You are on page 1of 13

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO.

7, JULY 2012 2877

Copper and Core Loss Minimization for Induction


Motors Using High-Order Sliding-Mode Control
Jorge Rivera Domínguez, Christian Mora-Soto, Susana Ortega-Cisneros,
Juan José Raygoza Panduro, and Alexander G. Loukianov

Abstract—A novel nonlinear affine model for an induction mo- Starting from the pioneering work of Blaschke [1], field-
tor with core loss is developed in the well-known (α, β) stationary oriented control has been a classical control technique for
reference frame, where the core is represented with a resistance in induction motors. This control technique consists of applying
parallel with a magnetization inductance. Then, an optimal rotor
flux modulus is calculated such that the power loss due to stator, nonlinear state transformation and feedback for asymptotic
rotor, and core resistances is minimized, and as a consequence, the decoupling of the rotor velocity and rotor flux modulus along
motor efficiency is raised; therefore, this flux modulus is forced with proportional–integral control loops for each channel [2].
to be tracked by the induction motor along with a desired rotor Recently, various control approaches have been studied for per-
velocity by means of a high-order sliding-mode controller, the formance improvement. Active research areas include adaptive
supertwisting algorithm. Using a novel Lyapunov function, the
closed-loop stability of the system is demonstrated. Moreover, I/O feedback linearization [3], adaptive backstepping [4], [5],
a classical sliding-mode observer is designed for the estimation sliding mode [6]–[9], and artificial neural networks [10], [11],
of unmeasurable variables like rotor fluxes and magnetization among others. It is worth mentioning that all of these works and
currents. For the load torque, a Luenberger observer is designed. the references therein are based on a mathematical induction
The performance of the proposed controller is finally studied motor model that does not consider power core losses, implying
by simulation and experimental tests. It was observed that the
steady-state optimal flux signal corresponds to the load torque that the induction motor presents a low-efficiency performance.
profile. This fact suggests that the flux demand is the necessary one In order to achieve a high efficiency in power consumption,
to produce the electric torque that can cancel out the load torque. one must consider at least the power core losses in addition to
Index Terms—Induction motors, magnetic losses, variable copper losses and then designing a control law under conditions
structure systems. obtained when minimizing the power core and copper losses.
Nowadays, this problem has become an important issue of
study since, in the last two decades, it has been a global
I. I NTRODUCTION
need for better use of electric energy and considering that

I NDUCTION motors are widely used in industrial appli-


cations due to their simple mechanical construction, low
service requirements, and lower cost with respect to dc mo-
induction motors are the major electric energy consumers in
the industry; therefore, the efficiency improvement in induction
motors has also been another active research area. In the effort
tors that are also widely used in the industrial field. On the of maximizing the motor efficiency or reducing the power loss,
other hand, induction motors constitute a classical test bench the main works are toward a better physical design of the motor
in the automatic control theory framework due to the fact [12]–[14] and improving its performance [15]–[17] by means
that they represent a coupled multiple-input–multiple-output of closed-loop controllers. In the case of closed-loop controller
nonlinear system, resulting in a challenging control problem. designs, one must consider the power core loss, where the main
two approaches are the loss-model- and power-measure-based
methods [4]. In the first method, the losses are defined in terms
of induction motor state variables that are minimized by select-
ing a flux level using what is best known as the loss-model-
Manuscript received March 17, 2010; revised December 4, 2010,
March 5, 2011, and August 20, 2011; accepted August 30, 2011. Date of
based controller [18], [19]. In the second approach, also known
publication October 13, 2011; date of current version February 17, 2012. as search controllers, the flux is decreased until the electrical
This work was supported by the Programa de Mejoramiento del Profesorado input power settles down to the lowest value for a given torque
(México) under Grant PROMEP 103.5/09/1465. This paper was presented in
part at the 11th International Workshop on Variable Structure Systems, México
and velocity [20]–[22]. It is clear that the first method yields to
City, México, June 26–28, 2010. better results since it is based on mathematical analysis, while
J. Rivera Domínguez, C. Mora-Soto, and J. J. Raygoza Panduro are the second one cannot be reliable since it is realized by means
with the Departamento de Electrónica, Centro Universitario de Ciencias
Exactas e Ingenierías, Universidad de Guadalajara, Guadalajara 44430, of an experimental method. With respect to loss-model-based
México (e-mail: jorge.rivera@cucei.udg.mx; christianms3@hotmail.com; controllers, there are two main approaches for modeling the
juan.raygoza@cucei.udg.mx). core, as a parallel resistance or as a series one. In the model with
S. Ortega-Cisneros and A. G. Loukianov are with the Center for
Research and Advanced Studies (CINVESTAV), National Polytechnic a parallel resistance, it is fixed in parallel with a magnetization
Institute (IPN), Guadalajara Campus, Zapopan 45015, México (e-mail: inductance, increasing the four electrical equations to six in the
susana.ortega@gdl.cinvestav.mx; louk@gdl.cinvestav.mx). (α, β) stationary reference frame [23]. The model with a series
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. resistance is obtained from the parallel resistance model, where
Digital Object Identifier 10.1109/TIE.2011.2171170 simplifications as the elimination of the magnetization current

0278-0046/$26.00 © 2011 IEEE


2878 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 7, JULY 2012

are carried on [24]. Again, it is clear that the parallel resis- previous one [34]. In the following, these enhancements are
tance model can yield to better results since no simplifications enumerated.
are made. There are works related to the parallel resistance 1) The determination of the optimal rotor fluxes has been
model, like the work of Dong and Ojo [16], which lacks verified to properly meet conditions for the minimization
closed-loop stability analysis for the designed controller and of cost functions.
observer, or the work of Lim and Nam [17], where the currents 2) The observer’s injection and the control law are now
flowing through the resistance that represents the core are given in the form of theorems.
eliminated. 3) Remark 1 is introduced for the justification of working in
In general, these last two works do not deal with a nonlinear the (α, β) reference frame.
affine model of the induction motor with core loss. These works 4) Remark 2 is introduced for the proposing of a possible
lack formal procedures for designing control laws and lack sensorless control scheme.
classical Lyapunov stability analysis; moreover, the estimated 5) A block diagram of the closed-loop system is presented.
variables are not guaranteed to converge to the real ones, 6) A more detailed simulation study was carried on.
resulting in unclear and doubtable control algorithms. On the 7) An experimental study is now performed.
other hand, the fifth-order model formulated as a nonlinear 8) Appendix I is introduced in order to present the details
affine system has demonstrated to be a suitable model due to for obtaining the nonlinear affine model for the induction
the existence of several works that are based on such model, motor with core.
as the ones mentioned at the beginning of this introduction, 9) Appendix II is introduced for the proofs of theorems.
resulting in clear and formal control algorithms. Therefore, it 10) The separation principle is introduced in Appendix III.
is interesting to investigate the nonlinear affine model of the The rest of this paper is structured as follows. In Section II,
induction motor with core loss in order to be able to analyze it the motor model for induction motors with core loss is pre-
under the framework of nonlinear systems [25], [26], among sented in nonlinear affine form. The optimal rotor fluxes for
others. Concepts like equilibrium points, Lyapunov stability, reducing the core and copper losses are calculated in Section III.
center manifold theorem, and stability of perturbed systems, In Section IV, a supertwisting controller is designed for the
among other concepts, can easily be treated. Moreover, one can induction motor with core loss for the tracking of a desired rotor
exploit different control techniques like feedback linearization, velocity and the optimal rotor flux modulus. Simulations and
adaptive control, output tracking, adaptive observers, and robust experimental tests are presented in Section V, and finally, some
control, among others, that rely on this type of nonlinear comments conclude this work in Section VI.
systems, as one can find them in books [27]–[29], among
others.
On the other hand, the sliding-mode control technique [30] II. I NDUCTION M OTOR M ODELING W ITH C ORE L OSS
has been applied in the last years in several industrial, robotic, In the following, a core loss version of the induction motor
and automotive applications. Its popularity is due to its robust- will be presented as a nonlinear affine system. Under the
ness properties to plant parameter variations and to the fact assumptions of equal mutual inductance and linear magnetic
that the order of the system to be stabilized is reduced by one. circuit, the induction motor equations in the stator reference
The nature of the control action is of the discontinuous type; frame (α, β) are given by [34]
characterized by a high-frequency switching signal, this has the
advantage of not requiring a pulse width modulation (PWM) dω Tl
= η0 (ψα iβ − ψβ iα ) −
stage but, at the same time, presents the disadvantage that the dt J
high frequency is limited to the maximum frequency of the dψα
= −η4 ψα − N pωψβ + η4 Lm iα,Lm
operation of the actuator, resulting in small output oscillations dt
that can be harmful specially to mechanical systems; this phe- dψβ
= −η4 ψβ + N pωψα + η4 Lm iβ,Lm
nomenon is called chattering. A solution to this problem is the dt
high-order sliding-mode technique [31]. This control technique diα,Lm η1
maintains the same sliding-mode properties (in this sense, first- = −(η1 + η2 )iα,Lm + ψα + η2 iα
dt Lm
order sliding mode) with the advantage of eliminating the diβ,Lm η1
chattering problem. The actual disadvantage of this control = −(η1 + η2 )iβ,Lm + ψβ + η2 iβ
dt Lm
technique is that the stability proofs are based on geometrical
diα
methods since the Lyapunov function proposing results in a dif- = −(Rs η3 + η5 )iα − η1 η3 ψα
ficult task [32]. The work presented in [33] proposes quadratic- dt
like Lyapunov functions for a special case of second-order + (η5 + η1 η3 Lm )iα,Lm + η3 vα
sliding-mode controller, the supertwisting controller, making diβ
possible to obtain an explicit relation for the controller design = −(Rs η3 + η5 )iβ − η1 η3 ψβ
dt
parameters. + (η5 + η1 η3 Lm )iβ,Lm + η3 vβ (1)
In an effort of solving some of the disadvantages of the works
in [16] and [17], the authors have presented some interested where ω is the rotor velocity, vα and vβ are the stator voltages,
advancements in the work [34]. Then, the contributions of the iα and iβ are the stator currents, iα,Lm and iβ,Lm are the
present work are, in fact, the enhancements with respect to the magnetization currents, ψα and ψβ are the rotor fluxes, T l is
RIVERA DOMÍNGUEZ et al.: COPPER AND CORE LOSS MINIMIZATION FOR INDUCTION MOTORS 2879

