electronics
Article
Performance Improvement of DTC-SVM of PMSM with
Compensation for the Dead Time Effect and Power Switch Loss
Based on Extended Kalman Filter
Doo-Il Son, Jun-Seo Han, Je-Suk Park, Hee-Sun Lim
Citation: Son, D.-I.; Han, J.-S.; Park,
J.-S.; Lim, H.-S.; Lee, G.-H.
Performance Improvement of
DTC-SVM of PMSM with
Compensation for the Dead Time
Effect and Power Switch Loss Based
on Extended Kalman Filter.
Electronics 2023, 12, 966. https://
doi.org/10.3390/electronics12040966
Academic Editors: Wei Hua and
Sheldon Williamson
Received: 12 January 2023
Revised: 4 February 2023
Accepted: 14 February 2023
Published: 15 February 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Electronics 2023, 12, 966. https://doi.org/10.3390/electronics12040966
and Geun-Ho Lee *
Electric Motor Control Laboratory, Graduate School of Automotive Engineering, Kookmin University,
Seoul 02707, Republic of Korea
* Correspondence: motor@kookmin.ac.kr; Tel.: +82-2-910-6409
Abstract: Two algorithms have been extensively studied for motor control: Field Oriented Control
(FOC) and Direct Torque Control (DTC). Both control algorithms use a Voltage Source Inverter (VSI) to
drive a Permanent Magnet Synchronous Motor (PMSM). To prevent short-arm short-circuit accidents
when driving PMSM using VSI, a dead time is used to turn off the TOP and BOTTOM switches of
each arm at the same time. However, this dead-time technique causes an unexpected pole voltage
to be applied to the PMSM on the VSI output voltage, causing distortion and resulting in control
nonlinearity. The disturbance voltage that causes nonlinearity is difficult to measure directly with the
sensor. Therefore, this paper analyzes the nonlinearity of the controller due to the distorted voltage
caused by the dead time during PMSM operation using the DTC algorithm and predicts the distorted
output voltage using the extended Kalman Filter (EKF) to improve control stability. As a result, The
algorithm proposed in this paper has verified the improvement of torque ripple and stator flux ripple
through experiments and simulations.
Keywords: PMSM; direct torque control; Kalman filter; dead-time; IGBT; inverter; FOC; DTC-SVM
1. Introduction
In the field of motor control, FOC (field-oriented control) and DTC (direct-torque
control) have been predominantly studied. The DTC algorithm has been in development
since the 1980s. The DTC algorithm was widely used in the field of induction motors
in their early days. As high efficiency and high density became important in the motor
industry, the DTC algorithm also began to be applied to the synchronous motor field.
The DTC algorithm is a control method that follows the torque and magnetic flux of the
motor using a hysteresis controller and a lookup table (LUT) for the torque command value
and stator flux command value [1–3]. However, the conventional DTC algorithm using a
hysteresis controller and LUT has a significant disadvantage in that torque ripple occurs
because the load angle between the rotor flux and the stator flux cannot be maintained
constantly during motor rotation. To overcome this, many control schemes have been
proposed. Among them, a DTC-SVM (Direct Torque Control Space Vector Modulation)
technique that is able to instantaneously keep the load angle constant according to the
position of the stator flux has been developed [4–9]. In this way, a DTC-SVM control
technique that places the stator flux at an accurate location has been proposed to reduce
torque ripple, but additionally, for accurate DTC-SVM control, the position and magnitude
of the stator flux must be accurately estimated. Most DTC-SVM algorithms use motor
parameters and current in the method of estimating the stator flux and position. Therefore,
the magnitude of the stator flux is estimated using the three-phase current value of the
motor measured using a current sensor mounted on the VSI (Voltage Source Inverter). If
the noise is included and measured, the exact location and value of the stator flux cannot
be estimated, causing torque ripple or poor control characteristics [10–13]. In this case, the
https://www.mdpi.com/journal/electronics
Electronics 2023, 12, 966 2 of 17

