Model Predictive Control of Induction Motor with
Delay Time Compensation: An Experimental
Assessment
Muslem Uddin, Saad Mekhilef, Mutsuo Nakaoka
Power Electronics and Renewable Energy Research
Laboratory (PEARL), Department of Electrical Engineering
University of Malaya
50603 Kuala Lumpur, Malaysia
Email: muslem_eee04@siswa.um.edu.my
saad@um.edu.my
Abstract—This paper proposes a delay time
compensation method in the model predictive control
(MPC) of induction motor (IM) at high and low speed
considering the selection of optimum switching vector to
actuate the three-phase voltage source inverter (VSI).
The proposed control method compensates the delay
time to improve the performance of the system, and
consequently maintain the accurate tracking of the
references at different speed regions. The control scheme
utilizes the discrete nature of the system, and uses the
possible switching vectors of the converter in an intuitive
manner. Therefore, based on minimum quality function
the optimum switching vector is selected for the next
sampling time actuation of the power converter. The
control scheme is validated through the MATLAB
simulation and experimental validation in DS1104 R&D
Controller Platform. The simulation and experimental
results prove the feasibility of the proposed method with
encouraging performance.
Keywords— Predictive Control; Power Converter; Delay
Time Compensation; Induction Motor; Torque & Flux.
I. INTRODUCTION
AC machine drives are mostly important in the industrial
applications. To control the electrical drives, two types of
control methods were widely used during the last two
decades in the industries, such as; direct torque control
(DTC) [1], [2] and field oriented control (FOC). The losses
associated with the DTC can be controlled with proper
utilization of space vector modulation, and imposing the
predictive torque control (PTC) algorithm [3]. As the
predictive control method is simple and easy to implement
due to the outstanding development of the digital signal
processor, this control method has been earned a great
concern to research community. However, this predictive
control can be applied to power electronic converter,
because this control can utilize the discrete nature of power
converter to predict the future behavior of the system.
The authors wish to thank the financial support from the University of
Malaya through HIR-MOHE project UM.C/HIR/MOHE/ENG/17.
978-1-4799-6735-3/15/$31.00 ©2015 IEEE
Marco Rivera
Department of Industrial Technologies
Universidad de Talca
Curico, Chile
Email: marcoesteban@gmail.com
Recently, many works have been investigated efficiently
with the predictive control algorithm which have been
proved the robustness and feasibility of the predictive
control method for power converter [4].
Therefore, a predictive current control fed by voltage
source inverter [5]; predictive torque and flux control of IM
fed by indirect matrix converter (IMC) with unity power
factor control at the input side [3]; unity power factor
control at the grid side of an active front end rectifier [6],
[7]; multilevel inverter fed induction motor predictive
control [8]; direct matrix converter (DMC) fed predictive
torque control of IM with reactive power compensation [9];
predictive current control of a three-phase four-leg inverter
[10]; predictive control of bidirectional AC-DC converter
[11]; torque ripple reduction of IM with predictive direct
torque control method [12], and a weighting factor
optimization method in the predictive control algorithm for
reduction of torque ripple [13], [14] have been investigated
in the recent decades. A complete review for the
improvement of input current as well as total harmonic
distortion (THD) analysis at the input and output side of the
IMC with predictive control algorithm was investigated in
[15]. Also, a comprehensive review has been presented to
summarize the model predictive control (MPC) method
applied to power electronics application in [16], and
described the applications of MPC method on four different
power electronics converters and drives.
The experimental verification is associated with a delay
time of microprocessors, which deteriorates the system
