CISTEM’18 - Algiers, Algeria, October 29-31, 2018

Integral Backstepping Control for Double Star

Induction Machine (DSIM)
Samir Zeghlache Ali Djerioui
Noureddine Layadi
Laboratoire d’Analyse des Signaux et Laboratoire de Génie Electrique
Laboratoire de Génie Electrique
Systèmes Departement of Electrical Engineering
Departement of Electrical Engineering
Departement of Electronics Faculty of technology, University
Faculty of technology, University
Faculty of technology, University Mohamed Boudiaf of M’Sila
Mohamed Boudiaf of M’Sila
Mohamed Boudiaf of M’Sila BP 166, Ichbilia, M’Sila, 28000,
BP 166, Ichbilia, M’Sila, 28000,
BP 166, Ichbilia, M’Sila, 28000, Algeria
Algeria
Algeria [email protected]
[email protected]
[email protected]
Fouad Berrabah
Azeddine Houari
Mohamed-Fouad Benkhoris Departement of Electrical Engineering
IREENA Laboratory
IREENA Laboratory Faculty of technology, University
University of Nantes, Saint-Nazaire,
University of Nantes, Saint-Nazaire, Mohamed Boudiaf of M’Sila
France
France BP 166, Ichbilia, M’Sila, 28000,
[email protected]
Mohamed-Fouad.Benkhoris@univ- Algeria
nantes.fr [email protected]

Abstract—This paper presents an integral backstepping In this paper, an integral backstepping controller is
control (IBSC) for double star induction machine (DSIM) in designed for speed, flux and current regulators. Integral
order to steer the speed and flux to the desired references. The action allows successful rejection of disturbances and
DSIM is powered by two three phase voltage source inverters ensures the elimination of the static error. Some tests are
(VSIs) using pulse width modulation (PWM). The stability of carried out to show the robustness of the proposed approach.
the closed loop is verified by exploitation of Lyapunov’s theory. Stability analysis of the system is demonstrated using the
To proof the effectiveness of the proposed control, a Lyapunov’s theory. Compared with existing works, the main
comparative study is made with rotor field oriented control contributions of this paper can be summarized as follows:
(RFOC) based on regulators proportional integral (PI).
Simulation results via MATLAB/SIMULINK show that the • In this paper, the application of the integral
IBSC has a fast dynamic, better load disturbances rejection backstepping control on a double-star induction
and better tracking performance. machine is carried out for the first time.
Keywords—double star induction machine, integral • The present work has added an integral action to the
backstepping control, RFOC control law developed in [11], which can significantly
reduce the static error in case of critical or faulty
I. INTRODUCTION operation.
The double star induction machine belongs to the • Compared to work in [12], the proposed control has
multiphase machines category. It has been proposed for been applied on a more complex machine, where the
different fields of industry that need high power such as coupling between the six stator phases is strong,
electric hybrid vehicles, locomotive traction, ship propulsion especially in the case of operation in abnormal
and many other applications where the safety condition is conditions, in addition, the application of the
required such as aerospace and offshore wind energy proposed control on a DSIM is more advantageous
systems. DISM not only guarantees a decrease of rotor because nowadays the multiphase induction machine
harmonics currents and torque pulsations but also has many is more used in the industry than the traditional
other advantages such as: reliability, power segmentation and induction motor.
higher efficiency. DSIM has a greater fault tolerance; it can
continue operating even with open-phase faults thanks to the • Unlike to [13], where authors propose a sliding mode
greater number of degrees of freedom that it owns compared direct field-oriented control of six-phase induction
to the three-phase machines [1-7]. machine, the chattering phenomenon was eliminated.

