You are on page 1of 10

1444 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 61, NO.

3, MARCH 2014

Sensorless Control of Induction-Motor Drive


Based on Robust Kalman Filter and
Adaptive Speed Estimation
Francesco Alonge, Member, IEEE, Filippo D’Ippolito, Member, IEEE, and
Antonino Sferlazza, Student Member, IEEE

Abstract—This paper deals with robust estimation of rotor flux must be obtained. However, performance is greatly affected
and speed for sensorless control of motion control systems with by parameter uncertainty; thus, the behavior of both the con-
an induction motor. Instead of using sixth-order extended Kalman troller and the estimator, which are designed using a model-
filters (EKFs), rotor flux is estimated by means of a fourth-order
descriptor-type robust KF, which explicitly takes into account mo- based approach, rapidly deteriorates in the presence of these
tor parameter uncertainties, whereas the speed is estimated using uncertainties. Obviously, the behavior of the whole control
a recursive least squares algorithm starting from the knowledge of system is particularly sensitive to that of the state estimator.
the rotor flux itself. It is shown that the descriptor-type structure To cope with these uncertainties, online parameter identifica-
allows for a direct translation of parameter uncertainties into vari- tion, i.e., adaptive or robust design techniques, can be em-
ations of the coefficients appearing in the model, and this improves
the degree of robustness of the estimates. Experimental findings, ployed for designing either the estimator or the controller,
carried out on a closed-loop system consisting of a low-power or both.
induction-motor-load system, a proportional–integral-type con- To this regard, in [1] and [2], both rotor and stator resis-
troller, and the proposed estimator, are shown with the aim of tances are estimated online using neural networks (NNs) or two
verifying the goodness of the whole closed-loop control system. extended Kalman filters (EKFs), respectively, whereas in [3],
Index Terms—Adaptive speed estimation, induction motor, ro- only stator resistance is estimated using an EKF; all of these
bust Kalman filter, sensorless control. works assume that stator, rotor, and mutual inductances are
well known. In [1], rotor and stator resistances are estimated
N OMENCLATURE by two different schemes involving two different NNs that are
usd , usq Stator voltages in a fixed reference frame [V]. constructed starting from a model of the induction motor, and
speed is estimated using the same approach described in [4],
isd , isq Stator currents in fixed reference frame [A].
in which the authors state that estimated speed is sensitive to
ψsd , ψsq Rotor fluxes in fixed reference frame [Wb].
noise, thereby requiring filtering; experimental results obtained
Ls (Lr ) Stator (rotor) inductance [H].
by processing the data acquired from a closed-loop drive show
Lm Mutual inductance [H].
that, in the presence of load torque, the speed does not track
Rs (Rr ) Stator (rotor) resistance [Ω].
the measured one. In [2], rotor and stator resistances are esti-
Tr = Lr /Rr Rotor time constant [s].
mated by two different seventh-order EKFs that estimate the
σ Total leakage factor.
augmented state of the induction-motor-load system, consist-
σs Stator leakage factor.
ing of stator current and rotor flux components, speed, load
σr Rotor leakage factor.
torque and, alternatively, stator resistance and rotor resistance.
ω Electrical angular rotor speed [el.rad/s].
However, the braided EKF, consisting of two seventh-order
ωr Mechanical angular rotor speed [rad/s].
KFs, is too complex; moreover, in practical applications it is
Ts Sampling time.
not possible to know the existence of the persistent excitation
In Identity matrix (nth order).
condition a priori, which is a mandatory condition for an
exact parameter estimation. This appears clearly in the results
I. I NTRODUCTION shown in [2], aimed at proving the need of identifying both
rotor and stator resistances. In [5], it is described the real-time
N OWADAYS, field-oriented control of induction-motor
electrical drives is widely used when high performances implementation of a biinput EKF estimator, which deals with
the estimation of the whole state of the induction motor together
Manuscript received July 17, 2012; revised December 20, 2012 and with stator and rotor resistances, in the wide speed range.
March 20, 2013; accepted March 24, 2013. Date of publication April 5, 2013; In [3], a conventional EKF is studied, and the corresponding
date of current version August 23, 2013. results are given in the presence of various scenarios, but
The authors are with the Department of Energy, Information Engineering,
and Mathematical Models, University of Palermo, 90128 Palermo, Italy sensitivity analysis in the presence of parameter variations is
(e-mail: francesco.alonge@unipa.it; filippo.dippolito@unipa.it; antonino. not carried out, whereas in [6], it is shown that EKF is sensitive
sferlazza@unipa.it). to parameter variations, particularly at low speeds. Moreover,
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. the experimental results shown in [2] and [3] are obtained
Digital Object Identifier 10.1109/TIE.2013.2257142 by processing data acquired from voltage/frequency-controlled

0278-0046 © 2013 IEEE


Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
ALONGE et al.: SENSORLESS CONTROL OF INDUCTION-MOTOR DRIVE 1445