performance. Therefore, this paper proposes a delay time
compensation method in the predictive control algorithm to
increase the system performance.
This paper is organized in the following manner: the
second section describes the modeling of the induction
motor; the third section summarizes the proposed delay time
compensated predictive control method with subsections:
first order approximations for predictive variables, the
importance of delay time compensations, consideration of
long prediction horizons for determination of the predicted
variables in the second next (k+2) sampling time instant, an
543
 .
y (k + 1) − y (k ) (7)
y= Ts
The predictive variables (stator flux and rotor flux) of the
induction motor can be determined by applying the first
order Euler’s approximations as following:
ψ s (k ) = ψ s (k − 1) + vo (k )Ts − Rsi(k )Ts (8)
ψ r ( k ) = i ( k )( L m −
Ls Lr L
) + r ψ s (k )
(9)
Lm Lm
Predictive stator flux, stator current and torque in the next
(k+1)th sampling time instant become,
Fig. 1. Power converter topology
ψ s (k + 1) = ψ s (k ) + v0 (k )Ts − Rs i(k )Ts (10)
overall quality function of the MPC method, and a T Ts 1 k (11)
comprehensive predictive control scheme. In the fourth i (k + 1) = (1 + s )i (k ) + [ {( r − k jω )ψ (k ) + v (k )}]
r m r o
τσ τ σ + Ts rσ τ r
section, the simulation and experimental results are discussed
3 (12)
to prove the feasibility of the proposed method. The last Te ( k + 1) = p{ψ s ( k + 1) × i ( k + 1)}
section is concluded with a comprehensive conclusion. 2
where, rσ = Rs + Rr k r2 , τ = Lr , k = L m , k = Lm ,
r s r
Rr Ls Lr
II. MODELING OF INDUCTION MOTOR
σ = 1 − k s k r and τ σ = σLs / rσ .
The power converter topology fed by induction motor
(IM) is presented in Fig. 1. The three-phase system B. Delay time compensation
( ya , yb and yc ) can be represented as a two dimensional
The computational time needed in the predictive control
complex space vector (SV) as below: algorithm to predict the variables, and processors delay
y = y α + jy β (1) deteriorates the performance of the predictive control at the
experimental investigation. To solve this delay problem, it
where, y = 1 (2 y − y − y ) , y = 1 ( y − y ) (2)
α
3
a b c β b c can be considered the prediction horizon at (k+2)th sampling
3
time to predict the variables which are compared with the
In Eq. 2, αβ -reference frame is considered as the references, and determine the quality functions. The
stationary reference frame for expected space vector. The optimum switching vector is selected corresponding to the
model of the IM referred to stator is obtained as described in minimum quality function, and applied it in the next (k+1)th
[17]. The fixed coordinate stator and rotor voltage equations sampling time actuation. As a result, one sampling time is
are presented as, available to compensate the time delay produced by the
. processor.
vo = Rs i + Ls ψ s (3)
. C. Determination of the predictive model for (k+2)th
vr = Rr ir + Lr ψ r − jpωψ s (4)
sampling time instant
where i , stator current; ir , rotor current; R s , stator
resistance; R r , rotor resistance; p , number of pole pairs, Applying the Euler’s Approximation, similarly, for the
second next (k+2)th sampling time instant, the predictive
and ω , rotor angular frequency. stator flux, stator current, and torque become as follows:
The stator flux and rotor flux can be presented as below: (13)
ψ s ( k + 2) = ψ s ( k + 1) + vo ( k )Ts − Rs i(k + 1)Ts
ψ s = Ls i + Lmir , ψ r = Lr ir + Lmi (5)
Ts Ts 1 kr
where, ( L s , L r ), self-inductances; L m , mutual inductance. i ( k + 2) = (1 + )i ( k + 1) + [ {( − k r jω m )ψ r ( k + 1) + v o ( k )}]
τσ τ σ + Ts Rσ τr
The developed electrical torque in the IM can be represented (14)
by stator current and stator flux as the following equation: 3
Te (k + 2) = p{ψ s (k + 2) × i (k + 2)}
(15)
3 (6) 2
Te = p (ψ s × i )
2
D. Quality function
III. PROPOSED PREDICTIVE CONTROL WITH
DELAY TIME COMPENSATION The quality function in the predictive control of IM with
delay compensation is presented as below,
A. Discrete-time model predictive torque and flux control g IM ( k + 2) = Te (k + 2) − Tref + λ * ψ s ( k + 2) − ψ ref (16)