the load torque, with N p as the number of pole pairs, Rs is the in the following forms:
stator resistance, Lm is the magnetizing inductance, and J is
the rotor moment of inertia. Moreover (Rs Lr + Rr Lm + Rc Lm )iα,Lm
ψα,o =
Rs + Rr + Rc
3Lm N p Rc Rc
η0 = η1 = η2 = Rc (Lr − Lm )(iα,Rc − iα,s )
2J(Lr − Lm ) Lr − Lm Lm +
Rs + Rr + Rc
1 Rr Rc
η3 = η4 = η5 = (Rs Lr + Rr Lm + Rc Lm )iβ,Lm
Ls − Lm Lr − Lm Ls − Lm ψβ,o =
Rs + Rr + Rc
where Lr = Llr + Lm and Ls = Lls + Lm are the rotor and
stator inductances, respectively, and Rr and Rc are the rotor Rc (Lr − Lm )(iβ,Rc − iβ,s )
+ .
and core resistances, respectively, with Lls and Llr as the stator Rs + Rr + Rc
and rotor leakage inductances, respectively.
Remark 1: It is well known that the induction motor rep- The Hessian matrix that results from (4) is determined as
resented in the (d, q) reference frame behaves similarly to an  3(R +R +R ) 
s r c
externally excited dc motor, i.e., the rotor velocity and rotor flux (Lr −Lm )2 0
subsystems are decoupled, implying that the produced torque is 3(Rs +Rr +Rc )
0 (Lr −Lm )2
proportional to a single current component, in this case, the one
in q. Also, it is well known that this decoupling is only valid that can easily be verified to be a positive definite matrix.
under the strong assumption that the rotor flux has reached a
steady-state constant value. Therefore, in order to avoid this
strong assumption and to cope with an unknown optimal rotor IV. S UPERTWISTING S LIDING -M ODE C ONTROL OF
flux reference signal, the (α, β) or stationary reference frame is I NDUCTION M OTORS W ITH C ORE L OSS
selected.
The control problem is to force the rotor angular veloc-
ity ω and the square of the rotor flux modulus ψm = ψα2 +
III. O PTIMAL ROTOR F LUX C ALCULATION ψβ2 to track some desired references ωr and ψm,r , respec-
FOR M AXIMUM E FFICIENCY tively, ensuring, at the same time, load torque rejection. This
control problem will be solved in this section, along with
The copper and core losses are obtained by the corresponding the design of an observer for the estimation of unmeasured
resistance and current relations. Therefore, the power loss in the variables.
copper and core is expressed as follows:
   
PL = 1.5Rs i2α,s + i2β,s + 1.5Rr i2α,r + i2β,r A. Controller Design
  In order to solve the posed control problem using the super-
+ 1.5Rc i2α,Rc + i2β,Rc (2) twisting sliding-mode approach, we first derive the expressions
of the tracking error dynamics z1 = ω − ωr and z2 = ψm −
where iα,Rc and iβ,Rc are the currents flowing through the
ψm,r which are the outputs to be forced to zero. For performing
resistance that represents the core. Since PL is a positive
a maximum efficiency in energy consumption, one assigns
definite function, this can be considered as a cost function and
then to be minimized with any desired variables; in this case, the 2 2
most suitable are the rotor fluxes. Let us assign Ψr = (ψα ψβ ) ψm,r = ψα,o + ψβ,o . (5)
as the rotor flux vector and Ψr,o = (ψα,o ψβ,o ) as the optimal
rotor flux vector to be determined. Now, by applying a classical Using (1), the error dynamics can be written as follows:
procedure for minimizing cost functions outlined in [35], one Tl
can calculate the optimal rotor fluxes that minimize the cost ż1 = η0 (ψα iβ − ψβ iα ) − − ω̇r
J
function PL by meeting the following two conditions. First
   ż2 = −2η4 ψm + 2η4 Lm (ψα iα,Lm − ψβ iβ,Lm ) − ψ̇m,r .
∂PL  ∂PL ∂PL 
= = (0 0) (3)
∂Ψr Ψr =Ψr,o ∂ψα ∂ψβ Ψr =Ψr,o Note that the relative degrees are two for output z1 and three for
z2 ; therefore, we first propose a desired dynamic for z2 in the
i.e., the gradient is zero at the minimum, and second following form:
  ∂ 2 PL 
∂ 2 PL 
∂ 2 PL 
∂ψα 2 ∂ψα ψβ  −2η4 ψm +2η4 Lm (ψα iα,Lm −ψβ iβ,Lm )− ψ̇m,r = k2 z2 +z3
=  >0 (4)
∂Ψ2r Ψr =Ψr,o
2 2
∂ P
∂ψβ ψα
L ∂
∂ψ 2
P L

β Ψr =Ψr,o
with k2 < 0. The new variable z3 is then calculated as
i.e., the Hessian matrix is positive definite at the minimum. The
resulting optimal rotor flux components from (3) are obtained z3 = −2η4 ψm +2η4 Lm (ψα iα,Lm −ψβ iβ,Lm )− ψ̇m,r −k2 z2 .
2880 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 7, JULY 2012