drives. Note that the EKF is applied also for sensorless control In Section II, a mathematical model of the motor and a de-
of synchronous ac drives [7]. scription of the uncertainties are shown. The formulation of the
An alternative solution to EKFs is presented in [8], where problem is considered in Section III. Then, in Sections IV and
the design and implementation of unscented KFs (UKFs) for V, the RDKF and the speed estimator are shown. In Section VI,
induction-motor sensorless drives is investigated. This filter a procedure is described to determine those system parameters
requires higher computational effort. useful for tuning the estimator. In Section VII, closed-loop
In [4], rotor flux and stator currents are estimated by means of experimental results are shown in order to validate the approach
a sliding mode observer (SMO), processed by the difference of previously described. Finally, Section VIII deals with some
the measured and observed stator currents. For those systems conclusions.
affine with respect to the input, the SMO is robust against all
disturbances, including parameter uncertainties that belong to
the space generated by the column of the forcing matrix, but II. M ATHEMATICAL M ODEL OF THE M OTOR AND
the produced estimates are affected by chattering, as shown U NCERTAINTY D ESCRIPTION
in [4].
The mathematical model of an induction motor in the de-
In [9]–[11], model reference adaptive control techniques
scriptor form is given by
are proposed for designing observers, whereas in [12], the
same techniques are employed for designing a controller. Other isd Lr dψrd
model reference adaptive system (MRAS) observers are de- σLs + = −Rs isd + usd (1)
dt Lm dt
scribed in the literature, and some of them are compared in [13].
From the control engineering point of view, the approach isq Lr dψrq
proposed in this paper explicitly assumes the objective of σLs + = −Rs isq + usq (2)
dt Lm dt
the robustness of the estimator against variations of all the
parameters of the motor, without the need of estimating some of dψrd Lm 1
= isd − ψrd − ωψrq (3)
them. The first step to reach this objective is that of formulating dt Tr Tr
the state estimation problem in two steps: In the first step, for a
dψrq Lm 1
given speed, the rotor flux and stator currents are estimated by = isq − ψrq + ωψrd (4)
means of a linear fourth-order robust descriptor KF (RDKF), dt Tr Tr
starting from measured stator currents and the supplied voltages dωr
computed by the controller; in the second step, the speed is J = −fv ωr + kt (ψrd isq − ψrq isd ) − tl (5)
dt
estimated by solving a total least squares problem starting
from the dynamic equations of the rotor flux components (cf. z = [isd isq ]T (6)
also [10] and [11]). The descriptor form of the KF is used
here because the coefficients of the model are functions of the where kt = 2pLm /(3Lr ), J is the inertia moment, fv is the
physical electromagnetic parameters that are simpler than those viscous friction coefficient, and z(t) is the output vector.
appearing in the conventional form. Consequently, physical The model (1)–(6) is nonlinear and multivariable, and is
parameter variations can be directly translated into variations affected by parametric uncertainties. Moreover, the load torque
of the coefficients appearing in the model, and this intrinsically tl is unknown. For estimating speed, two approaches could
leads to a certain degree of robustness of the DKF. Moreover, be employed. The first approach is based on the assumption
in order to take explicitly into account parameter uncertainties, that speed varies slowly with respect to the electromagnetic
the RDKF is designed according to [14] and [15]. variables; this suggests that ω̇ = 0 should substitute (5), thus
The advantages of the described procedure is that the me- obtaining the fifth-order model. The second approach leads to
chanical equation is not included in the state estimation pro- a sixth-order model, in which the load torque is assumed as
cedure, thus avoiding the use of nonlinear estimation methods, slowly varying, and consequently, the five-order state of the
such as EKF, and the connected lack of an observability prop- system (1)–(6) is augmented by the variable tl whose dynamics
erty of the model in certain operating conditions [16]; instead, is expressed by ṫl = 0. The resulting model is nonlinear and
only two linear least squares problems must be solved for requires an EKF.
estimating the state of the system. Moreover, neither load torque As stated in the introduction, in this paper, we use an alter-
estimation nor parameter estimation is required, guaranteeing, native procedure for estimating speed based on the assumption
in any case, the robustness of the estimation. Finally, all of the that speed is a parameter in (3) and (4) of the model. This is
parameter variations are simultaneously taken into account, and the reason why we consider only the model consisting of the
this occurs independently of the causes of their variation. The linear equations (1)–(4) and (6) for designing the RDKF. The
controller designed for the experiments consists of four simple mathematical model (1)–(4) and (6) in the compact matricial
proportional–integral (PI) control loops; only the speed con- form, including stochastic uncertainties, can be written in the
troller is equipped with an antiwind-up scheme because speed following descriptor form:
is subject to a large range of variations, and its variations are
slower than other variables. Experimental results are carried out
Ẽ ẋ(t) = F̃ (t)x(t) + B̃u(t) + Q̃w(t) (7)
on a closed-loop control system that uses the state estimation as
feedback variables for computing the PI-type control law. z(t) = Hx(t) + Rv(t) (8)
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
1446 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 61, NO. 3, MARCH 2014

where x(t) and u(t) are the state and input vectors, respec- whereas (10) remains unchanged because H and R are con-
tively, given by x = [isd isq ψrd ψrq ]T , u=[usd usq ]T ; stant. Matrices δE k and δF k are given by
w(t) and v(t) are the system and measurement noise assumed ⎡ ⎤
zero-mean white noise uncorrelated between them and with δ(σLs ) 0 δ LLmr 0
⎢ ⎥
the other variables, having covariance matrices equal to I 4 ⎢ δ LLmr ⎥
δE k = ⎢ 0 δ(σLs ) 0 ⎥
and I 2 , respectively; Q̃ and R are diagonal square matrices ⎣ ⎦
of suitable dimensions; and matrices Ẽ, B̃, F̃ (t), and H are 0 0 0 0
given by 0 0 0 0
δF k = δE k + Ts δ F̃ k
⎡ Lr ⎤ ⎡ ⎤
σLs 0 Lm 0 1 0 where
⎢ 0 Lr⎥ ⎢0 1⎥ ⎡ δ(R ) ⎤
Ẽ = ⎢
⎣ 0
σLs 0 Lm⎥, B̃ = ⎣ ⎦ 0 0 0
0 1 0 ⎦ 0 0 s
⎢ 0 δ(Rs ) 0 0 ⎥
0 0 0 1 0 0 ⎢ ⎥
δ F̃ k = ⎢ δ Lm 0 δ 1
0 ⎥.
⎡ ⎤ ⎣ Tr Tr ⎦
−Rs 0 0 0
 0 δ Lm
0 δ 1
⎢ 0 −R 0 0 ⎥ Tr Tr
F̃ (t) =⎢
s ⎥, H= 1 0 0 0
.
⎣ LTm 0 − T1r −ω(t)⎦ 0 1 0 0
r Generally, uncertainties for descriptor-type models are repre-
0 Lm
Tr ω(t) − T1r sented in the following standard form:
  
