. Syst. Mag., vol. 27,no. 2,pp. 24-45,2007.
[61 S. Yamamura, AC Motors For High-Peiformance Applications: [22] "IEEE Standard Test Procedure for Polyphase Induction Motors and
Analysis and Control, New York,USA: Marcel Dekker,Inc,1986. Generators," iEEE Std 112-2004 (Revision of iEEE Std 112-1996), pp.
[71 B.-K. Choi, and H.-D. Chiang, "On the Local Identifiability of Load 0_1-79,2004.
Model Parameters in Measurement-based Approach," Journal of
Electrical Engineering & Technology, vol. 4,no. 2,pp. 149-158,2009.

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on December 01,2022 at 12:09:37 UTC from IEEE Xplore. Restrictions apply.

You might also like