dead time used to prevent a short circuit when turning on/off the power switch element of
the VSI is a representative disturbance, and many methods to reduce this disturbance have
been proposed [14–20]. Most of these methods compensate for the dead time disturbance
that occurs according to the current direction when using the FOC algorithm. In addition,
regarding the corresponding compensation method, the disturbance voltage generated by
the diode of the power switch device, such as the MOSFET and IGBT of the inverter, was
not considered. In a later study, the diode voltage drop value of the power switch device
was also considered, but the corresponding value could not be measured in real-time [20].
Therefore, this paper analyzes the effect of dead-time disturbance voltage on DTC control
and uses an EKF (extended Kalman filter) to estimate and compensate for the dead-time
disturbance voltage value that is difficult to measure. The Kalman filter was first applied
to synchronous motors in 1983, and the algorithm has been continuously researched in
subsequent decades [21]. Continuous research has been conducted in various algorithm
fields of PMSMs (Permanent Magnet Synchronous Motors) using the Kalman filter. [22–25]
In addition, dead-time compensation methods using Kalman filters have been continuously
studied [26–30]. Most studies have focused on dead-time compensation based on the FOC
algorithm. In this paper analysis of the dead-time voltage distortion that occurs when the
DTC algorithm is used, and the paper proposes a method to compensate for it using an
extended Kalman Filter. The effectiveness of the algorithm is demonstrated by comparing
the performance of the proposed algorithm and the conventional algorithm. In addition,
the voltage compensation of the proposed algorithm has been proven to improve the
performance of the controller through experiments and simulations.

2. Direct Torque Control of PMSM

2.1. Mathematical Model of PMSM
The dynamic characteristics of the DTC algorithm are analyzed and the PMSM math-
ematical model projected to the rotor flux reference coordinate system for application
to control is expressed in Equation (1). In addition, the d- and q-axis stator flux linkage
equation is expressed in Equation (2), and the PMSM torque equation is expressed in
Equation (3).
d
      
Vd Rs + dt Ld − wr L q id 0
= d + (1)
Vq wr L d Rs + dt Lq iq wr φ f
      
φd L 0 id φ
= d + f
φq 0 Lq iq 0
(2)
φd0 = Ld id

Te = 23 P(φd iq − φq id )
(3)
= 23 P(φ f iq + ( Ld − Lq )id iq )
where Vd and Vq represent stator voltage values of the d- and q-axis in the rotor coordinate
system, respectively. id , iq , φd , φq , Ld , Lq , φ f , Rs represent the d- and q-axis stator current, the
d- and q-axis stator flux, the d- and q-axis stator inductance, rotor flux, and stator resistance,
respectively. If the above torque, obtained by Equation (3), is arranged in terms of the
torque load angle, it can be expressed in Equation (4)
3 φs 1
Te = P (φ Lq sin δ + φs ( Ld − Lq ) sin 2δ) (4)
2 Ld Lq f 2
Through the torque obtained in Equation (4), it can be seen that, if the stator flux
linkage φs is kept constant, the torque can be controlled through the torque load angle δ.
Since the electrical response is very short compared to the mechanical response time, the
load angle can be adjusted by instantaneously adjusting the position of the stator magnetic
flux using the stator voltage vector.
Electronics 2023, 12, 966 3 of 17

2.2. Conventional DTC of PMSM

In order to apply the PMSM mathematical model to the DTC control technique,
Equations (1) and (2) can be discretized with a fixed control period Ts and arranged in
Equation (5). In addition, if this is arranged based on the stator flux linkage standard, it can
be expressed as Equation (6). As the speed increases, the magnitude of the stator resistance
is relatively very small, so if ignored, it can be seen that the position of the stator flux at the
point of (k + 1) can be adjusted according to the d-axis voltage and q-axis voltage command
through the corresponding equation.
φd (k +1)−φd (k )
Vd (k + 1) = Ts − wr Lq Iq (k)
(5)
φq (k +1)−φq (k )
Vq (k + 1) = Ts + wr ( Ld Id (k) + φ f )
 