 
Now, defining the vector ζ1 = (ζ1,1 , ζ2,1 )T = (z1 , z3 )T , its dy- 1
vβ = − kβ,1 |ζ3,2 |sign(ζ3,2 ) − vβ,1
namic, along with the one of z2 , using the equations in (1) η3
results as
v̇β,1 = −kβ,2 sign(ζ3,2 ) (10)
ż2 = k2 z2 + g2 ζ1
where kα,1 , kα,2 , kβ,1 , and kβ,2 are constant design parameters
ζ˙1 = f1 (ω, ψα , ψβ , iα,Lm , iβ,Lm ) + G(ψα , ψβ )ζ2 (6) and vα,1 and vβ,1 are integral actions. One can assume that the
functions f¯2,1 = f2,1 − iα,r and f¯2,2 = f2,2 − iβ,r are globally
where g2 = (0, 1), ζ2 = (iα , iβ )T , f1 (·) = (f11 (·), f12 (·))T , bounded by
and
  |f¯2,1 | ≤ δ1 |ζ3,1 |1/2 |f¯2,2 | ≤ δ2 |ζ3,2 |1/2 (11)
−η0 ψβ η0 ψα
G(ψα , ψβ ) =
2η2 η4 Lm ψα 2η2 η4 Lm ψβ
for some constants δ1 > 0 and δ2 > 0. If the controller gains
with f11 (·) = −T l/J − ω̇r and are selected to satisfy the following relations

f12 (·) = 4η42 ψm − 4η42 Lm (ψα iα,Lm + ψβ iβ,Lm ) 5δ1 kα,1 + 4δ12
kα,1 > 2δ1 kα,2 > kα,1
  2(kα,1 − 2δ1 )
η1
+ 2η4 Lm ψα −(η1 + η2 )iα,Lm + ψα 5δ2 kβ,1 + 4δ22
Lm kβ,1 > 2δ2 kβ,2 > kβ,1 (12)
2(kβ,1 − 2δ2 )
+ 2η4 Lm iα,Lm (−η4ψα − N pωψβ + η4 Lm iα,Lm )
 
η1 then the posed control problem is solved.
+ 2η4 Lm ψβ −(η1 + η2 )iβ,Lm + ψβ Proof: The proof is given in Appendix II. 
Lm
Analyzing the term f12 in (7), along with (1) and (5), results
+ 2η4 Lm iβ,Lm (−η4ψβ + N pωψα + η4 Lm iβ,Lm ) evidence that ψ̇m,r and ψ̈m,r depend on the control inputs and
+ 2k2 η4 ψm − 2k2 η4 Lm (ψα iα,Lm + ψβ iβ,Lm ) the derivative of the control inputs, respectively. In order to
avoid a control redesign, we propose to calculate ψ̇m,r and ψ̈m,r
+ k2 ψ̇m,r − ψ̈m,r . (7) using robust exact differentiators [36]. The well-known robust
exact differentiator structure is as follows:
For the stabilization of (6), one proposes a desired dynamic as
φ̇1 = κ1,1
ζ˙1 = f (·) + G(ψα , ψβ )ζ2 = K1 ζ1 (8)
κ1,1 = κ1,2 − λ1,1 |φ1 − ψm,r |1/2 sign(φ1 − ψm,r )
where K1 is a Hurwitz matrix. From (8), one can
calculate the current vector as reference signal, i.e., κ̇1,2 = −λ1,2 sign(φ1 − ψm,r )
ζ2,r = G−1 (·)(K1 ζ1 − f1 (·)) with ζ2,r = (iα,r , iβ,r )T . Due
φ̇2 = κ2,1
to det(G) = −2η0 η2 η4 Lm ψm , one can note that G−1 exists
under the obvious assumption that ψm = 0. The reference κ2,1 = κ2,2 − λ2,1 |φ2 − κ1,1 |1/2 sign(φ2 − κ1,1 )
current signal is such that, if circulating through the motor
stator windings, the output tracking errors z1 and z2 will decay κ̇2,2 = −λ2,2 sign(φ2 − κ1,1 )
asymptotically to zero. In order to force the stator currents to
where λ1,1 , λ1,2 , λ2,1 , λ2,2 > 0. Here, the input signal to be
be equal to the stator reference current signals, one defines
the variable ζ3 = (ζ3,1 , ζ3,2 )T = ζ2 − ζ2,r , where its dynamic, differentiated is ψm,r , and the outputs are ψ̇m,r = κ1,1 and
along with the ones in (6), results as ψ̈m,r = κ2,1 .

ż2 = k2 z2 + g2 ζ1
B. Observer Design
ζ˙1 = K1 ζ1 + G(ψα , ψβ )ζ3
In this section, we consider and fix the drawbacks of the
ζ̇3 = f2 (·) + η3 V − ζ̇2,r (9) proposed control (10). The first problem is the measurability
of the rotor fluxes and magnetization currents. This problem
with f2 (·) = (f2,1 , f2,2 )T , V = (vα , vβ )T , and is solved using a sliding-mode observer. The second problem
  concerns the estimation of the load torque, where a classical
−(Rs η3 +η5 )iα −η1 η3 ψα +(η5 +η1 η3 Lm )iα,Lm
f2 (·) = . Luenberger observer is designed.
−(Rs η3 +η5 )iβ −η1 η3 ψβ +(η5 +η1 η3 Lm )iβ,Lm
The proposed sliding-mode observer for rotor fluxes and
Theorem 1: Consider ζ3,1 and ζ3,2 as sliding functions and magnetization currents is proposed based on (1) as follows:
vα and vβ as supertwisting sliding-mode controllers
  dψ̂α
1 = −η4 ψ̂α − N pω ψ̂β + η4 Lm îα,Lm + ρα να
vα = − kα,1 |ζ3,1 |sign(ζ3,1 ) − vα,1 dt
η3
dψ̂β
v̇α,1 = −kα,2 sign(ζ3,1 ) = −η4 ψ̂β + N pω ψ̂α + η4 Lm îβ,Lm + ρβ νβ
dt
RIVERA DOMÍNGUEZ et al.: COPPER AND CORE LOSS MINIMIZATION FOR INDUCTION MOTORS 2881

dîα,Lm η1
= −(η1 + η2 )îα,Lm + ψ̂α + η2 iα + λα να
dt Lm
dîβ,Lm η1
= −(η1 + η2 )îβ,Lm + ψ̂β + η2 iβ + λβ νβ
dt Lm
dîα
= −(Rs η3 + η5 )îα − η1 η3 ψ̂α
dt
+ (η5 + η1 η3 Lm )îα,Lm + η3 vα + να
dîβ
= −(Rs η3 + η5 )îβ − η1 η3 ψ̂β
dt
+ (η5 + η1 η3 Lm )îβ,Lm + η3 vβ + νβ (13)
Fig. 1. Block diagram of the closed-loop control system.

where ψ̂α and ψ̂β are the rotor flux estimates, îα,Lm and îβ,Lm For the load torque estimation, we consider that it is slowly
are the magnetization current estimates, îα and îβ are the varying as that in [9] and [37], so one can assume that it is
stator current estimates, ρα , ρβ , λα , and λβ are the observer constant, i.e., Ṫ l = 0. This fact can be valid since the elec-
design parameters, and να and νβ are the injected inputs to the tric dynamic of the motor is faster than the mechanical one.
observer that will be defined in the following lines. Now, one Therefore, one proposes the following observer based on rotor
defines the estimation errors ψ̃α = ψα − ψ̂α , ψ̃β = ψβ − ψ̂β , velocity and stator current measurements:
ĩα,Lm = iα,Lm − îα,Lm , ĩβ,Lm = iβ,Lm − îβ,Lm , ĩα = iα −
îα , and ĩβ = iβ − îβ , whose dynamics can be expressed as dω̂
= η0 (ψ̂α iβ − ψ̂β iα ) −
T̂ l
+ l1 (ω − ω̂)
dt J
dψ̃α
= −η4 ψ̃α − N pω ψ̃β + η4 Lm ĩα,Lm − ρα να dT̂ l
dt = l2 (ω − ω̂) (17)
dt
dψ̃β
= −η4 ψ̃β + N pω ψ̃α + η4 Lm ĩβ,Lm − ρβ νβ where ω̂ and T̂ l are the rotor velocity and load torque estimates,
dt
respectively, and l1 and l2 are constant design parameters.
dĩα,Lm η1 Defining the estimation errors as eω = ω − ω̂ and eT l = T l −
= −(η1 + η2 )ĩα,Lm + ψ̃α − λα να
dt Lm T̂ l, one can determine the estimation error dynamic
dĩβ,Lm η1       
= −(η1 + η2 )ĩβ,Lm + ψ̃β − λβ νβ ėω −l1 − J1 eω ψ̃α iβ − ψ̃β iα
dt Lm = + η0 .
ėT l −l2 0 eT l 0
dĩα (18)
= −(Rs η3 + η5 )ĩα − η1 η3 ψ̃α
dt When the estimation errors for the rotor fluxes in (14) are zero,
+ (η5 + η1 η3 Lm )ĩα,Lm − να (18) reduces to
    