The first order Euler’s approximations can be represented

ψ ref are the reference torque and reference
where, Tref ,
as below: flux, respectively. λ , is the weighting factor, and represent

544
the flux control has a priority control rather than the torque
control.
E. Delay time compensated predictive control scheme
The predictive control scheme and algorithm for
induction motor control are presented in Figs. 2(a) and 2(b),
respectively. The predictive controller satisfies the
following steps:
• First: stator voltage, vo (k ) ; stator current, i(k); and
speed, ωm (k ) of induction motor are measured in
the k th sampling time instant.
(a)
(b)
Fig. 2. (a) Predictive control scheme and (b) algorithm
545
• Second: stator reference flux (ψref) and reference
speed ( ωref ) are known values. Speed controller is
used to set the reference torque ( Tref ).
• Third: estimation of the stator and rotor flux.
• Fourth: predictive torque Te ( k + 1) and predictive
stator flux ψ s (k + 1) are predicted in the next
sampling time period (k+1)th based on measured
variables. And, this predictive torque and flux are
used to predict the (k+2)th predictions of the same
variables for all eight possible switching vectors.
• Fifth: the (k+2)th predictive values are compared
with their respective references, and determine the
quality functions for all the possible switching
states.
• Lastly, the optimum switching vector corresponds
to the minimum cost function is selected for the
next sampling time actuation.
IV. SIMULATION AND EXPERIMENTAL RESULTS
The proposed control method is verified in MATLAB
Simulink as well as experimental validation with DS1104
R&D Controller at different speed regions to justify the
performance of the proposed control scheme. The
parameters used in the simulation and experimentation are
given in Table I.
A. Simulation Results
The Figs. 3 and 4 represent the results of IM control
applied with model predictive control (MPC) method at high
speed and low speed regions, respectively. In both
simulations, the motor start at 0.01s with the reference torque
of 6 Nm in Figs. 3(a) and 4(a). The predictive torque follows
the reference torque with high tracking response in both
verifications. In Figs. 3(a) and 4(a), when the motor obtain
its refrence high speed of 125 rad/s, and 37.125 rad/s,
respectively, the predictive torque becomes at minimum, and
follows the reference torque, accurately.
Furthermore, when the reference speeds have been
changed to negetive direction, the measured speeds are
started to follow the reference speeds at exact time of 1.7s
and 1.2s at high and low speed, respectively, with reverse
high torque of 6 Nm in both the analysis. Once the motor
attained its reverse high speed of -125 rad/s and -37.125
rad/s, the predictive torque becomes zero. At time of 3.5s, a
load torque of 3 Nm has been applied. Consequently, the
predictive controller forced the motor to develop the torque
of 3 Nm to metigate the load demand at high and low speed
regions presented in Figs. 3(a) and 4(a).
Also, the predictive stator flux tracks the reference flux
very well, and maintain the magnitude of 1.0 Wb throughout
the simulation time in Figs. 3(a) and 4(a). The Figs. 3(b) and
4(b) show the αβ − components of the stator current of IM
corresponding to reference changes in the simulations which
show a very encouraging behavior with the phase
displacement of 90o between the two components at forward
and reverse speed regions.