φd (k + 1) = Vd (k + 1) + wr Lq Iq (k + 1) Ts + φd (k )
h i (6)
φq (k + 1) = Vq (k + 1) − wr ( Ld Id (k ) + φ f ) Ts + φq (k)

In the case of the existing DTC algorithm, when the torque command value and stator
flux command value are input into the controller using these characteristics, the state values
of torque and stator flux are output through the hysteresis controller, and according to
the state values, the LUT (lookup table) outputs the specified voltage command. Table 1
displays the LUTs. In most existing DTC algorithms, the torque hysteresis controller is
composed of three levels and the stator flux hysteresis controller is composed of two
levels. The configuration of the conventional DTC controller and the stator flux trajectory
according to the voltage command vector of the conventional DTC controller are shown in
Figures 1 and 2. As can be seen in Figure 2, in the case of the conventional DTC controller,
when a certain torque is output, the stator flux is not located in an accurate position,
resulting in poor control dynamics.

Table 1. LUT (lookup table) used for conventional DTC control.

Stator Flux Section

Flux Torque
θs1 θs2 θs3 θs4 θs5 θs6
1 V2 V3 V4 V5 V6 V1
1 0 V7 V0 V7 V0 V7 V0
−1 V6 V1 V2 V3 V4 V5
1 V3 V4 V5 V6 V1 V2
Electronics 2023, 12, x FOR PEER REVIEW 4 of 19
−1 0 V0 V7 V0 V7 V0 V7
−1 V5 V6 V1 V2 V3 V4

Blockdiagram
Figure1.1.Block
Figure diagramof
ofConventional
ConventionalDTC
DTCcontrol
controltechnique.
technique.
 Electronics 2023, 12, 966 4 of 17

Figure 1. Block diagram of Conventional DTC control technique.

Figure 2.
Figure 2. Stator
Stator flux
flux trajectory
trajectory according
according to
to the
the voltage
voltage vector
vector command
command under
under conventional
conventional DTC
DTC
using LUT (lookup table).
using LUT (lookup table).

2.3.
TableDTC-SVM of PMSM
1. LUT (lookup table) used for conventional DTC control.
The DTC-SVM control method is used Stator toFlux
reduce the torque ripple caused by not
Section
Electronics 2023,
Flux locating
12, x FOR PEER REVIEW
Torque the stator flux in an accurate position in the conventional DTC algorithm. 5 ofThe
19
torque command value
θs1 θs2 flux command
and stator
θs3 θs4 input to θthe
value are
θ
s5 controllers6in the