The integral backstepping is a kind of backstepping This paper is organized as follows: In section 2, the
method with integral action, recently, several searches based description of the DSIM and the dynamics model are given.
on this control method were made; [8] presents an integral In section 3, the integral backstepping control is designed.
backstepping adaptive H controller for a class of switched Simulation results are shown and analyzed in section 4.
nonlinear systems with unknown parameters and external Finally, the conclusion is given in section 5.
disturbance, [9] proposes an IBSC combined with high gain
observer for the optimization of grid connected doubly-fed II. DSIM MODELLING
induction generator. Moreover, for electric vehicle The DSIM has two stators shifted by an electrical angle 
application, [10] developed an IBSC for a permanent magnet and mobile squirrel cage rotor composed by three phases.
synchronous motor (PMSM), this controller is adapted via Each star is composed by three immovable windings [4].
the estimation of the unknown load torque and friction Fig. 1 shows an explicit schema which represents the DSIM
effects. windings. The windings series (Sa1, Sb1, Sc1), (Sa2, Sb2, Sc2),

978-1-5386-4988-6/18/$31.00 ©2018 IEEE

 CISTEM’18 - Algiers, Algeria, October 29-31, 2018

(Ra, Rb, Rc) represent stator1, stator2 and rotor respectively, d 1ª Lm º

 is the angle shift between the two stators,  is the one ωr = « p 2 ϕ r (isq1 + isq 2 ) − pTL − K f ωr » (6)
dt J ¬ Lm + Lr ¼
between rotor and stator1, in this research we put =30° with
accordance to [3]. In order to establish the mathematical d Rr L R
ϕr = − ϕ r + m r (isd 1 + isd 2 ) (7)
model of DSIM, the following assumptions are made: dt Lm + Lr Lm + Lr
• The air-gap is uniform.
Where: vsd1, vsq1 are stator1 voltages components. vsd2, vsq2
• Magnetic linearity (negligible saturation). are stator2 voltages components. isd1, isq1 are stator1 currents
• No space harmonic. components. isd2, isq2 are stator2 currents components. Ls1, Ls2,
Lr and Lm are stator1, stator2, rotor and mutual inductance,
• The stator windings are identical. respectively. Rs1, Rs2 and Rr are respectively stator1, stator2
and rotor resistance. Tr= Lr/ Rr is the rotor time constant. TL
• The two stars have the same parameters. is the applied load torque. r is the rotor speed. J, Kf denote
• The magnetic circuit is symmetrical. the rotor inertia and friction coefficient. s is the stator
pulsation. gl is the slip pulsation. r is the rotor flux. p
• The slotting effect is negligible. denotes the number of pole pairs. The subscripts d, q
This paper takes into consideration eddy currents; their designate direct and quadrate indices according to the usual
mathematical model is represented by the following d-q axis components in the synchronous rotating frame
equation:
III. INTEGRAL BACKSTEPPING CONTROL
J ed = aE (1)
A. Step 1: Speed and Flux Control
Where Jed is the eddy current density, a is the electrical
conductivity of materials, E is the electric field strength. The aim of this step is to lead the flux ϕ r and the speed
ωr to their desired references ϕ r* and ωr* respectively. The
tracking errors of speed and flux are given by:
Ra
S b1 t
­Z1 = ωr* − ωr + Kω (ωr* − ωr ) dt
³0
or

°
θ (8)
Rot

® t
°¯Z 2 = ϕr* − ϕr + Kϕ ³ (ϕ r* − ϕ r ) dt
2

S a2
Sb

0
2
t or
St a Where Kω and Kϕ are a positive design constants.
α The dynamics of tracking errors became:

Stator1
S a1
­Z1 = ω r − ω r + Kω (ωr − ωr )
* *

® (9)
¯Z 2 = ϕ r − ϕ r + Kϕ (ϕr − ϕ r )
* *

Rb Rc
Using (6), (7) and (9), the dynamics of tracking errors
became:
Sc1 ­Z = ω * − 1 ª p 2 Lm ϕ (i + i ) − pT − K ω º
Sc 2 ° 1 r
J «¬ Lm + Lr
r sq1 sq 2 L f r»
¼
°°
+ Kω (ωr − ωr )
* (10)
®
° Rr L R
Fig. 1. DSIM windings.
°Z 2 = ϕ r* + ϕ r − m r (isd 1 + isd 2 ) + Kϕ (ϕr* − ϕr )
°̄ Lm + Lr Lm + Lr
The d-q dynamic healthy model of squirrel cage double
star induction machine with a reference frame fixed to the
rotor is given by: The first Lyapunov’s function candidate adapted to the
rotor flux and speed is defined by:
d 1
isd 1 = [vsd 1 − Rs1isd1 + ωs (Ls1isq1 + Trϕr ωgl )] (2) 1
dt Ls1 V1 = (Z12 + Z 22 ) (11)
2
d 1
isq1 = [vsq1 − Rs1isq1 − ωs ( Ls1isd 1 + ϕr )] (3)
dt Ls1 The derivative of (11), according to the time is:
d 1
isd 2 = [vsd 2 − Rs 2isd 2 + ωs (Ls 2isq 2 + Tr ϕr ωgl )] (4)
dt Ls 2 V1 = Z1 Z1 + Z 2 Z 2 (12)
d 1
isd 2 = [vsd 2 − Rs 2isd 2 + ωs (Ls 2isq 2 + Tr ϕr ωgl )] (5) In order to have the derivative of Lyapunov’s function
dt Ls 2
negative definite we pose:
 CISTEM’18 - Algiers, Algeria, October 29-31, 2018

­°Z1 = − K1 Z1
t
­Z 3 = isd* 1 − isd 1 + K isd 1 ³ (isd* 1 − isd 1 ) dt
® (13) ° 0

°
¯°Z 2 = − K 2 Z 2
t

°Z 4 = isq* 1 − isq1 + K isq1 ³ (isq* 1 − isq1 ) dt

0 (20)
® t
°Z 5 =i *
sd 2 − isd 2 + K isd 2 ³ (i *
sd 2 − isd 2 ) dt
Where, K1 and K 2 are positive gains. This chose is ° 0

°Z =i *
− isq 2 + Kisq 2 ³ (i
t
*
− isq 2 ) dt
correct since we have: ¯ 6 sq 2 0 sq 2

V1 = − K1 Z12 − K 2 Z 22 < 0 (14) Where K isd 1 , K isq1 , K isd 2 and K isq 2 are a positive design
constants.
From (10) and (13) we obtain: The time derivate of (20) gives:

­− K Z = ω * − 1 ª p 2 Lm ϕ (i + i ) − pT − K ω º ­Z d * d
= isd 1 − isd 1 + K isd 1 (isd* 1 − isd 1 )
J «¬ Lm + Lr
° 1 1 f r»
r r sq1 sq 2 L
¼ ° 3 dt dt
° °
° °Z 4 d d
= isq* 1 − isq1 + K isq1 (isq* 1 − isq1 )
° + Kω (ωr − ωr )
*
(15) ° dt dt (21)
® ®
° R L R °Z d d
*
°− K 2 Z 2 = ϕr + L + L ϕr − L + L (isd 1 + isd 2 )
r m r
= isd* 2 − isd 2 + K isd 2 (isd* 2 − isd 2 )
° 5 dt dt
° m r m r
°
° °Z 6 d d
= isq* 2 − isq 2 + Kisq 2 (isq* 2 − isq 2 )
¯ + Kϕ (ϕr − ϕ r )
*
¯ dt dt

Assuming that: From (2) - (5), (21) can be written as follows:

∗
­°isq1 + isq 2 = isq ­Z d * 1
® (16) ° 3
= isd 1 − [vsd 1 − Rs1isd 1 + ωs (Ls1isq1 + Trϕr ωgl )]
dt Ls1
¯°isd 1 + isd 2 = isd
∗ °
° + K isd 1 (isd* 1 − isd 1 )
°
And: ° d * 1
°Z 4 = isq1 − [vsq1 − Rs1isq1 − ωs (Ls1isd 1 + ϕ r )]
° dt Ls1
­∗ isq∗ °
∗ °° + K isq1 (isq* 1 − isq1 ) (22)
°°isq1 = isq 2 = 2 ®
® (17) °Z d * 1
°∗ ∗ isd∗ ° 5
= isd 2 − [vsd 2 − Rs 2isd 2 + ωs (Ls 2isq 2 + Trϕr ω gl )]
°̄isd 1 = isd 2 = °
dt Ls 2
2 ° + K isd 2 (isd* 2 − isd 2 )
°
Where: isd∗ and isq∗ are the references of current ° d * 1
°Z 6 = isq 2 − [vsq 2 − Rs 2 isq 2 − ωs ( Ls 2isd 2 + ϕ r )]
components that stabilize speed and flux. By substituting ° dt Ls 2
(16) into (15), we obtain the following virtual control: °
°̄ + K isq 2 (isq* 2 − isq 2 )