−δF k δE k+1 M f,k 0 −N f,k N e,k+1
Remark 1: The coefficients appearing in the model (1)–(4) = Δk
0 δH k 0 M h,k 0 N h,k
and, consequently, in (7) and (8), have simple expressions in (12)
terms of the physical parameters of the motor. For example, in
where Δk is a bounded arbitrary contraction with Δk ∞ ≤
looking at model (1)–(4), a variation in σLs produces variations
1, and M f,k , M h,k , N f,k , N e,k+1 , and N h,k are known
in the first two terms of (1) and (2), whereas in considering the
matrices. In our case, since δH k = 0 ∀k ∈ Z, we have
usual model of the motor, the same variation in parameter σLs
produces variations in all of the coefficients of the differential M f,k = I 4 , M h,k = I 2 , N f,k = max δF k 
k
equations expressing the dynamics of the stator currents and in
N e,k+1 = max δE k , N h,k = 0.
two terms of the equations expressing the dynamics of the rotor k
flux components. Although this introduces robustness into the
Now, with the aim of avoiding the formulation of problems
descriptor form of the KF with respect to the conventional one,
that are too complex, it is convenient to analyze the parameter
in this paper, a RDKF is designed for estimating stator currents
variations due to the increase in temperature and magnetic sat-
and rotor flux.
uration. Temperature variation produces variation in rotor and
Remark 2: Matrices B̃ and H are not affected by
stator resistances. However, rotor resistance varies also with the
uncertainties.
slip and, consequently, with load. All of these variations must
Remark 3: Matrix F̃ (t) is time-varying because it depends
be taken into account. In order to take into account magnetic
on the speed ω(t). Matrix Ẽ is always nonsingular.
saturation effects with accuracy, very complex mathematical
Starting from model (7) and (8), the following discrete-time
models must be constructed [17], but these models generally
stochastic model is obtained by using the Euler method:
lead to complex controllers. In many cases, saturation is taken
into account, assuming that it leads to a reduction of the mutual
Exk+1 = F k xk + Buk + Qwk (9) inductance Lm . Consequently, saturation affects the values of
the rotor and stator inductances given by
z k = Hxk + Rv k (10)
Lr = Lσr + Lm (13)
where k := kTs (k ∈ Z) is the current discrete time in which Ls = Lσs + Lm (14)
Ts is the sampling time, and E = Ẽ, F k = Ẽ + Ts F̃ k , F̃ k =
where Lσr = σr Lm and Lσs = σs Lm are the rotor and stator
F̃ (kTs ), B = Ts B̃, and Q = Ts Q̃.
leakage inductances, respectively, which are assumed to be
constant because we are interested in the saturation of the flux
main path. Because the values of the given leakage induc-
A. Uncertainty Description tances are small with respect to Lm , we have δ(σLs )  σLs ,
Assuming that the values of the electromagnetic parameters δ(Lm /Lr )  Lm /Lr and δ(Lm /Tr ) ∼ = (Lm /Lr )δ(Rr ).
are different from the nominal ones, denoting with δE k and It follows that:
δF k the corresponding variations of matrices E and F k , (9)
E + δEk ∼
= E, Fk + δFk ∼
= E + Ts (F˜k + δ F̃k ).
becomes
In order to set magnetic parameter variations, after the
(E + δE k )xk+1 = (F k + δF k )xk + Buk + Qwk (11) computation of Lσr and Lσs from the nominal values of the
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
ALONGE et al.: SENSORLESS CONTROL OF INDUCTION-MOTOR DRIVE 1447

parameters, for a given percentage variation of Lm , the values IV. K ALMAN FILTERING
of Lr and Ls are obtained using (13) and (14).
A. DKF
Let us suppose that ωk is known. In addition, in (9) and (10),
III. P ROBLEM F ORMULATION
matrix [E T H T ]T is full column rank, the recursive filtered
The problem we deal with in this paper is that of estimating estimate x̂k|k of state xk , i.e., the solution of the problem (15)
the rotor flux components and the speed of an induction-motor- and (16), is given by the following algorithm [14].
load system, by solving two linear least squares subproblems. At instant k = 0, the algorithm is initialized with
More precisely, assuming at instant k the knowledge of the
speed, denoted by ω̂k , the first subproblem is that of estimating P 0|0 = [P −1 T −1
0 +H R H]
−1
x̂0|0 = P 0|0 + H T R−1 z 0 .
the state of the model (9) and (10) by means of a DKF, by
Then, at step k, update {x̂k|k , P k|k } to {x̂k+1|k+1 , P k+1|k+1 }
solving the following minimization problem:
as follows:
min x0 2P −1 +z 0 −Hx0 2R−1 for k = 0 (15)  −1
x0 P k+1|k+1 = E T (Q + F k P k|k F Tk )−1 E + H T R−1 H
 0
min xk − x̂k|k 2P −1 + Exk+1 (21)
xk ,xk+1 k|k 
 −1
− (F k xk + Buk )2Q−1 + z k+1 − Hxk 2R−1 , x̂k+1|k+1 =P k+1|k+1 E T Q + F k P k|k F Tk

for k > 0. (16)