1 way as in the
same V2 conventional V3 DTC algorithm.V4 V5
However, V6
the DTC-SVM V1
controller
outputs the torque error value as ∗
1 0 V 7 V 0the load angleV 7 command V 0 using theV
δ PI7 controller,V not
0 the
command of the voltage command is output, through which the stator flux value can be
hysteresis controller.
−1 at an accurate
output In addition,
V6 position through
V1 in the case of
V2 the
the current stator flux
motorVrotation
3 controller, the angle command
V4 and the current
speed V5
of the
1 voltage command
V is output, through which the stator flux
stator flux estimation value. It outputs the stator voltage vector command value andVangle
3 V 4 V 5 V 6 value
V 1 can be output
2 at
−1 an0accurate position through the current motor rotationVspeed and the V0current stator flux
command value toVSVM
0 using theV7 load angleVcommand
0 and7 voltage command angleVcom-
7
estimation value. It outputs the stator voltage vector command value and angle command
mand−1 output fromVthe 5 stator flux V6controller. V The
1 diagramVfor 2 the DTC-SVM V3 controlVblock
4
value to SVM using the load angle command and voltage command angle command output
and the stator flux position difference generated when using the conventional DTC algo-
from the stator flux controller. The diagram for the DTC-SVM control block and the stator
rithm and whenofusing
2.3. DTC-SVM PMSM the DTC-SVM algorithm are shown in Figures 3 and 4. Regarding
flux position difference generated when using the conventional DTC algorithm and when
the conventional DTC algorithm in Figure 4a, it can be seen thatripple
usingThe
the DTC-SVM
DTC-SVM control
algorithm method is used
are shown into reduce
Figures 3 the
andtorque  the stator
4. Regarding caused fluxbyvector
not lo-
the conventional
catingalgorithm
generated
DTC theusing
stator
influx
the LUTindoes
Figure an accurate
4a, itnot
canreach position
the that
be seen statorin
the the
flux flux OC
conventional
vector
stator vector DTC
where algorithm.
the
generated load The
angle
using the
→
torque
δLUT command
candoes
be kept value
constant.
not reach
and stator
This causes
the stator
flux command
an error
flux vector OCvalue
where
value
when are input
estimating
the load
to
angle δthe
the controller
canstator
be kept φs err
in* the
fluxconstant.
same way as in the conventional DTC algorithm. However,∗the DTC-SVM controller out-
, This causes
causing an error
torque value
ripple. when estimating
In Figure the stator
4b, in the case of theflux φs err, causing
DTV-SVM algorithm,torquetheripple.
state ofIn
puts
the sixthe
Figure 4b,torque
power the error
in switch ofvalue
casedevices as the
the DTV-SVM
can load angle
algorithm,
be changed command
within theonestate
cycleδ* to
of using
the the the
six power
position PI switch
controller, not
devices
stator flux at
the
 controller. In addition, in the case of the stator flux controller,
hysteresis →
the angle
acan
position OB , where
be changed withinthe oneload
cycleangle δ can
to position thebestator flux at ainposition
maintained a constant OB,state.
where the load
angle δ can be maintained in a constant state.

Blockdiagram
Figure3.3.Block
Figure diagramofofDTC-SVM
DTC-SVMcontrol.
control.
Electronics 2023, 12, 966 5 of 17

Figure 3. Block diagram of DTC-SVM control.

(a) (b)
Figure 4. Stator linkage flux vector trajectory according to voltage vector command under DTC-
Figure 4. Stator linkage flux vector trajectory according to voltage vector command under DTC-SVM
SVM control (a) Position of stator flux in the case of conventional DTC control (b) Position of stator
control (a) Position of stator flux in the case of conventional DTC control (b) Position of stator flux in
flux in the case of DTC-SVM control.
the case of DTC-SVM control.

3.3. Problems
Problems Caused
Causedby byDead-Time
Dead-Timeand andPower-Switch
Power-SwitchLosses
Lossesin
inDTC-SVM
DTC-SVM
Deadtime
Dead timeisisthe
thetime
timetaken
takentototurn
turnoffoffthe
thetop/bottom
top/bottom switch
switch to
to prevent
prevent burnout
burnoutduedue
to a short arm when turning the top/bottom switch of each leg on or off in
to a short arm when turning the top/bottom switch of each leg on or off in order to apply order to apply
aapole
polevoltage
voltageto toeach
eachphase
phaseof
ofthe
themotor
motorin inthe
theVSI.
VSI.The
Theoverall
overallconfiguration
configurationof ofthe
theVSI
VSI
isisshown
shownin inFigure
Figure5a,5a,and
andaasingle
singleleg
legisisshown
shownin inFigure
Figure5b.
5b.The
Thedead
deadtime
timeisisshown
showninin
Figure6a,b.
Figure 6a,b.
Turning the switch on or off during each phase applies the pole voltage output
command of the controller to the motor. In the dead time, unexpected pole voltage is
applied to the motor by conducting the diode of the power switch device according to the
direction of the current. This is shown in Figure 7. Additionally, when the diode conducts,
the loss of the forward voltage, which the diode itself has, is reflected and applied to the
motor. While this is relatively small in size and can be ignored in the high-speed operation
range, it is difficult to ignore in the low-speed range. In addition, the application of such
Electronics 2023, 12, x FORunexpected
PEER REVIEW disturbance voltage eventually causes a distortion of the phase current and 6 of 1
torque ripple in the DTC-SVM algorithm, which estimates stator flux using the current
value, thereby degrading performance.