dĩβ ėω −l1 − J1 eω
= −(Rs η3 + η5 )ĩβ − η1 η3 ψ̃β ėT l
=
−l2 0 eT l
(19)
dt
+ (η5 + η1 η3 Lm )ĩβ,Lm − νβ . (14) where l1 and l2 can easily be determined in order to yield
limt→∞ eω (t) = 0 and limt→∞ eT l (t) = 0.
Theorem 2: Suppose that the stator currents are available An important issue is analyzed in Appendix III, the sep-
for measurements and that the injection observer signals are aration principle, that allows us to implement the designed
selected as controller with measured and estimated states. A block diagram
of the proposed control strategy is shown in Fig. 1.
να = lα sign(ĩα ) νβ = lβ sign(ĩβ ) (15) Remark 2: Sensorless control of electric machines is an
interesting problem to solve since it avoids the use of a me-
where lα and lβ are constant design gains. If these gains are chanical sensor for rotor position and velocity measurements.
selected to satisfy the following relations There is a clear advantage of sensorless control when there
  is no place or budget for such a mechanical sensor. Although
 
lα > η1 η3 ψ̃α − (η5 + η1 η3 Lm )ĩα,Lm  sensorless control is out of the scope for this work, we outline a
  procedure for rotor velocity estimation as shown in Fig. 2 based
 
lβ > η1 η3 ψ̃β − (η5 + η1 η3 Lm )ĩβ,Lm  (16) on the model (1). The first step is to calculate the stator fluxes
based on stator current and voltage measurements. In order to
then the estimation errors of the unmeasurable states, i.e.,  = eliminate the dc offset produced by integrators, one can follow
(ψ̃α ψ̃β ĩα,Lm ĩβ,Lm )T , tend asymptotically to zero. the work in [38]. Once the stator fluxes are retrieved, these can
Proof: The proof is given in Appendix II.  be used, along with stator currents, for the calculation of the
2882 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 7, JULY 2012

Fig. 3. Load torque considered in closed-loop simulations.

Fig. 2. Block diagram for rotor velocity estimation for a sensorless scheme.

magnetization currents. Then, a sliding-mode observer can be


used for stator current estimation, where the intention of this
observer is to extract the rotor fluxes by using the equivalent
control method. In a similar way, a sliding-mode observer can
be designed for rotor fluxes; then, by the equivalent control
method, one can calculate the flux terms multiplied by the
rotor velocity. Finally, using the calculated rotor fluxes, one can
estimate the rotor velocity.
Fig. 4. Open-loop performance. (a) Open-loop comparison of the rotor ve-
V. S IMULATIONS AND E XPERIMENTAL R ESULTS locity in the case of considering and not considering the core. (b) Open-
loop comparison of the rotor flux modulus in the case of considering and not
In this section, we first compare the proposed control scheme considering the core.
with different scenarios, by means of simulations, and then, we
describe the experimental results.

A. Simulation Results
In this part, we verify the performance of the proposed
control scheme by means of numeric simulations.
We consider an induction motor with the following nominal
parameters: Rr = 10.1 Ω, Rs = 14 Ω, Rc = 1 kΩ, Ls = 400 ×
10−3 H, Lr = 412.8 × 10−3 H, Lm = 377 × 10−3 H, and J =
0.01 kg · m2 .
Hence, η1 = 27 932.96 Ω/H, η2 = 2652.51 Ω/H, η3 =
43.47 H−1 , η4 = 282.12 Ω/H, and η5 = 43 478.26 Ω/H.
The gain parameters have been chosen for the controller Fig. 5. Closed-loop rotor velocity performance. (a) Rotor velocity tracking in
the case of not considering the core in the control design. (b) Rotor velocity
as kα,1 = 9100, kα,2 = 100, kβ,1 = 9000, kβ,2 = 100, K1 = tracking in the case of considering the core in the control design.
diag(−550, −100), and k2 = −3200. For the robust differen-
tiator, λ1,1 = 10, λ1,2 = 10, λ2,1 = 10, and λ2,2 = 10. For the transients in the model with core can be attributable to the fact
rotor flux and magnetization current observer, lα = 6, lβ = 6, that the start-up stator currents are rapidly increasing due to
ρα = −0.005, ρβ = −0.005, λα = −0.001, and λβ = −0.001. exponentially increasing magnetization currents. We focus at
For the load torque observer, l1 = −70, and l2 = −70. the steady-state values of the squared magnitude of the rotor
The load torque profile is shown in Fig. 3, while the reference fluxes; that for the fifth-order model is approximately 1.6 Wb2
velocity signal increases from 0 to 188.5 rad/s in the first 10 s while the model here obtained predicts a value of 0.16 Wb2 .
and then remains constant for subsequent time. When the rotor The obtained steady-state values are commonly used in many
velocity reaches a constant value, it is of interest to observe works (see, for instance, [3]–[6], [9], [10], [39], and [40]) as
the responses of the power lost and the square of the rotor flux a reference signal for the square of the rotor flux modulus;
modulus to load torque variations. therefore, these values are used in this work for the evaluation
The classical fifth-order model and the one obtained here of the power loss performance in each case, as shown in Fig. 8.
were open-loop simulated, where the outputs are shown in Meanwhile, Fig. 5 shows the performance of the rotor ve-
Fig. 4. Both systems predict the same steady-state rotor velocity locity in closed loop when simulating the seventh-order model
with different transient performances. The reason for faster with a fifth-order-model-based controller (without core) and
RIVERA DOMÍNGUEZ et al.: COPPER AND CORE LOSS MINIMIZATION FOR INDUCTION MOTORS 2883

Fig. 6. Closed-loop optimal squared rotor flux modulus performance in the Fig. 8. Comparison of the power loss in the copper and core using the optimal
case of considering and not considering the core in the control design. flux modulus and the steady-state open-loop values for the flux modulus
predicted in Fig. 4(b).

TABLE I
I NDUCTION M OTOR P ERFORMANCE TO S TATOR R ESISTANCE VARIATIONS

TABLE II
I NDUCTION M OTOR P ERFORMANCE TO ROTOR R ESISTANCE VARIATIONS

Fig. 7. Comparison of the power loss in the copper and core using an
optimal flux modulus when considering a fifth- and a seventh-order-model-
based controller design. Finally, one takes into consideration variations for rotor and
stator resistances. These parameters commonly suffer from
with a seventh-order-model-based controller (with core), where large variations with respect to their nominal values during op-
it can be appreciated a good tracking performance in both cases. eration due to heating. Therefore, in this work, one evaluates the
The controller design and optimal flux calculation in the case of closed-loop performance of the induction motor to resistance
not considering the core (fifth-order model) were similar to the variations. The simulation results are summarized in Tables I
ones designed in this work. The tracking of an optimal squared and II.
rotor flux modulus is shown in Fig. 6, where the simulation With Δz1 as the percent tracking error in rotor velocity and
conditions are similar to the ones above mentioned for Fig. 5. Δψm,r as the increment of the optimal rotor flux modulus with
Here, it can be appreciated that, in the case when considering respect to a steady-state value obtained without plant parameter
the core, the optimal flux levels result in lower values with re- variations, (ψm,r,ss = 0.0473 Wb2 ). From these tables, one can
spect to that when not considering it. Moreover, the optimal flux observe that the tracking of the rotor velocity performs well in
with core considerations selects an adequate value according to the presence of stator or rotor resistance variations. On the other
the load torque levels. This fact is evident due to the staircase- hand, the optimal rotor flux reference signal ψm,r significantly
shape flux level in Fig. 6 that is similar to the load torque profile increases in the presence of such variations. Nevertheless, the
in Fig. 3. induction motor can still perform well since the open-loop
On the other hand, when not considering the core, the flux value of the rotor flux modulus (in the fifth-order model) that
demand is greater than that when considering it. The exceeded is commonly used as a reference signal is, in this case, 3280%
flux will demand unnecessary higher current increments, im- (1.6 Wb2 ) far away from ψm,r,ss .
plying an increment in the power lost in the copper and core.
This fact is shown in Fig. 7.
B. Experimental Results
Now, by using the proposed control strategy, in Fig. 8, the
power lost in the copper and core is shown in the case of using The experimental setup is shown in Fig. 9. Here, the
the optimal flux modulus and the predicted open-loop steady- VARIAC is a three-phase variable transformer fed from a three-
state values (1.6 and 0.16 Wb2 ) whether considering or not phase voltage source. By rotating the knob of the VARIAC,
the core [Fig. 4(b)]. This is a common practice when dealing the amplitude of the three-phase voltage source is regulated.
with the control of the rotor flux in induction motors. From These voltages are fed to the power module (Semikron IGBT
this figure, one can observe a low power loss in the copper Power Electronics Teaching System) that incorporates a three-
and core when using the optimal flux; therefore, there is a phase rectifier and inverter; in that way, a regulated dc voltage
clear advantage over selecting steady-state open-loop rotor flux is available for the three-phase inverter. The control algorithm
modulus values. and space vector PWM generation are programmed in Simulink
2884 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 7, JULY 2012

Fig. 11. Estimated load torque in real time.