× (F k x̂k|k + Buk )+H T R−1 z k+1 .
With reference to the second subproblem, using the Eulero
method, the last two equations of (9), expressing the dynamics (22)
of the rotor flux components, can be rewritten in the following
Algorithms (21) and (22) can be obtained as a solution of the
matrix form:
  following regularized least squares problem:
−Ts ψrq (k) ψrd (k + 1) − w1 ψrd (k) − w2 isd (k)  
Ts ψrd (k)
ωk =
ψrq (k + 1) − w1 ψrq (k) − w2 isq (k) min = xT Qx + (Ax − b)T W (Ax − b) (23)
x
(17)
where xT Qx is a regularization term, and Q = QT > 0 and
where w1 = 1 − (1/Tr )Ts and w2 = (Lm /Tr )Ts , and the val-
W = W T ≥ 0 are weight matrices; x ∈ Rn is the unknown
ues of the rotor flux components are given from the solution of
vector; A ∈ Rn×n is the data matrix; and b ∈ Rn×1 is the
the first subproblem (see Section V). From (17), the formulation
observation vector. The solution of (23) is
of the second subproblem is as follows:
x̂ = [Q + AT W A]−1 AT W b. (24)
min Φ̂k ω̂k − ŷ k 2 for k ≥ 0 (18)
ω̂k
The problem (16) can be rewritten in the regularized least
where squares form (23) with the following identifications:
  
−Ts ψ̂rq (k) −F k E F k x̂k|k + Buk
Φ̂k = A← , b←
Ts ψ̂rd (k) 0 H z k+1
  
ψ̂rd (k + 1) − w1 ψ̂rd (k) − w2 isd (k) −Q−1 0 −P −1 0
ŷ k = . W ← −1 , Q← k|k
ψ̂rq (k + 1) − w1 ψ̂rq (k) − w2 isq (k) 0 R 0 0

Since our objective is to take explicitly uncertainties [cf. −xk − x̂k|k
x← . (25)
model (11)], we need to modify the formulation (15) and (16) xk+1
into the following robust version:
Consequently, (24) with the identification (25) leads to the
min x0 2P −1 + z 0 − Hx0 2R−1 for k = 0 (19) time and measurement update of the KF (21) and (22).
x0 0

min max xk − x̂k|k 2P −1 B. RDKF
xk ,xk+1 δF k ,δE k k|k

+ (E + δE k )xk+1 Consider the following robust version of the optimization


− ((F k + δF k )xk + Buk )2Q−1 problem (23):
+ z k+1 − Hxk 2R−1 , for k > 0. min max = x2Q + (A + δA)x − (b + δb)2W (26)
x δA,δb
(20)
where {δA, δb} are uncertainties modeled by
Equation (18) can be solved using least squares methods, as
will be shown in Section V. [δA δb] = M Δ[δN a δN b ] (27)
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
1448 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 61, NO. 3, MARCH 2014

where Δ∞ ≤ 1. The solution of (26) is given by Then, at step k, update {x̂k|k , P k|k } to {x̂k+1|k+1 ,
P k+1|k+1 } as follows:
x̂ = [Q̂ + AT Ŵ A]−1 AT Ŵ b + λ̂N Ta N Tb (28)
P k+1|k+1
 −1
where {Q̂, Ŵ } are defined as follows: T T −1 T −1
= Ê k+1 Q̂k + F̂ k P k|k F̂ k Ê k+1 + Ĥ R̂k+1 Ĥ
Q̂ = Q − λ̂−1 N Ta N a (29)
(33)
Ŵ = W + W M (λ̂I − M T W M )M T W (30) x̂k+1|k+1

T T −1
λ̂ is a nonnegative scalar parameter obtained by solving the = P k+1|k+1 Ê k+1 Q̂k + F̂ k P k|k F̂ k
following optimization problem [14]:
T −1
× F̂ k x̂k|k + B̂uk + Ĥ R̂k+1 z k+1
λ̂ = arg min G(λ) (31)
λ≥M T W M 
(34)
where
where
G(λ) = x(λ)Q2 −λN a x(λ) − N b 2 +Ax(λ)−b2W (λ) 
Q̂k 0
 −1  T  Q̂k = Q − λ̂−1
k I 4, Q̂k =
x(λ) := Q(λ) + AT W (λ)A A W (λ)b + λN Ta N b 0 I4
  
Q(λ) := Q − λ−1 N Ta N a  E  Fk
Ê k+1 = , F̂ k =
λ̂−1
k Ne λ̂−1
k N f,k
W (λ) := W + W M (λI − M T W M )M T W .
 
H B
Now, the minimax problem (20) can be rewritten as problem Ĥ = , B̂ =
02×4 04×2
(26) by means of the following identifications: 
  −1 R̂k+1 0
−F k E F k x̂k|k + Buk R̂k+1 = R − λ̂k I 2 R̂k+1 = (35)
A← , b← 0 I2
0 H z k+1
  and λ̂k is obtained by minimizing the function G(λ) [cf.
−Q−1 0 −P −1 0
W ← −1 , Q ← k|k
(31)] with the identification (32) over the interval λ̂k > λl =
0 R 0 0
diag{Q−1 , R−1 }.
 