(a) (b)
Figure 5. (a) VSI (voltage source inverter) configuration block diagram. (b) VSI each phase powe
Figure 5. (a) VSI (voltage source inverter) configuration block diagram. (b) VSI each phase power
switch configuration diagram.
switch configuration diagram.
 (a) (b)
Electronics 2023, 12, 966 6 of 17
Figure 5. (a) VSI (voltage source inverter) configuration block diagram. (b) VSI each phase power
switch configuration diagram.

(a) (b)
Electronics 2023, 12, x FOR PEER REVIEW 7 of
Figure 6. (a) ON/OFF signal of TOP/BOTTOM power switch element in an ideal case 19
without dead
Figure 6. (a) ON/OFF signal of TOP/BOTTOM power switch element in an ideal case without dead
time. (b) ON/OFF signal of TOP/BOTTOM power switch element with dead time.
time. (b) ON/OFF signal of TOP/BOTTOM power switch element with dead time.

Turning the switch on or off during each phase applies the pole voltage output com-
mand of the controller to the motor. In the dead time, unexpected pole voltage is applied
to the motor by conducting the diode of the power switch device according to the direc-
tion of the current. This is shown in Figure 7. Additionally, when the diode conducts, the
loss of the forward voltage, which the diode itself has, is reflected and applied to the mo-
tor. While this is relatively small in size and can be ignored in the high-speed operation
range, it is difficult to ignore in the low-speed range. In addition, the application of such
unexpected disturbance voltage eventually causes a distortion of the phase current and
torque ripple in the DTC-SVM algorithm, which estimates stator flux using the current
value, thereby degrading performance.

(a) (b)

Figure 7.Figure
(a) A situation
7. (a) A in
situation −V
which in / 2−Vvoltage
which
DC is applied
DC /2 voltage to the motor
is applied to thewhen the
motor current
when theflows
current flows
in the positive direction and is in a dead-time condition (b) A situation in which V /2 voltage is
in the positive direction and is in a dead-time condition (b) A situation in which V DC / 2 voltage
DC
applied to the motor when the current flows in the negative direction and is in a dead-time condition.
is applied to the motor when the current flows in the negative direction and is in a dead-time con-
dition. The disturbance voltage V applied to the motor by the dead time can be expressed
dead
as Equation (7). As can be seen from Figure 7, the unexpected disturbance voltage is
Theapplied
disturbance voltage
as the diode Vdead applied
is conducted at theto the time
dead motor by according
point the deadtotime can be of the
the direction
three-phase
expressed current
as Equation (7). applied
As can tobethe motor.
seen from Figure 7, the unexpected disturbance
voltage is applied as the diode is conducted at the dead time point according to the
Tdead
V
direction of the three-phase current dead = −
applied to (VDC ) ∗ sign(i phase ) (7)
T the motor.
sample