­ ∗ J ( Lm + Lr ) ª  p Kf º
°isq = p 2 L ϕ «¬ωr + k1 z1 + kω eω + J TL + J ωr »¼
The increasing of Lyapunov’s function which analysis
° m r the system stability is defined by:
® (18)
° ∗ ( Lm + Lr ) ª ∗ Rr º 1
°isd = ϕ r + ϕr + k2 z2 + kϕ eϕ » V2 = [V1 + (Z32 + Z 42 + Z52 + Z 62 )] (23)
¯ Lm Rr «¬ Lm + Lr ¼ 2

With: Its time derivative is:

­eω = ω r − ω r
*
V2 = V1 + Z 3 Z3 + Z 4 Z 4 + Z 5 Z 5 + Z 6 Z 6 (24)
® (19)
¯eϕ = ϕ r − ϕ r
*

The system global stability is achieved if only V2

definite negative therefore Z 3 , Z 4 , Z 5 and Z 6 are chosen as
B. Step 2:Currents control in the first step:
In this stage, the control law is established by adjusting
the four stator currents: isd 1 , isq1 , isd 2 and isq 2 generated by ­°Z 3 = − K3 Z 3 ; Z 5 = − K 5 Z 5
the first step, so we specify also their error signals: ® (25)
¯°Z 4 = − K 4 Z 4 ; Z 6 = − K 6 Z 6
 CISTEM’18 - Algiers, Algeria, October 29-31, 2018

Where K 3 , K 4 , K 5 and K6 are positives gains that

stabilize the currents in closed-loop, usually they are
obtained by adjustment. By substituting (25) and (14) into
(24), V2 became:

V2 = − K1 Z12 − K 2 Z 22 − K 3 Z 32 − K 4 Z 42 − K 5 Z 52 − K 6 Z 62 < 0 (26)

We show clearly that (26) checks the stability according

to the Lyapunov’s theory. By using (22) and (25), we obtain
the final control represented by the following components of
stator voltages:

­ d * Rs1 Trϕ r ωsω gl

°vsd 1 = Ls1[ dt isd 1 + L isd 1 − ωs isq1 − Ls1
° s1

°
° + K isd 1 (isd* 1 − isd 1 ) + K 3 Z 3 ]
° (a)
° d * Rs1 ϕω
°vsq1 = Ls1[ isq1 + isq1 + ωs isd 1 + r s
° dt L s1 Ls1
°
° + K isq1 (isd* 1 − isd 1 ) + K3 Z 3 ]
°
® (27)
° d * Rs 2 Tr ϕr ωsω gl
°vsd 2 = Ls 2 [ isd 2 + isd 2 − ωs isq 2 −
° dt L s2 Ls 2
°
° + K isd 2 (isd* 2 − isd 2 ) + K5 Z 5 ]
°
° d * R ϕω
°vsq 2 = Ls 2 [ isq 2 + s 2 isq 2 + ωs isd 2 + r s
° dt L s2 Ls 2
°
°¯ + K isq 2 (isq* 2 − isq 2 ) + K 6 Z 6 ]

IV. SIMULATION RESULTS AND INTERPRETATION

In this section, the effectiveness and robustness of the
(b)
proposed control is clarified through simulation results by
using MATLAB/SIMULINK. The DSIM studied in this
paper is fed by two PWM-VSI, its parameters are as follows: Rotor flux
Voltage: 230-380 V, Power: 4.5kw, Frequency f=50 Hz. The 1.2
nominal electrical and mechanical parameters are given in
table 1. A load torque (TN=15N.m) is applied at t=2sec. 1