Fig. 3. Simulation results (for,
reference torque [Nm], measured speed vs reference speed [rad/s], and
predictive stator flux vs reference flux [Wb]; (b) αβ − components,
α − component, and β − component of stator current [A].
(a)
(b)
Fig. 4. Simulation results (for, ω m = 37.175 rad/s) (a) predictive torque vs
reference torque [Nm], measured speed vs reference speed [rad/s], and
predictive stator flux vs reference flux [Wb]; (b) αβ − components,
α − component, and β − component of stator current [A].
In forward speed, α -component is lagged by 90o than the
β -component, but at the reverse speed, α -component leads
the β -component by 90o. Therefore, the simulation results
have been proved the effectiveness of the proposed method
by achieving the well tracking of references and good
performances.
B. Experimental results
The Fig. 5 shows an overall laboratory experimental test
setup to validate the simulation results with experimental
implementation in DS1104 R&D controller platform.
(b)
In the experiment the gate signals are generated from the
ωm = 125 rad/s) (a) predictive torque vs DS1104 R&D Controller board by compiling the control
part of the simulation in the DS1104 R&D controller
platform. The three phase stator currents, voltages, and
speed are measured with the current sensors, voltage
sensors, and speed sensors, respectively which are fed to
ADC ports of DS1104 R&D controller board to complete
the closed-loop of the proposed control algorithm. The
experimental results of IM at high speed and low speed
references are presented in the Figs. 6 and 7, respectively.
In Fig. 6(a), the reference speed is taken as 125 rad/s
which is compared to measured speed to get the error signal.
This error of speed acts as the input of the speed controller
which generates the reference torque of 6 Nm, and the
predictive torque follows the reference torque accurately, in
the Fig. 6(a). Also, the predictive stator flux tracks the
reference stator flux of 1.0Wb in Fig. 6(a).
The corresponding αβ−components of the experimental
stator current at high speed of IM in Fig. 6(b) validate the
simulation results. Moreover, the phase changing mode
between the αβ−components has also been presented with
zoomed when the motor changes its speed direction from
forward to reverse, and at load changing moment.
(a)
546
 Similarly, at low speed, the predictive torque, speed,
and predictive flux show the rapid tracking of the reference
torque, reference speed, and stator flux references,
respectively, by keeping the permissible torque ripple which
is presented in Fig. 7(a). Also, the corresponding
αβ −components of experimental stator current have been
shown in Fig. 7(b).
The experimental results proved that the proposed
delay compensated predictive controller has been achieved
the satisfactory results even at very low speed region of IM.

Fig. 5. Experimental setup

(a)

(a)

(b)
(b)
Fig. 7. Experimental results (for low speed, ωm = 37.125 rad/s): (a)
Fig. 6. Experimental results (for high speed, ωm = 125 rad/s): (a)
predictive torque vs reference torque [Nm], measured speed vs reference
predictive torque vs reference torque [Nm], measured speed vs reference speed [rad/s], and predictive stator flux vs reference flux [Wb]; (b)
speed [rad/s], and predictive stator flux vs reference flux [Wb]; (b) αβ − components, α − component, and β − component of stator current
αβ − components, α − component, and β − component of stator current [A].
[A].