Fig. 9. Experimental setup.

Fig. 12. Real-time results for rotor velocity tracking. Zoom from 6.999 to 7 s
on the right.

1024 pulses per revolution of resolution. The board’s library


includes its own method for decoding the position from the
encoder digital signals. The velocity is then calculated by taking
the rate of change of the position w.r.t. the time and filtered
by a first-order Butterworth low-pass filter having an 8-rad/s
edge frequency, in order to attenuate the measurement noise.
The stator phase currents are measured by Hall-type sensors
(LEM HX 10-P). The sampling period has been fixed at Ts =
230 μs, which is an admissible value in applications, as already
observed in [37] and [41]. The motor was loaded at start-up by a
dc generator connected to a variable-power resistor. The applied
initial torque was set to 5 N · m, and by modifying the variable-
power resistor, it was possible to introduce a decrement of
Fig. 10. Block diagram of the experimental setup. 2 N · m at t = 7 s as Fig. 11 shows its estimation.
Fig. 12 shows the rotor velocity output tracking result. The
in Matlab and implemented with a DSP board (dSPACE 1104). reference velocity signal increases from 0 to 188.5 rad/s in
This board comes along with a library that easily incorporates the first 5 s and then remains constant for subsequent time. The
with Simulink. The signals from the phase currents are acquired transient and steady-state results are similar to the ones reported
by analog-to-digital converters included in the DSP board. in Section V-A.
The mechanical rotor position is provided by an encoder and Fig. 13 shows the rotor flux modulus output tracking re-
acquired by a dedicated encoder connector available in the DSP sults. The reference signal is the calculated optimal rotor flux
board. Once the DSP executes the control algorithm in each magnitude (5); meanwhile, the rotor flux modulus is calculated
sampling step, it generates six digital signals for switching the with estimated variables. One can appreciate the same behavior
insulated-gate bipolar transistors (IGBTs) of the inverter. These predicted by the simulation study. Once the rotor velocity has
digital signals are TTL level and are converted to a CMOS level reached a constant value at 5 s (see Fig. 12), the flux level also
of 15 V. This voltage level is the required one for switching on reaches a constant value; then, at 7 s, a step decrement in the
the IGBTs. A block diagram of this setup is shown in Fig. 10. load torque is introduced (see Fig. 11), and the optimal rotor
The induction motor has the same parameter values of the flux modulus responds by reducing its value in a step fashion.
digitally simulated motor in Section V-A. As mentioned earlier, This fact suggests that the flux demand is the appropriate one to
the motor rotor position is measured by an optical encoder with cancel out the load torque, implying that the stator currents will
RIVERA DOMÍNGUEZ et al.: COPPER AND CORE LOSS MINIMIZATION FOR INDUCTION MOTORS 2885

Fig. 13. Real-time results for rotor flux modulus tracking. Zoom from 6.999
to 7 s on the right. Fig. 15. Real-time results for stator currents and voltages. (a) Stator
current α. (b) Stator current β. (c) Stator voltage α. (d) Stator voltage β.

TABLE III
VARIATIONS IN THE L OAD T ORQUE , ROTOR F LUX
M ODULUS , AND P OWER L OST

and 13) and the power lost (see Figs. 7 and 14) are strongly
influenced by the load torque as previously discussed. It is
worth mentioning that, for the quantification of Fig. 14(a), a
Fig. 14. Real-time results for power loss in the copper and core. (a) When filter was applied to this signal.
using an optimal flux. (b) When using an arbitrary constant value for the Remark 3: It is worth mentioning that the DSP here used is
reference flux modulus.
for research purposes, and in the literature, it has been reported
demand reduced values as one can verify in Fig. 15(a) and (b). that similar sliding-mode controllers have been implemented
One can appreciate the same behavior shown in Section V-A, in commercial DSPs from Texas Instruments Incorporated at
i.e., the flux reference reduces its value when the load torque is a sampling time of 125 μs in [42] and 300 μs in [43], which
reduced. means that the recent advancements in digital microprocessor
With respect to the calculation of the power lost (3), the technology have rendered cheaper, simpler, and more flexible
corresponding power lost in the stator copper is determined the implementation of controllers.
with stator current measurements; meanwhile, the power lost
in the rotor copper is determined with the estimation of rotor VI. C ONCLUSION
fluxes and magnetization currents in order to calculate an
estimation of the rotor currents [see (22)]. Finally, the power In this paper, a representation of three-phase induction mo-
lost in the core is calculated using both measurements and es- tors with core loss has been obtained as a nonlinear affine
timations in order to calculate the currents flowing through the system. Based on this model, optimal reference rotor fluxes
resistance that represents the core. These currents are obtained have been determined for power loss minimization in copper
by reformulating the last two equations in (20) as follows: and core. On the basis of the obtained model, an observer-based
controller, which guarantees asymptotic reference tracking for
iα,Rc = iα,s + iα,r − iα,Lm the velocity and optimal flux in the presence of an unknown
iβ,Rc = iβ,s + iβ,r − iβ,Lm . load torque, has been designed using a supertwisting sliding-
mode controller along with a novel Lyapunov function for
Fig. 14 shows the power lost when using the optimal flux and classical stability analysis. With the simulation study, one could
a constant flux. One can observe that the power lost when using verify that the steady-state optimal flux shape corresponds
the optimal flux is lower than the one when using an arbitrary to the load torque profile. This fact suggests that the flux
constant value. demand is the necessary one to produce the electric torque
Fig. 15 shows the corresponding stator currents and voltages, that can cancel out the load torque. This means that exceeded
where one can appreciate the current reduction at 7 s, i.e., when flux demands when using open-loop steady-state values as
the load torque is reduced. a reference will demand unnecessary higher current values,
Finally, Table III shows the load torque increments used in implying an increment in the power lost in the copper and
simulation (Fig. 3) and real time (Fig. 11) and the correspond- core. Also, a real-time implementation was carried out where
ing variations in the rotor flux modulus and power lost. Here, the good performance of the induction motor could be verified.
one can appreciate that the rotor flux modulus (see Figs. 6 Some interesting issues as the robustness of the observer and
2886 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 7, JULY 2012