In addition, when the diode T

Vdead = − deadconducts in the forward direction during dead time,
(VDC ) * sign(i ) (7)
a diode voltage drop component Tsampleis added and phase
applied to the motor according to the
magnitude of the current. The magnitude of the voltage drop component of the diode and
In addition, when thenonlinearly
MOSFET changes diode conducts
accordingin the
toforward direction
the magnitude during
of the deadThis
current. time, a
is shown in
Figure drop
diode voltage 8a,b. component is added and applied to the motor according to the mag-
nitude of theIncurrent.
order toThe
know the exactof
magnitude magnitude
the voltage of drop
the disturbance
componentvoltage, the current
of the diode and value
MOSFET Idiode at thenonlinearly
changes point when according
the diode istoconducted
the magnitude must of
bethe
known exactly.
current. ThisHowever,
is shownininorder to
accurately
Figure 8a,b. measure the magnitude of the phase current of the motor, it must be measured at
the point when the voltage vector is applied as a zero vector. With these characteristics, the
complex device drive method that senses the current value at the switching frequency, i.e.,
the highest point or lowest point of the carrier wave, is adopted and used. This is a control
method that reflects the characteristics of the DSP (Digital Signal Processor) that schedules
tasks according to the CPU (Central Processing Unit) clock. Accurately measuring the
current at the point at which the above-mentioned diode is conducting means accurately
 dead

expressed as Equation (7). As can be seen from Figure 7, the unexpected disturbance
voltage is applied as the diode is conducted at the dead time point according to the
direction of the three-phase current applied to the motor.

Electronics 2023, 12, 966 Tdead 7 of 17

Vdead = − (VDC ) * sign(iphase ) (7)
Tsample

In addition,
reading the ON/OFFwhen the diode
status, whichconducts in the forward
is instantaneously direction
variable during
according to thedead time, a
pole voltage
diode voltage drop component is added and applied to the motor according
command. However, this measurement method is complicated and vulnerable to noise to the mag-
nitude of the current. The magnitude of the voltage drop
during sensing, so it is difficult to measure an accurate value. component of the diode and
MOSFET changes nonlinearly according to the magnitude of the current. This is shown in
Figure 8a,b. Vdead_ f = Vdead + Vdiode (8)

(a) (b)
Figure 8. (a) Turn-on characteristics of the MOSFET (b) Turn-on characteristics of the diode.
Figure 8. (a) Turn-on characteristics of the MOSFET (b) Turn-on characteristics of the diode.

In this
In order to know
paper, the exact magnitude
the disturbance of the disturbance
voltage generated during dead voltage,
time isthe current using
estimated value
Idiodeobserver.
an at the pointThewhen the diode
disturbance is conducted
voltage must be known
to be estimated can be exactly.
expressed However, in order
as Equation (8),
where Vdead ismeasure
to accurately the unexpected disturbance
the magnitude pole
of the voltage
phase applied
current to the
of the motor
motor, by each
it must phase
be meas-
current the point V
ured at direction, diode is
when thevoltage
the voltagevector
loss caused by theaspower
is applied a zeroswitch
vector.elements and charac-
With these diodes,
and Vdead_the
teristics, f iscomplex
the totaldevice
disturbance voltage during
drive method deadthe
that senses time. Vdead_value
current f can at
bethe
expressed
switchingas
Equation (9) through the coordinate transformation on the d-q-axis coordinate
frequency, i.e., the highest point or lowest point of the carrier wave, is adopted and used. plane used
for
Thiscontrol.
is a control method that reflects the characteristics of the DSP (Digital Signal
Vdead_ f ∗ sign(i a )
" #  
r
Vd_dead
r = T (θr ) Vdead_ f ∗ sign(ib ) 
Vq_dead
Vdead_ f ∗ sign(ic )
(9)
 Vdead_ f ∗ sign(i a )
 