0.8
TABLE I. MACHINE PARAMETERS [3]

Parameters Identifiers & values 0.6

Output power Pn = 4.5 kw
Stator1 resistance Rs1 = 3.72  0.4
Stator2 resistance Rs2 = 3.72  phir*
Rotor resistance Rr = 2.12  phir (RFOC)
0.2
Stator1 inductance Ls1 = 0.022 H
phir (IBSC)
Stator2 inductance Ls2 = 0.022 H
Rotor inductance Lr = 0.006 H 0
Mutuel inductance Lm = 0.3672 H
0 1 2 3 4 5
Inertia J = 0. 0625 Kgm2 time (s)
damping coefficient Kf = 0.001 Nm.(rd/s)-1 (c)
Number of pole pairs p=1
 CISTEM’18 - Algiers, Algeria, October 29-31, 2018

(a)
(d)

Fig. 2. Simulation results of IBSC and RFOC

Fig. 2.a shows the speed responses for the two control
methods, in both signals the speed follows its reference value
(300rd/s) without overshoot and without oscillations, but it is
clearly shows that the proposed IBSC has faster response
than RFOC and imposes a short transient regime. The
proposed control also provides better stability with the
smallest average static error. No ripples in the
electromagnetic signals as illustrated in Fig. 2.b proving that
both control schemes are able to surmount the external load
torque rejection. At start-up the electromagnetic torque
presents oscillations and a peak of 53 N.m, 82 N.m for
RFOC and IBSC, respectively. After the transient regime
where IBSC has the better dynamic response, the torque
compensates the friction losses and load torque. Fig. 2.c
show that the two control methods are suitable to lead the
flux to it desired reference but the IBSC has the fastest
dynamic response than RFOC. Finally, Fig. 2.d shows the (b)
behavior of the line current isa1 of the first stator which has a
sinusoidal waveform affected by the switching frequency Fig. 3. DSIM performance with IBSC (Te=1.5TN, 2.5sec < t < 3.5sec)
generated by the inverter, during the dynamic state, the
DSIM consumes very important currents, when inserting the
load, the amplitude of currents reach 7A for both control
methods. The obtained results summarize and reflect the
response swiftness of the IBSC compared to the RFOC.
To evaluate the robustness of the integral backstepping
control of a double star induction machine, against external
disturbances, step changes of load torque (Cr=1.5CrN
between t= 2.5sec and t=3.5sec).
Fig. 3 shows that the abrupt variation of the load torque
has a slight influence on the performance of the machine,
which proves the robustness of the proposed control against
load torque disturbances.
In order to prove the robustness of the proposed control
against parametric variation, disturbances of 50% of nominal
values are introduced into the rotor resistive parameter
(increasing of 50% in Rr) and into the mechanical parameter
(increasing of 50% in J) followed by a reversal of speed at
t=2.5sec where the reference speed is fixed at +200rd/s and (a)
-200 rd/s.
 CISTEM’18 - Algiers, Algeria, October 29-31, 2018

[2] A. Meroufel, S. Massoum, A. Bentaallah, P. Wira, F. Z. Belaimeche

Electromagnetic torque and A. Masssoum, “Double Star Induction Motor Direct Torque
100
Control with Fuzzy Sliding Mode Speed Controller”, Rev. Roum. Sci.
T
L Techn.–Électrotechn. et Énerg., vol.62, no.1, pp. 31-35, 2017.
T (IBSC) [3] H. Rahali, S. Zeghlache and L. Benalia, “Adaptive Field-Oriented
e