controller with respect to plant parameter variations and the For the dynamical stator current equations, one can replace (24)
sensorless control are currently under study. in the first and second equations in (20) and make use of Lls =
Ls − Lm ; then, one can calculate the rate of change of the stator
A PPENDIX I currents
N ONLINEAR A FFINE M ODEL OF I NDUCTION diα,s Rs + Rc Rc ψα,r
M OTORS W ITH C ORE L OSS =− iα,s −
dt (Lm − Ls ) Lm − Ls Lr − Lm
In [34], the following equations are derived from an equiva- Rc Lm iα,Lm Rc
+ + iα,Lm
lent electric diagram of an induction motor referred to the stator Lm − Ls Lr − Lm Lm − Ls
that takes into account the core: vα,s
+
Lm − Ls
diα,s diα,Lm
vα,s = Rs iα,s + Lls + Lm diβ,s Rs + Rc Rc ψβ,r
dt dt =− iβ,s −
diβ,s diβ,Lm dt (Lm − Ls ) Lm − Ls Lr − Lm
vβ,s = Rs iβ,s + Lls + Lm Rc Lm iβ,Lm Rc
dt dt + + iβ,Lm
diα,r diα,Lm Lm − Ls Lr − Lm Lm − Ls
0 = Rr iα,r + N pωψβ,r + Llr + Lm vβ,s
dt dt + . (25)
diβ,r diβ,Lm Lm − Ls
0 = Rr iβ,r − N pωψα,r + Llr + Lm
dt dt The mechanical equation remains the same as in the fifth-order
Lm diα,Lm model [3], [9] since the balancing of torques presented in the
0 = −iα,s − iα,r + + iα,Lm
Rc dt induction motor is identical in both cases.
Lm diβ,Lm
0 = −iβ,s − iβ,r + + iβ,Lm (20)
Rc dt
A PPENDIX II
where iα,r and iβ,r are the rotor currents. The rotor fluxes are P ROOFS OF T HEOREMS
determined as
A. Proof for Theorem 1
ψα,r = Llr iα,r + Lm iα,Lm The sliding-mode function dynamic appearing in the last
ψβ,r = Llr iβ,r + Lm iβ,Lm (21) equation in (9) closed-loop by the control action (10)
results in
making use of Llr = Lr − Lm and solving for the rotor
currents ζ̇3,1 = −kα,1 |ζ3,1 |sign(ζ3,1 ) + vα,1 + f¯2,1
ψα,r Lm iα,Lm v̇α,1 = −kα,2 sign(ζ3,1 )
iα,r = −
Lr − Lm Lr − Lm
ψβ,r Lm iβ,Lm ζ̇3,2 = −kβ,1 |ζ3,2 |sign(ζ3,2 ) + vβ,1 + f¯2,2
iβ,r = − . (22)
Lr − Lm Lr − Lm v̇β,1 = −kβ,2 sign(ζ3,2 ). (26)
For the third and fourth equations in (20), one can make use
of (21) in order to explicitly write the rate of change of the Now, following the same procedure outlined in [33], one can
rotor fluxes and then to replace (22), yielding to the following apply to (26) and analyze in the same way the following
expressions: candidate Lyapunov function:
  1 2
dψα,r ψα,r Lm iα,Lm V = 2kα,2 |ζ3,1 | + vα,1
= −Rr − − N pωψβ,r 2
Lr − Lm Lr − Lm
1
2
dt
 
dψβ,r ψβ,r Lm iβ,Lm + kα,1 |ζ3,1 |1/2 sign(ζ3,1 ) − vα,1
= −Rr − + N pωψα,r . (23) 2
dt Lr − Lm Lr − Lm 1 2
+ 2kβ,2 |ζ3,2 | + vβ,1
Replacing (22) in the fifth and sixth equations in (20) and 2
solving for the rate of change of the magnetization currents 1
2
+ kβ,1 |ζ3,2 |1/2 sign(ζ3,2 ) − vβ,1
result in 2
   = ξα Pα ξα + ξβT Pβ ξβ
T
diα,Lm Rc ψα,r Lm iα,Lm
= iα,s + −
dt Lm Lr − Lm Lr − Lm where ξαT = (|ζ3,1 |1/2 sign(ζ3,1 ) vα,1 ), ξβT =
Rc
− iα,Lm (|ζ3,2 | sign(ζ3,2 ) vβ,1 ), and
1/2
Lm  
  
diβ,Lm Rc ψβ,r Lm iβ,Lm 1 4kα,2 + kα,1 2
−kα,1
= iβ,s + − Pα =
dt Lm Lr − Lm Lr − Lm 2 −kα,1 2
 
Rc 1 4kβ,2 + kβ,1 −kβ,1
2
− iβ,Lm . (24) Pβ =
2 −kβ,1 2
.
Lm
RIVERA DOMÍNGUEZ et al.: COPPER AND CORE LOSS MINIMIZATION FOR INDUCTION MOTORS 2887

Its time derivative along the solution of (26) results as obtained by replacing the calculated equivalent controls,
follows: resulting in a linear time-variant dynamic system
 
1 f¯ Ao,11 Ao,12
V̇ = −   ξ T Qα ξα +  2,1  qα,1
T
ξα ˙ = Ao (ω), Ao = (28)
1/2  α  1/2  Ao,21 Ao,22
ζ3,1  ζ3,1 
1 f¯ with
−   ξβT Qβ ξβ +  2,2  qβ,1
T
ξβ  
1/2   1/2  ρα η1 η3 − η4 −N pω
ζ3,2  ζ3,2  Ao,11 =
N pω ρβ η1 η3 − η4
where  
η4 Lm − ρα γ 0
  Ao,12 =
0 η4 Lm − ρβ γ
kα,1 2
2kα,2 + kα,1 −kα,1
Qα =  
2 −kα,1 1 η1
+ λα η1 η3 0
  Ao,21 = Lm
η1
kβ,1 2
2kβ,2 + kβ,1 −kβ,1 0 Lm + λβ η1 η3
Qβ =  
2 −kβ,1 1 −η1 − η2 − λα γ 0
Ao,22 =
T
qα,1 = ( 2kα,2 + 12 kα,1
2
− 12 kα,1 ) 0 −η1 − η2 − λβ γ
T
qβ,1 = ( 2kβ,2 + 12 kβ,1
2
− 12 kβ,1 ) . γ = η5 + η1 η3 Lm .

Applying the bounds for the perturbations as given in (11), At this point, one can choose the design parameters in the
the expression for the derivative of the Lyapunov function is following two fashions. First, as in the work presented in [44],
reduced to where a polynomial with desired poles is proposed pd (s) =
(s − p1 )(s − p2 )(s − p3 )(s − p4 ) such that the coefficients of
kα,1 T k
V̇ = −   ξ Q̃α ξα −  β,1  ξβT Q̃β ξβ the characteristic equation that results from the matrix Ao
1/2  α  1/2 
2 ζ3,1  2 ζ3,2  are equalized with the ones related with pd (s), i.e., det(sI −
Ao ) = pd (s), resulting in design parameters as functions of ω,
where this will guarantee that limt→∞ (t) = 0 and the tracking of any

 desired rotor velocity. Second, as in the work presented in [4],
4kα,2
2
2kα,2 + kα,1 − kα,1 + kα,1 δ1 −kα,1 + 2δ1 where it is assumed that the rotor velocity is constant, one can
Q̃α =
−kα,1 + 2δ1 1 easily determine the design parameters. In this case, the motor

 operation is restricted to slow varying-in-time rotor velocities.
4kβ,2
2
2kβ,2 + kβ,1 − kβ,1 + kβ,1 δ2 −kβ,1 + 2δ2
Q̃β = .
−kβ,1 + 2δ2 1
A PPENDIX III
Then, by choosing the control gains as in (12), one can easily S EPARATION P RINCIPLE
verify that Q̃α > 0 and Q̃β > 0, implying that the derivative of In the following, it will be shown that the feedback
the Lyapunov function is negative definite.  
When the sliding mode occurs, i.e., ζ3 = 0, the sliding-mode 1
v̂α = − kα,1 |ζ̂3,1 |sign(ζ̂3,1 ) − v̂α,1
dynamic results as follows: η3
v̂˙ α,1 = −kα,2 sign(ζ̂3,1 )
ż2 = k2 z2 + g2 ζ1
 