−δF k δE δF k x̂k|k Computation of λ̂k can be carried out by means of the given
δA ← , δb ← optimization procedure. However, in [15], it is proposed to
0 0 0
choose λ̂k as follows:
  
−N f,k N e N f,k x̂k|k
Na ← , Nb ← λ̂k = λ̂ =
(1 + 0, 5)λl , for λl = 0
(36)
0 0 0
0, for λl = 0
 
−M f 0 −xk − x̂k|k which allows the offline computation of several of the given
M← , x← (32)
0 Nf xk+1 matrices.
Remark 4: From (33) and (34), it is easy to verify
and the initial conditions are
that, for descriptor systems without uncertainties (M f,k =
A ← H, b ← z 0 , δA ← 0, δb ← 0, Q ← P −1
0 ,
0, M h,k = 0, N f,k = 0, N e,k+1 = 0, N h,k = 0), this algo-
rithm collapses to the DKF (21) and (22).
M ← M h , N a ← 0, N b ← 0. Remark 5: The main difference between the robust filter
and the standard one is that, in the robust algorithm, the new
From (28) and the identifications (32), the filtered robust
recursion operates on system and noise covariance matrices,
optimum estimate x̂k|k is obtained from the following recursive
modified with respect to the given nominal values. More
algorithm.
precisely, the algorithm updates these matrices to the values
At instant k = 0, the algorithm is initialized with
necessary for obtaining robust estimation.
−1
P 0|0 = [P −1 T
0 + H R̂ H]
−1

−1 V. S PEED E STIMATION
x̂0|0 = P 0|0 + H T R̂ z 0
The problem (18) could be solved by means of the ordinary
R̂ = R − λ̂−1
−1 I 2 least squares (OLS) method as follows:
−1 −1 T
where λ̂−1 is obtained by minimizing the function G(λ) [cf. Pω,k+1 = τ Pω,k + Φ̂k Φ̂k (37)
(31)] with the identification (32) over the interval λ̂−1 > λl = T
R−1 . ω̂k+1 = ω̂k + Pω,k+1 Φ̂k Φ̂k ω̂k − ŷ k (38)
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
ALONGE et al.: SENSORLESS CONTROL OF INDUCTION-MOTOR DRIVE 1449

cal model of the electromagnetic circuit of the motor, expressed


in complex state variables:
 
dis Ls − σLs 1
σLs = − Rs + is − jω ψ̃ r + ψ̃ r + us
dt Tr Tr
(41)
 
dψ̃ r Ls − σLs 1
Fig. 1. Block diagram of the whole estimator. = is + jω ψ̃ r + ψ̃ r (42)
dt Tr Tr
where τ < 1 is the forgetting factor, and the algorithm is
where is = isd + jisq , ψ̃ r = ψ̃rd + j ψ̃rq , and us = usd +
initialized with Pω,0 = P0 for some P0 > 0 and ω̂0 = 0.
jusq .
Remark 6: Forgetting factor τ is introduced in (37), in order
−1 In order to identify the parameters of model (41) and (42),
to avoid that, for τ = 1, Pω,k increases linearly, which implies
the Levenberg–Marquardt algorithm (LMA) has been used
that Pω,k converges to zero for k → ∞; this in turn implies that,
[19], which provides a numerical solution to the problem of
during speed transients, the estimated speed tracks the actual
minimizing a nonlinear function over a space of parameters of
one very slowly.
the function.
As it is well known, the OLS algorithm assumes that Φ̂k is
For the application of the LMA, the aforementioned nonlin-
not affected by errors, and errors are confined to ŷ k . However,
ear function consists of
this hypothesis does not correspond to our case because estima-

tion and modeling errors cause errors also in Φ̂k . Therefore, in 
1  N
this application, the employment of total least squares (TLS) is S(β) =  isd,k − îsd,k 2 + isq,k − îsq,k 2 (43)
better because it also takes into account the errors in the data N
k=1
matrix. In fact, estimated rotor flux present in Φ̂k is affected
both by modeling errors and noise, similar to the observation where (isd,k , isq,k ) are the direct and in-quadrature stator cur-
vector. Consequently, instead of (18), a TLS problem is solved rents measured at instant k, and (îsd,k , îsq,k ) are the corre-
by minimizing the following modified cost function: sponding values computed by solving model (41) and (42), for
a given set of stator voltages and the actual parameter vector β
Φ̂k ω̂k − ŷ k 2 defined as follows:
min . (39)
ω̂k 1 + ω̂k2
β = [Rs Ls σLs T r ]T . (44)
The adaptation law that minimizes (39) is
T The algorithm is stopped when the parameter vector β is such
ω̂k+1 = ω̂k − αk ΓTk Φ̂k + αk Φ̂k Φ̂k ω̂k (40) that function (43) is less than the chosen step-size tolerance.
In order to obtain all of the electromagnetic parameters, an
where αk is a positive constant, and Γk is given by often used procedure consists on the assumption Lr = Ls [20];
Δk then, it is possible to compute rotor resistance from the rotor
Γk = , Δk = Φ̂k ω̂k − ŷ k . time constant Tr . Alternatively, we can assume σr = 2σs [21],
1 + ω̂k2
to compute σ from σLs and Ls , and then compute σr and σs as
In [18], it is proven that the origin ω̂ = 0 always belongs to follows:
the convergence domain of TLS. Hence, the use of a null initial 
condition is the best choice if no prior information is given. The σ 9 3
σs = + − (45)
block diagram of the complete estimator is shown in Fig. 1. 2(1 − σ) 16 4
σr = 2σs (46)
VI. PARAMETER E STIMATION P ROCEDURE FOR
E STIMATOR T UNING and finally, we compute Lm and Lr by the equations