547
TABLE I: SIMULATION AND EXPERIMENTAL PARAMETERS [3] S. Muslem Uddin, S. Mekhilef, M. Rivera, and J. Rodriguez, "A
FCS-MPC of an induction motor fed by indirect matrix
converter with unity power factor control," in Proc. 8th IEEE
Description Variables Values Conference on Industrial Electronics and Applications (ICIEA)
Sampling time Ts 25 µs Melbourne, Australia, 2013, pp. 1769-1774.
[4] P. Cortés, M. P. Kazmierkowski, R. M. Kennel, D. E. Quevedo,
DC voltage Vdc 500 V and J. Rodríguez, "Predictive control in power electronics and
Supply frequency f s 50 Hz drives," IEEE Transactions on Industrial Electronics, vol. 55,
pp. 4312-4324, 2008.
ωref 125 rad/s and
Reference speed [5] J. Rodriguez, J. Pontt, C. A. Silva, P. Correa, P. Lezana, P.
37.175 rad/s
Cortés, and U. Ammann, "Predictive current control of a voltage
Nominal torque Tnom 6 Nm source inverter," IEEE Transactions on Industrial Electronics,
vol. 54, pp. 495-503, 2007.
Reference Flux ψ ref 1.0 Wb [6] D. E. Quevedo, R. P. Aguilera, M. A. Pérez, P. Cortés, and R.
Stator resistance Rs 21 Ω Lizana, "Model predictive control of an AFE rectifier with
dynamic references," IEEE Transactions on Power Electronics,
Stator inductance Ls 1.053 mH vol. 27, pp. 3128-3136, 2012.
Rotor resistance Rr 22.63 Ω [7] S. Muslem Uddin, P. Akter, S. Mekhilef, M. Mubin, M. Rivera,
and J. Rodriguez, "Model predictive control of an active front
Rotor inductance Lr 1.081 mH end rectifier with unity displacement factor," in Proc. IEEE
Mutual Inductance Lm 0.9963 mH International Conference on Circuits and Systems (ICCAS)
Kuala Lumpur, Malaysia, 2013, pp. 81-85.
Number of poles p 2 [8] P. Correa, M. Pacas, and J. Rodriguez, "Predictive torque
Moment of inertia J 0.04 Kg m2 control for inverter-fed induction machines," IEEE Transactions
on Industrial Electronics, vol. 54, pp. 1073-1079, 2007.
Weighting Factor λ Tnom ψ ref [9] R. Vargas, U. Ammann, B. Hudoffsky, J. Rodriguez, and P.
Wheeler, "Predictive torque control of an induction machine fed
by a matrix converter with reactive input power control," IEEE
Transactions on Power Electronics, vol. 25, pp. 1426-1438,
2010.
V. CONCLUSION [10] M. Rivera, V. Yaramasu, A. Llor, J. Rodriguez, B. Wu, and M.
Fadel, "Digital predictive current control of a three-phase four-
The simulation and experimental results show that the leg inverter," IEEE Transactions on Industrial Electronics, vol.
predictive control is a promising control tool that is robust 60, pp. 4903-4912, 2013.
and powerful to control the power converters and electrical [11] M. Parvez, S. Mekhilef, N. M. Tan, and H. Akagi, "Model
machine drives. The proposed delay time compensated predictive control of a bidirectional AC-DC converter for V2G
and G2V applications in electric vehicle battery charger," in
model predictive control method utilizes the discrete and IEEE Transportation Electrification Conference and Expo
inductive nature of the power converter and induction motor (ITEC), 2014, pp. 1-6.
load. In this control, the long prediction horizon, i.e, the [12] J. Beerten, J. Verveckken, and J. Driesen, "Predictive direct
second next (k+2)th predictive variables are predicted, and torque control for flux and torque ripple reduction," IEEE
Transactions on Industrial Electronics, vol. 57, pp. 404-412,
compared with the references to determine the quality 2010.
functions for all the possible eight switching vectors of the [13] M. Uddin, S. Mekhilef, M. Mubin, M. Rivera, and J. Rodriguez,
converter to overcome the inevitable control delay. The "Model Predictive Torque Ripple Reduction with Weighting
delay associated with the processors in the experiment has Factor Optimization Fed by an Indirect Matrix Converter,"
Electric Power Components and Systems, vol. 42, pp. 1-11,
no effect on control performance when delay time 2014.
compensation has been taken in consideration in the [14] M. Uddin, S. Mekhilef, M. Rivera, and J. Rodriguez,
predictive control algorithm which results in well tracking "Predictive Indirect Matrix Converter Fed Torque Ripple
of the reference variables at high speed, even at low speed Minimization with Weighting Factor Optimization," In Proc. of
International Power Electronics Conference (IPEC), pp. 3574-
regions of the induction motor. 3581, 2014.
[15] M. Rivera, C. Rojas, A. Wilson, J. Rodriguez, J. Espinoza, C.
Baier, and J. Muñoz, "Review of predictive control methods to
improve the input current of an indirect matrix converter," IET
REFERENCES Power Electronics, pp. 1-9, 2013.
[16] S. Vazquez, J. Leon, L. Franquelo, J. Rodriguez, H. Young, A.
Marquez, and P. Zanchetta, "Model Predictive Control: A
[1] Y.-S. Lai, J.-C. Lin, and J. J. Wang, "Direct torque control Review of Its Applications in Power Electronics," IEEE
induction motor drives with self-commissioning based on Industrial Electronics Magazine, vol. 8, pp. 16-31, 2014.
Taguchi methodology," IEEE Transaction on Power [17] J. Rodriguez, J. Kolar, J. Espinoza, M. Rivera, and C. Rojas,
Electronics, vol. 15, pp. 1065-1071, 2000. "Predictive torque and flux control of an induction machine fed
[2] K.-B. Lee and F. Blaabjerg, "Sensorless DTC-SVM for by an indirect matrix converter," in Proc. IEEE International
induction motor driven by a matrix converter using a parameter Conf. on Ind. Tech., 2010, pp. 1857-1863.
estimation strategy," IEEE Transactions on Industrial
Electronics, vol. 55, pp. 512-521, 2008.

548