1
ζ˙1 = K1 ζ1 (27) v̂β = − kβ,1 |ζ̂3,2 |sign(ζ̂3,2 ) − v̂β,1
η3
where it can be seen that ζ1 and z2 tend asymptotically to zero, v̂˙ β,1 = −kβ,2 sign(ζ̂3,2 )
solving in that way the posed control problem.
with ζ̂31 = ζ2 − ζ̂2,r (w, ε) and ε = (T̂ l, ψ̂α , ψ̂β , îα,Lm ,
B. Proof of Theorem 2 îβ,Lm ), asymptotically stabilizes the residual dynamic in (9),
i.e., the sliding-mode dynamic (27).
From the derivative of the following Lyapunov candidate Based on the work presented in [45], where a separation
function Vo = (1/2)(ĩ2α + ĩ2β ) along the trajectories of (14), principle for the stabilization of a certain class of nonlinear
one can easily see that the gains (16) will make V̇o < 0 and, systems is proposed, the main results for fulfilling this principle
as a consequence, the convergence of ĩα and ĩβ toward zero in are summarized as follows.
finite time. When the sliding mode occurs, i.e., ĩα = ĩβ = 0, R1) The system must be stabilized by a globally bounded
one can calculate the equivalent control for the injected state feedback control. The global boundedness of the
signals from ĩ˙ α = 0 and ĩ˙ β = 0 as να,eq = −η1 η3 ψ̃α + control protects the state of the plant from peaking when
(η5 + η1 η3 Lm )ĩα,Lm and νβ,eq = −η1 η3 ψ̃β + (η5 + the high-gain observer estimates are used instead of the
η1 η3 Lm )ĩβ,Lm ; then, the sliding-mode dynamic can be true states.
2888 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 7, JULY 2012

R2) The high-gain observer must be globally and exponen- [18] F. Fernandez-Bernal, A. Garcia-Cerrada, and R. Faure, “Model based
tially stable. loss minimization for dc and ac vector-controlled motors including
core saturation,” IEEE Trans. Ind. Appl., vol. 36, no. 3, pp. 755–763,
In the work presented in [46], it has been demonstrated that May/Jun. 2000.
the second-order sliding-mode controllers as (10) are globally [19] I. Kioskeridis and N. Margaris, “Loss minimization in induction motor
adjustable-speed drives,” IEEE Trans. Ind. Appl., vol. 43, no. 1, pp. 226–
bounded feedback controllers for nonlinear affine systems; 231, Feb. 1996.
therefore, R1 is fulfilled. Moreover, the sliding-mode dynamic [20] C. Chakraborty, M. C. Ta, T. Uchida, and Y. Hori, “Fast search controllers
(27) is globally asymptotically stable by this controller. In [30], for efficiency maximization of induction motor derives based on dc link
power measurement,” in Proc. Power Convers. Conf., 2002, pp. 402–408.
it is stated that the discontinuous control action implements [21] G. Sousa, B. Bose, and J. Cleland, “Fuzzy logic based on-line efficiency
high (theoretically infinite) gain; therefore, the sliding-mode optimization control of an indirect vector-controlled induction motor
observer (13) with (15) can be considered as a high-gain ob- drive,” IEEE Trans. Ind. Appl., vol. 42, no. 2, pp. 192–198, Apr. 1995.
[22] D. Kirschen, D. Novotny, and T. Lipo, “On-line efficiency optimization
server along with the load torque observer (17) whose gains can of a variable frequency induction motor drive,” IEEE Trans. Ind. Appl.,
be selected sufficiently large. Moreover, it can be easily seen vol. IA-21, no. 3, pp. 610–616, May 1985.
that the sliding-mode dynamic (28), along with the estimation [23] E. Levi, A. Boglietti, and M. Lazzar, “Performance deterioration in indi-
rect vector controlled induction motor drives due to iron losses,” in Proc.
error for the load torque (19), is globally exponentially stable. Power Electron. Spec. Conf., 1995, pp. 1312–1318.
[24] W. J. Jinh and K. Nam, “A vector control scheme for EV induction motors
with series iron loss model,” IEEE Trans. Ind. Appl., vol. 45, no. 4,
R EFERENCES pp. 617–624, Aug. 1998.
[1] F. Blaschke, “The principle of field orientation as applied to the new [25] H. Khalil, Nonlinear Systems. New York: Macmillan, 1992.
transvector closed-loop control system for rotating field machines,” [26] M. Vidyasagar, Nonlinear Systems Analysis. Philadelphia, PA: SIAM,
Siemens Rev., vol. 39, no. 5, pp. 217–220, 1972. 2002.
[2] G. W. Chang, G. Espinosa-Perez, E. Mendes, and R. Ortega, “Tuning rules [27] A. Isidori, Nonlinear Control Systems. London, U.K.: Springer-Verlag,
for the PI gains of field-oriented controllers of induction motors,” IEEE 1995.
Trans. Ind. Electron., vol. 47, no. 3, pp. 592–602, Jun. 2002. [28] R. Marino and P. Tomei, Nonlinear Control Design: Geometric, Adaptive,
[3] R. Marino, S. Peresada, and P. Valigi, “Adaptive input–output linearizing and Robust. London, U.K.: Prentice-Hall, 1995.
control of induction motors,” IEEE Trans. Autom. Control, vol. 38, no. 2, [29] H. Nijmeijer and A. Schaft, Nonlinear Dynamical Control Systems. New
pp. 208–221, Feb. 1993. York: Springer-Verlag, 1990.
[4] M. Hajian, J. Soltani, G. Markadeh, and S. Hosseinnia, “Adaptive non- [30] V. Utkin, J. Guldner, and Shi, Sliding Mode Control in Electromechanical
linear direct torque control of sensorless IM drives with efficiency op- Systems. Cleveland, OH: CRC, 1999.
timization,” IEEE Trans. Ind. Electron., vol. 57, no. 3, pp. 975–985, [31] A. Levant, “Quasi-continuous high-order sliding-mode controllers,” IEEE
Mar. 2010. Trans. Autom. Control, vol. 50, no. 11, pp. 1812–1816, Nov. 2005.
[5] C.-M. Lin and C.-F. Hsu, “Recurrent-neural-network-based adaptive [32] A. Levant, “Homogeneity approach to high-order sliding mode design,”
backstepping control for induction servomotors,” IEEE Trans. Ind. Elec- Automatica, vol. 43, no. 4, pp. 823–830, May 2005.
tron., vol. 52, no. 6, pp. 1677–1684, Dec. 2005. [33] J. A. Moreno and M. A. Osorio, “A Lyapunov approach to second-order
[6] A. B. Proca and A. Keyhani, “Sliding-mode flux observer with online sliding mode controllers and observers,” in Proc. 47th IEEE Conf. Deci-
rotor parameter estimation for induction motors,” IEEE Trans. Ind. Elec- sion Control, 2008, pp. 2856–2861.
tron., vol. 54, no. 2, pp. 716–723, Apr. 2007. [34] J. Dominguez, C. Mora-Soto, S. Ortega, J. Raygoza, and A. De La Mora,
[7] M. Fnaiech, F. Betin, G.-A. Capolino, and F. Fnaiech, “Fuzzy logic “Super-twisting control of induction motors with core loss,” in Proc. 11th
and sliding-mode controls applied to six-phase induction machine with Int. Workshop VSS, Jun. 2010, pp. 428–433.
open phases,” IEEE Trans. Ind. Electron., vol. 57, no. 1, pp. 354–364, [35] R. Stengel, Optimal Control and Estimation. New York: Dover, 1994.
Jan. 2010. [36] A. Levant, “Robust exact differentiation via sliding mode technique,”
[8] C. Lascu, I. Boldea, and F. Blaabjerg, “A class of speed-sensorless sliding- Automatica, vol. 34, no. 3, pp. 379–384, Mar. 1998.
mode observers for high-performance induction motor drives,” IEEE [37] B. Castillo-Toledo, S. Di Gennaro, A. Loukianov, and J. Rivera, “Hybrid
Trans. Ind. Electron., vol. 56, no. 9, pp. 3394–3403, Sep. 2009. control of induction motors via sampled closed representations,” IEEE
[9] V. I. Utkin, “Sliding mode control design principles and applications to Trans. Ind. Electron., vol. 55, no. 10, pp. 3758–3771, Oct. 2008.
electric drives,” IEEE Trans. Ind. Electron., vol. 40, no. 1, pp. 23–36, [38] J. Holtz, “Sensorless control of induction machines—With or without
Feb. 1993. signal injection?,” IEEE Trans. Ind. Electron., vol. 53, no. 1, pp. 7–30,
[10] S. Gadoue, D. Giaouris, and J. Finch, “Sensorless control of induction Feb. 2006.
motor drives at very low and zero speeds using neural network flux [39] R. Shahnazi, H. M. Shanechi, and N. Pariz, “Position control of in-
observers,” IEEE Trans. Ind. Electron., vol. 56, no. 8, pp. 3029–3039, duction and dc servomotors: A novel adaptive fuzzy PI sliding mode
Aug. 2009. control,” IEEE Trans. Energy Convers., vol. 23, no. 1, pp. 138–147,
[11] H. Toliyat, M. Wlas, and Z. Krzemiriski, “Neural-network-based parame- Mar. 2008.
ter estimations of induction motors,” IEEE Trans. Ind. Electron., vol. 55, [40] M. Ghanes and G. Zheng, “On sensorless induction motor drives: Sliding-
no. 4, pp. 1783–1794, Apr. 2008. mode observer and output feedback controller,” IEEE Trans. Ind. Elec-
[12] R. Moulin, J. Leveque, L. Durantay, B. Douine, D. Netter, and tron., vol. 56, no. 9, pp. 3404–3413, Sep. 2009.
A. Rezzoug, “Superconducting multistack inductor for synchronous mo- [41] A. Benchaib, A. Rachid, E. Audrezet, and M. Tadjine, “Real-time sliding-
tors using the diamagnetism property of bulk material,” IEEE Trans. Ind. mode observer and control of an induction motor,” IEEE Trans. Ind.
Electron., vol. 57, no. 1, pp. 146–153, Jan. 2010. Electron., vol. 46, no. 1, pp. 128–138, Feb. 1999.
[13] M. Centner and U. Schafer, “Optimized design of high-speed induction [42] M. Hrasko, J. Vittek, R. Havrila, and I. Lokseninec, “Rotor flux observer
motors in respect of the electrical steel grade,” IEEE Trans. Ind. Electron., in pseudo-sliding mode for vector controlled induction motor drives,” in
vol. 57, no. 1, pp. 288–295, Jan. 2010. Proc. 12th Int. EPE-PEMC, Sep. 2006, pp. 1133–1136.
[14] W. Li, J. Cao, and X. Zhang, “Electrothermal analysis of induction motor [43] K.-K. Shyu, J.-K. Lin, V.-T. Pham, M.-J. Yang, and T.-W. Wang, “Global
with compound cage rotor used for PHEV,” IEEE Trans. Ind. Electron., minimum torque ripple design for direct torque control of induction mo-
vol. 57, no. 2, pp. 660–668, Feb. 2010. tor drives,” IEEE Trans. Ind. Electron., vol. 57, no. 9, pp. 3148–3156,
[15] L. Liu, K. Zhang, and S. Zhang, “Optimal efficiency control of induction Sep. 2010.
motor with core loss,” in Proc. IEEE Int. Conf. Appl. Supercond. Electro- [44] Y. Liu, X. Wu, J. Jim Zhu, and J. Lew, “Omni-directional mobile ro-
magn. Devices, 2009, pp. 180–182. bot controller design by trajectory linearization,” in Proc. Amer. Control
[16] G. Dong and O. Ojo, “Efficiency optimizing control of induction mo- Conf., 2003, vol. 4, pp. 3423–3428.
tor using natural variables,” IEEE Trans. Ind. Electron., vol. 53, no. 6, [45] A. Atassi and H. Khalil, “A separation principle for the stabilization of a
pp. 1791–1798, Dec. 2006. class of nonlinear systems,” IEEE Trans. Autom. Control, vol. 44, no. 9,
[17] S. L. and K. Nam, “Loss-minimising control scheme for induction mo- pp. 1672–1687, Sep. 1999.
tors,” Proc. Int. Elect. Eng.—Elect. Power Appl., vol. 151, no. 4, pp. 385– [46] A. Levant and L. Alelishvili, “Integral high-order sliding modes,” IEEE
397, Jul. 2004. Trans. Autom. Control, vol. 52, no. 7, pp. 1278–1282, Jul. 2007.
RIVERA DOMÍNGUEZ et al.: COPPER AND CORE LOSS MINIMIZATION FOR INDUCTION MOTORS 2889