For the tuning of the estimator, a set of parameters is iden- Ls


Lm = (47)
tified offline using the procedure described in the following. 1 + σs
The identified parameters are assumed as nominal parameters Lr = Lm (1 + σr ). (48)
for designing the DKF, the speed estimator, and the controller.
These parameters together with their hypothesized variation In this paper, the last method is employed.
ranges are taken into account for designing the RDKF. Con-
sequently, the parameters are not updated online.
VII. E XPERIMENTAL R ESULTS
Since the direct and in-quadrature components of the induced
part flux cannot be measured, only some parameters appearing In order to validate the estimator described earlier, closed-
in standard model (1)–(4) can be estimated. In order to obtain a loop experiments are shown, which were carried out on a
mathematical model in which only identifiable parameters ap- system consisting of a 750-W induction motor with two pole
pear, the expression of the scaled rotor flux ψ̃r = (Lm /Lr )ψr pairs and a powder brake system shown in Fig. 2, a three-phase
is replaced. This allows for obtaining the following mathemati- ac/dc converter–filter–source voltage inverter unit supplying the
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
1450 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 61, NO. 3, MARCH 2014

Fig. 3. Speed response of the system with feedback from the RDKF at high
reference speed and no load. The machine is fluxed at zero reference speed up
to 0.5 s, and then, it is started with a trapezoidal reference speed of 150 rad/s.

Fig. 2. Motor brake system.

TABLE I
M OTOR C HARACTERISTIC

TABLE II
M OTOR PARAMETERS Fig. 4. Current estimation error of the system with feedback from the RDKF
at high reference speed. Same conditions of Fig. 3.

motor, and a cascade controller consisting of four PI control


loops, two inner current loops, and two outer rotor flux and
speed loops [22], [23]. These PI controllers are designed as
described in [23], to obtain a bandwidth of 10 Hz for speed
and rotor flux loops, and 40 Hz for current loops. An antiwind-
up scheme is designed for a speed control loop [24]. The
module of stator current and voltage vectors are constrained
to Is,max = 10 A in order to avoid damage of the machine,
Vs,max = 0, 866VBU S V, i.e., the maximum modulus of the
rotating voltage vector that the inverter can generate according Fig. 5. Speed response of the system with feedback from the DKF at high
to the pulsewidth modulation technique employed. The motor, reference speed. Same conditions of Fig. 3.
fluxed at 0.5 Wb at t = 0, starts at t = 0.5 s. The whole
controller, including the proposed estimator, is implemented on In order to analyze robustness, both DKF and RDKF esti-
a DS1103 microcontroller that processes the controller itself at mators are designed assuming the following uncertainties: 50%
12 kHz, and allows data acquisition of the measured variables for Rr and Rs , and 30% for Lm . Note that neither cause nor
and their visualization on the cockpit provided by dSPACE rate of variation are needed for designing the estimator. Both
software. estimators were initialized assuming P 0 = 50I 4 , x0 = 0, R =
The measured variables are the speed computed starting from I 2 , and Q = diag{2 × 10−2 , 2 × 10−2 , 2 × 10−3 , 2 × 10−3 }.
data acquired by means of a 1024 ppr incremental encoder Figs. 3–6 show the responses of the closed-loop system in the
useful for comparing estimated and measured speeds, the latter presence of either robust or standard estimators, corresponding
filtered using a phase-locked loop (PLL) scheme and the two to a trapezoidal reference speed with a maximum speed of
stator currents given by two Hall effect transducers. 150 rad/s, at no load and with speed reversal.
The rated data of the motor are shown in Table I. The Figs. 9–12 show the responses of the closed-loop system at
parameters of the motor, identified as described in Section VI, no load, during low speed tests (3 rad/s). Both estimators are
are given in Table II. able to track the reference speed with a mean speed equal to
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
ALONGE et al.: SENSORLESS CONTROL OF INDUCTION-MOTOR DRIVE 1451

Fig. 6. Current estimation error of the system with feedback from the DKF at Fig. 9. Speed response of the system with feedback from the RDKF at no
high reference speed. Same conditions of Fig. 3. load and low reference speed. The machine is fluxed at zero reference speed. At
1 s, it is started with a step of 3 rad/s; and a further step of −3 rad/s, applied at
15 s, leads the reference speed to zero.

Fig. 7. Speed response of the system with feedback from the RDKF at high
reference speed. A 5-N · m load torque is applied at 2 s and removed at 13 s.
Fig. 10. Current estimation error of the system with feedback from the RDKF
at low-speed reference. Same operating conditions of Fig. 9.

Fig. 8. Speed response of the system with feedback from the DKF at high
reference speed. Same conditions of Fig. 7.
Fig. 11. Speed response of the system with feedback from the DKF at low
zero, but the RDKF displays better dynamic properties and is reference speed. Same operating conditions of Fig. 9.
slightly more noisy than the DKF.
An examination of Figs. 3 and 5 shows that both the estima- load; in fact, the maximum difference between measured and
tors give good results. In fact, in both cases, the speed tracks estimates speeds is in the interval [−2.5,0] rad/s, with a mean
the reference one, the mean error is zero, and the maximum displacement of about −1 rad/s for RDKF, whereas it is in
difference between measured and estimate speeds is less than the interval [−5, −3] rad/s, with a mean displacement of about
±1 rad/s. Spikes are due to the resolution of the encoder. Figs. 4 −4 rad/s for the DKF. The estimated current behavior is quite
and 6 show that both estimators are able to reproduce measured similar to that of the previous test and, then, are not reported
currents. Obviously, acting on the elements of matrix Q, it is here for the sake of brevity.
possible to conveniently filter these currents. Figs. 13 and 14 show the responses at 3 rad/s, also in presence
Figs. 7 and 8 show the closed-loop responses corresponding of a 4-N · m step load torque applied at 2 s and removed at
to a trapezoidal reference speed in the presence of a load torque 13 s. A comparison of that figures show that the RDKF has a
of 5 N · m applied at 2 s and removed at 13 s. A comparison good behavior also at load, with a maximum difference between
of that figures shows that the RDKF works better than DKF at measured and estimate speeds in the interval [−2, 2] rad/s, with
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
1452 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 61, NO. 3, MARCH 2014

Fig. 15. Speed response of the system with feedback from the RDKF at step
Fig. 12. Current estimation error of the system with feedback from the DKF reference speed and no load. The machine is fluxed at zero reference speed. At
at low speed reference. Same operating conditions of Fig. 9. 0.5 s, it is started with a step of 70 rad/s.