50 Control Using Supervisory Type-2 Fuzzy Control for Dual Star

Induction Machine”, International Journal of Intelligent Engineering
and Systems, vol.10, no.4, pp. 28-40, 2017.
[4] K. Iffouzar, M. F. Benkhoris, K. Ghedamsi, D. Aouzellag, “Behavior
0 analysis of a dual stars induction motor supplied by pwm multilevel
inverters”, Rev. Roum. Sci. Techn. – Électrotechn. Et Énerg., vol.61,
no.2, pp. 137–141, 2016.
[5] N. Mokhtari, M. F. Benkhoris, M. Merabtene and R. Le Doeuff,“
-50 Vectoriel modeling approach of multi-star machine supplied by
voltage source inverters”, International Conference on Electrical
Machines, ICEM'2004,Cracow, Poland, Sept 2004, pp. 5-8.
[6] K. Iffouzar, B. Amrouche, C. Otmane, M. F. Benkhoris, D. Aouzellag
-100 and K. Ghedamsi, “Improved direct field oriented control of
0 1 2 3 4 5 multiphase induction motor used in hybrid electric vehicle
application”, International Journal of Hydrogen Energy, vol.42, no.30,
time (s)
pp. 19296-19308, 2017.
(b) [7] N. Madani, M.F. Benkhoris, C. Millet and R. LE Doeuff,
“Sliding mode control of double stars asynchronous machine-two
PWM inverters set”, 11th International Power Electronics
Fig. 4. DSIM performance with IBSC (Rr=1.5RN , J=1.5JN)
and Motion Control Conferences EPE PEMC, Riga, Latvia, Sept
(Te=TN=15N.m at t=2sec) and (wr=-200 rd/s at t=2.5sec) 2004, pp. 2-4, CD-ROM.
[8] X. Cheng, X. Li and X. Liu, “Integral backstepping adaptive H
Fig. 4 clearly demonstrates that the proposed control control for a class of switched nonlinear systems”, Proceedings of the
(IBSC) guarantees complete robustness against parametric IEEE 29th chinese control and decision conference, Chongqing,
variation even in the presence of a load torque. At t=2.5sec, China, May 2017, pp. 1970-1975.
IBSC also provides a precise speed reversal with good [9] O. Adekanle, M. Guisser, M. E. Abdelmounim, “Integral
reference tracking (-200rd/s), as shown in Fig. 4.a, the speed Backstepping controller combined with High Gain Observer for the
optimization of grid-connected Doubly-Fed Induction Generator”,
variation affects the electromagnetic torque; a peak of - Proceedings of the IEEE international conference on wireless
59N.m is recorded at t=2.5sec, but after a fast dynamic, the technologies, embedded and intelligent systems, Fez, Morocco, April
torque compensates the friction losses and the load torque 2017, pp. 1-7.
(see Fig. 4.b). [10] S. Nadji, S. Benaicha, ”Robust integral backstepping control of
PMSM for electric vehicle application”, Proceedings of the 5th IEEE
international conference on electrical engineering, Boumerdes,
V. CONCLUSION Algeria, Oct 2017, pp.1-6.
In this paper, a nonlinear integral backstepping control [11] N. Layadi, S. Zeghlache, T. Benslimane and F. Berrabah, “
has been developed to control a double star induction Comparative Analysis between the Rotor Flux Oriented Control and
Backstepping Control of a Double Star Induction Machine (DSIM)
machine in order to improve its performance. Simulation under Open-Phase Fault”, AMSE Journals, Series Advances C,
results show that the proposed control is not only robust vol.72, no.4, pp. 292-311, 2018.
against variations in internal and external parameters but also [12] M. Fateh and R. Abdellatif, “Comparative study of integral and
guarantees high performances in both dynamic and steady classical backstepping controllers in IFOC of induction motor fed by
state mode where the tracking errors converge to zero. voltage source inverter”, International Journal of Hydrogen Energy,
Compared with RFOC, IBSC can track the reference values vol. 42, no. 28, pp. 17953-17964, 2017.
precisely and quickly. As future work, the IBSC will be [13] J. Listwan and K. Piekowski, “Sliding-mode direct field-oriented
injected into a fault-tolerant control system. control of six-phase induction motor”, Technical Transactions, no.( 2-
M), pp. 95-108, 2016.

REFERENCES
[1] Z. Tir, Y. Soufi, M. N. Hashemnia, O. P. Malik and K. Marouani,
“Fuzzy Logic Field Oriented Control of Double Star Induction Motor
Drive”, Electrical Engineering, vol. 99, no. 2, pp. 495-503, 2017.