Jorge Rivera Domínguez was born in El Rosario, Juan José Raygoza Panduro received the B.S. de-
México, in 1975. He received the B.Sc. degree from gree in communications and electronics from the
the Technological Institute of the Sea, Mazatlán, Universidad de Guadalajara, Guadalajara, México,
México, in 1999 and the M.Sc. and Ph.D. de- in 1989, the M.Sc. degree from the Center for
grees in electrical engineering from the Center for Research and Advanced Studies (CINVESTAV),
Research and Advanced Studies (CINVESTAV), National Polytechnic Institute (IPN), Zacatenco
National Polytechnic Institute (IPN), Guadalajara, Campus, México City, México, and the Ph.D. de-
México, in 2001 and 2005, respectively. gree in computer science and telecommunications
Since 2006, he has been with the Universidad from the Autonomous University of Madrid, Madrid,
de Guadalajara, Guadalajara, México, as a full-time Spain.
Professor with the Departamento de Electrónica, From 1996 to 2000, he was with IBM, where he
Centro Universitario de Ciencias Exactas e Ingenierías. His research interests participated in the technological transfer of manufacturing hard-disk heads
focus on regulator theory, sliding-mode control, discrete-time nonlinear control from the IBM manufacturing plant in San Jose, CA, to Guadalajara. He
systems, and their applications to electrical machines. He has published more is currently with the Departamento de Electrónica, Centro Universitario de
than 40 technical papers in international journals and conferences and has Ciencias Exactas e Ingenierías, Universidad de Guadalajara. He specializes in
served as Reviewer for different international journals and conferences. the design of digital architectures based on field-programmable gate arrays,
microprocessors, embedded systems, and bioelectronics. The main lines of
investigation which he works are electronic systems applied to biomedicine,
Christian Mora-Soto received the M.Sc. degree microprocessor design, digital control, and embedded systems.
in electronic engineering from the Universidad de
Guadalajara, Guadalajara, México, in 2010.
He is currently with the Departamento de Elec- Alexander G. Loukianov was born in Moscow,
trónica, Centro Universitario de Ciencias Exactas e Russia, in 1946. He received the Dipl.Eng. degree
Ingenierías, Universidad de Guadalajara. His current from the Polytechnic Institute, Moscow, in 1975
research work is focused on the control of electrical and the Ph.D. degree in automatic control from the
machines and drives. Institute of Control Sciences, Russian Academy of
Sciences, Moscow, in 1985.
In 1978, he was with the Institute of Control
Sciences, Russian Academy of Sciences, where he
was the Head of the Discontinuous Control Systems
Laboratory from 1994 to 1995. In 1992–1995, he
Susana Ortega-Cisneros received the B.S. degree was in charge of an industrial project between the
in communications and electronics from the Univer- Institute of Control Sciences, Russian Academy of Sciences, and the largest
sidad de Guadalajara, Guadalajara, México, in 1990, Russian car plant and also several international projects supported by Inter-
the M.Sc. degree from the Center for Research and national Association for the promotion of co-operation with scientists from
Advanced Studies (CINVESTAV), National Poly- the New Independent States of the former Soviet Union (INTAS) and Inter-
technic Institute (IPN), Zacatenco Campus, México national Cooperation-Copernicus (INCO-COPERNICUS), Brussels, Belgium.
City, México, and the Ph.D. degree in computer sci- In 1995–1997, he held a visiting position at the University of East London,
ence and telecommunications from the Autonomous London, U.K. Since April 1997, he has been with the Center for Research
University of Madrid, Madrid, Spain. and Advanced Studies (CINVESTAV), National Polytechnic Institute (IPN),
She is currently with CINVESTAV, IPN, Zapopan, Guadalajara Campus, Zapopan, México, as a Professor of electrical engineering
México. She specializes in the design of digital graduate programs. His research interests center in nonlinear systems control
architectures based on field-programmable gate arrays (FPGAs), DSPs, and and variable structure systems with sliding mode as applied to electric drives
microprocessors. The main lines of investigation which she works are digital and power systems control, robotics, space, and automotive control. He has
control, self-timed synchronization, electronic systems applied to biomedicine, published more than 90 technical papers in international journals and con-
embedded microprocessor design, digital electronics, and custom DSPs in ferences and has served as Reviewer for different international journals and
FPGAs. conferences.

You might also like