Fig. 13. Speed response of the system with feedback from the RDKF at low Fig. 16. Speed response of the system with feedback from the DKF at step
reference speed. The machine is fluxed at zero reference speed. At 1 s, it is reference speed and no load. Same operating conditions of Fig. 15.
started with a step of 3 rad/s. A load torque of 4 N · m is applied at 2 s and
removed at 13 s. Finally, the reference speed is put to zero.
RDKF works better then the DKF for step reference speed vari-
ations, particularly during transients. We want to point out again
that our experiments are carried out on an induction-motor
drive in which the estimated variables are employed for
closing the control loops. In this paper, we show results at
3 rad/s at no load and load, even if we also reach lower speeds
(1–2 rad/s) but with a worse speed waveform. In our opinion,
this is due to the nonlinear behavior of the brake, particularly at
low speed, which in certain conditions blocks the motor.

VIII. C ONCLUSION
In this paper, speed and rotor flux estimators are designed
for sensorless control of motion control systems with induction
Fig. 14. Speed response of the system with feedback from the DKF at low
speed reference. Same operating conditions of Fig. 13. motors. More precisely, the estimators consist of an intercon-
nection of an adaptive speed estimation scheme and either a
a mean displacement of about −1 rad/s. In addition, in these robust or standard descriptor-type KF. It is shown that the
difficult operating conditions, the RDKF appears better on the descriptor structure of the KF allows for a direct translation
dynamic point of view but, at the same time, more noisily with of parameter variations into coefficient variations of the sys-
respect to the DKF. tem model, which leads to simplifications in the describing
Finally, Figs. 15 and 16 show the speed responses of the uncertainties. The use of a speed estimator separate from flux
system with feedback from the RDKF and the DKF at step ref- one allows us to design a fourth-order linear KF estimator.
erence speed and no load. Only in this experiment, both estima- The DKF displays intrinsic robustness properties with respect
tors are designed assuming the nominal resistances increased to the conventional KF. However, the design of the DKF,
by 30%, and the mutual inductance decreased by 20%, with re- including explicitly robustness requirements, leads to better
spect to the values obtained with the previously described iden- results during load tests in both low and high speed ranges and
tification process. Examination of these figures shows that the during transients, for step reference speed, but at the expense
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.
ALONGE et al.: SENSORLESS CONTROL OF INDUCTION-MOTOR DRIVE 1453

of an increased noise level of the estimate. Note that if a great [18] G. Cirrincione, M. Cirrincione, J. Herault, and S. Van Huffel, “The MCA
accuracy in full-load conditions or low-speed operations are not EXIN neuron for the minor component analysis,” IEEE Trans. Neural
Netw., vol. 13, no. 1, pp. 160–187, Jan. 2002.
required, the DKF can be conveniently used. The whole esti- [19] P. E. Gill and W. Murray, “Algorithms for the solution of the nonlinear
mator scheme is suitable for implementation on a digital signal least-squares problem,” SIAM J. Numer. Anal., vol. 15, no. 5, pp. 977–
processor. Experiments carried out on a prototype show that the 992, Oct. 1978.
[20] G. R. Slemon, “Modelling of induction machines for electric drives,”
estimator scheme proposed in this paper is particularly suitable IEEE Trans. Ind. Appl., vol. 25, no. 6, pp. 1126–1131, Nov./Dec. 1989.
for sensorless control of induction-motor drive applications. [21] M. Cirrincione, M. Pucci, G. Cirrincione, and G.-A. Capolino, “Con-
strained minimization for parameter estimation of induction motors in
saturated and unsaturated conditions,” IEEE Trans. Ind. Electron., vol. 52,
R EFERENCES no. 5, pp. 1391–1402, Oct. 2005.
[1] B. Karanayil, M. F. Rahman, and C. Grantham, “Online stator and rotor [22] D. G. Holmes, B. P. McGrath, and S. G. Parker, “Current regulation
resistance estimation scheme using artificial neural networks for vector strategies for vector-controlled induction motor drives,” IEEE Trans. Ind.
controlled speed sensorless induction motor drive,” IEEE Trans. Ind. Electron., vol. 59, no. 10, pp. 3680–3689, Oct. 2012.
Electron., vol. 54, no. 1, pp. 167–176, Feb. 2007. [23] F. Alonge, F. D’Ippolito, F. Raimondi, and A. Urso, “Method for design-
[2] M. Barut, S. Bogosyan, and M. Gokasan, “Experimental evaluation of ing PI-type fuzzy controllers for induction motor drives,” Proc. Inst. Elect.
braided EKF for sensorless control of induction motors,” IEEE Trans. Ind. Eng.—Control Theory Appl., vol. 148, no. 1, pp. 61–69, Jan. 2001.
Electron., vol. 55, no. 2, pp. 620–632, Feb. 2008. [24] H.-B. Shin and J.-G. Park, “Anti-windup PID controller with integral state
[3] M. Barut, S. Bogosyan, and M. Gokasan, “Speed-sensorless estimation predictor for variable-speed motor drives,” IEEE Trans. Ind. Electron.,
for induction motors using extended Kalman filters,” IEEE Trans. Ind. vol. 59, no. 3, pp. 1509–1516, Mar. 2012.
Electron., vol. 54, no. 1, pp. 272–280, Feb. 2007.
[4] C. Lascu, I. Boldea, and F. Blaabjerg, “Very-low-speed variable-structure
control of sensorless induction machine drives without signal injection,”
IEEE Trans. Ind. Appl., vol. 41, no. 2, pp. 591–598, Mar./Apr. 2005. Francesco Alonge (M’02) was born in Agrigento,
[5] M. Barut, R. Demir, E. Zerdali, and R. Inan, “Real-time implementation Italy, in 1946. He received the Laurea degree in elec-
of bi input-extended Kalman filter-based estimator for speed-sensorless tronic engineering from the University of Palermo,
control of induction motors,” IEEE Trans. Ind. Electron., vol. 59, no. 11, Palermo, Italy, in 1972.
pp. 4197–4206, Nov. 2012. Since then, he has been with the University of
[6] F. Alonge and F. D’Ippolito, “Robustness analysis of an extended Kalman Palermo, where he is currently a Full Professor of
filter for sensorless control of induction motors,” in Proc. IEEE ISIE, automatic control with the Department of Energy,
2010, pp. 3257–3263. Information Engineering, and Mathematical Models.
[7] L. Idkhajine, E. Monmasson, and A. Maalouf, “Fully FPGA-based sen- His research interests include electrical drive control
sorless control for synchronous ac drive using an extended Kalman filter,” (including linear and nonlinear observers, stochastic
IEEE Trans. Ind. Electron., vol. 59, no. 10, pp. 3908–3918, Oct. 2012. observers, and parametric identification), robot con-
[8] S. Jafarzadeh, C. Lascu, and M. S. Fadali, “State estimation of induction trol, parametric identification and control in power electronics, and motion
motor drives using the unscented Kalman filter,” IEEE Trans. Ind. Elec- control of unmanned aerial vehicles in aeronautics.
tron., vol. 59, no. 11, pp. 4207–4216, Nov. 2012.
[9] M. F. Iacchetti, “Adaptive tuning of the stator inductance in a rotor-
current-based mras observer for sensorless doubly fed induction-machine Filippo D’Ippolito (M’00) was born in Palermo,
drives,” IEEE Trans. Ind. Electron., vol. 58, no. 10, pp. 4683–4692, Italy, in 1966. He received the Laurea degree in
Oct. 2011. electronic engineering and the Research Doctorate
[10] M. Cirrincione, M. Pucci, G. Cirrincione, and G.-A. Capolino, “A new degree in systems and control engineering from the
TLS-based MRAS speed estimation with adaptive integration for high- University of Palermo, Palermo, Italy, in 1991 and
performance induction machine drives,” IEEE Trans. Ind. Appl., vol. 40, 1996, respectively.
no. 4, pp. 1116–1137, Jul./Aug. 2004. He is currently a Research Associate with the
[11] M. Cirrincione, M. Pucci, G. Cirrincione, and G.-A. Capolino, “An adap- Department of Energy, Information Engineering, and
tive speed observer based on a new total least-squares neuron for induction Mathematical Models, University of Palermo. His
machine drives,” in Conf. Rec. 39th IEEE IAS Annu. Meeting, 2004, vol. 2, research interests include control of electrical drives,
pp. 1350–1361. adaptive and visual/force control of robot manipula-
[12] A. R. Teja, C. Chakraborty, S. Maiti, and Y. Hori, “A new model reference tors, and control of electrical power converters.
adaptive controller for four quadrant vector controlled induction motor Dr. D’Ippolito received the 2000 Kelvin Premium from the Institution of
drives,” IEEE Trans. Ind. Electron., vol. 59, no. 10, pp. 3757–3767, Electrical Engineers, for the paper Parameter identification of induction motor
Oct. 2012. model using genetic algorithms.
[13] K. Ohyama, G. M. Asher, and M. Sumner, “Comparative analysis of
experimental performance and stability of sensorless induction motor
drives,” IEEE Trans. Ind. Electron., vol. 53, no. 1, pp. 178–186, Feb. 2006. Antonino Sferlazza (S’12) was born in Palermo,
[14] J. Y. Ishihara, M. H. Terra, and J. C. Campos, “Robust Kalman filter for Italy, in November 1987. He received the Master’s
descriptor systems,” IEEE Trans. Autom. Control, vol. 51, no. 8, p. 1354, degree in automation engineering from the Univer-
Aug. 2006. sity of Palermo, Palermo, Italy, in 2011. He is cur-
[15] A. H. Sayed, “A framework for state-space estimation with uncertain rently working toward the Ph.D. degree in system
models,” IEEE Trans. Autom. Control, vol. 46, no. 7, pp. 998–1013, and control engineering in the Department of Energy,
Jul. 2001. Information Engineering, and Mathematical Models,
[16] C. C. De Wit, A. Youssef, J. Barbot, P. Martin, and F. Malrait, “Observ- University of Palermo.
ability conditions of induction motors at low frequencies,” in Proc. 39th His research interests include the development of
IEEE Conf. Decision Control, 2000, vol. 3, pp. 2044–2049. feedback control algorithms for nonlinear dynami-
[17] J. C. Moreira and T. A. Lipo, “Modeling of saturated AC machines includ- cal systems, optimization techniques, estimation of
ing air gap flux harmonic components,” IEEE Trans. Ind. Appl., vol. 28, stochastic dynamical systems, and applications of control of electrical drives,
no. 2, pp. 343–349, Mar./Apr. 1992. power converters, and mechanical systems.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on January 16,2024 at 20:25:20 UTC from IEEE Xplore. Restrictions apply.

You might also like