2π 2π

2 cos(θr ) cos(θr − 3 ) cos(θr + 3 )  Vdead_ f ∗ sign(ib ) 
=3
− sin(θr ) − sin(θr − 2π 3 ) − sin( θr + 3 )
2π
Vdead_ f ∗ sign(ic )
r
The disturbance voltage values, Vd_dead r
and Vq_dead , generated during the dead time
and obtained through Equation (9) are shown in Figure 9 on the d-q-axis coordinate plane
through MATLAB R2022a simulation. In addition, the disturbance voltage applied to each
phase of the motor at the dead time is expressed in Figure 10b. The ideal gate voltage
and the gate voltage considering the dead time and the pole voltage output from VSI to
the motor, when the sign of the current output to the phase is negative and positive are
shown in Figure 10a. The disturbance voltage causes the distortion of the current waveform
applied to the actual three phases of the stator. This can be confirmed in Figure 11. In the
case of the DTC algorithm, as was mentioned in the previous section, the stator flux is
estimated using the measured three-phase currents. Therefore, the distortion of the three-
phase current also appears in the trajectory of the stator flux estimated in the DTC−SVM
algorithm. In addition, it can be predicted that the distortion of the estimated stator flux
trajectory eventually causes torque ripple during the DTC−SVM algorithm. This was
demonstrated through simulations and experiments.
 Electronics 2023, 12, x FOR PEER REVIEW 9 of 19
Electronics 2023,
Electronics 12,12,
2023, x FOR
966 PEER REVIEW 9 of819
of 17

Figure 9. Distorted voltage due to dead time appearing on the d − q coordinates waveform
Figure 9. 9.Distorted voltage
voltagedue
duetotodead
dead time
time appearing on the
thed− q coordinates
dq−coordinates r
waveform
V r r , Vqr_ dead , and the motor rotor angle waveform θr .
Figure Distorted waveform Vd_dead ,
rV r d _ dead
V q_dead , V
d _ dead
, and the motor rotor angle waveform θ r
, and the motor rotor angle waveform
q _ dead
. θ
r .

(a) (b)
(a) (b)
Figure
Figure 10.10.
(a)(a) Influence
Influence of of dead−time
dead −time ononVSIVSI output
output voltage,
voltage, (b)(b) Disturbance
Disturbance voltage
voltage generated
generated in in
Figure 10.phase
each (a) Influence of dead−time
of the motor on VSI
by the dead time.output voltage, (b) Disturbance voltage generated in
each
each phase
phase of the
of the motor
motor by by
thethe dead
dead time.
time.
 Electronics 2023, 12, x FOR PEER REVIEW 10 of 19

Electronics 2023,
Electronics 2023, 12,
12, 966
x FOR PEER REVIEW 109 of 19
of 17

Figure 11. Three−phase current waveform with harmonics as the disturbance voltage generated
during dead time.
11. Three
Figure 11. Three−phase
−phase current
current waveform
waveform with
with harmonics
harmonics as the disturbance
disturbance voltage
voltage generated
generated
4. Proposed
during dead
Disturbance Voltage Compensation Method
dead time.
time.
during
The voltage distortion that occurs during the dead time changes depending on the
4. Proposed
operating Disturbance
state of the motor. Voltage Compensation
In addition,
Voltage an additional
Compensation Method
circuit configuration is required to
Method
measure the current
The voltage and voltage
distortion of the during
that occurs inverter;the otherwise,
dead time it changes
is difficult to measure
depending ondi-the
rectly.
operating Therefore, to overcome these problems, this paper proposes
operating state of the motor. In addition, an additional circuit configuration is required to
state of the motor. In addition, an additional circuit a method
configuration isof compen-
required to
sating the
measure
measure theDTC-SVM
the current
currentand three-phase
and voltageofcontrol
voltage thethe
of voltage
inverter;
inverter; output
otherwise,
otherwise,by designing
it isitdifficult
is difficult a disturbance
to measure
to measure ob-
directly.
di-
server based
Therefore, on the PMSM
to overcome these equation
problems, of state and motor
this paper parameters.
proposes a methoda of compensating the
rectly. Therefore, to overcome these problems, this paper proposes method of compen-
DTC-SVM The proposed
three-phase disturbance
control observer
voltage uses
output the
by DC voltage
designing aand the three-phase
disturbance observer current
based
sating the DTC-SVM three-phase control voltage output by designing a disturbance ob-
configured
on the PMSM in the VSI without
equation of state theandneed to add
motor a hardware circuit. In the previous sections,
parameters.
server based on the PMSM equation of state and motor parameters.
it was The demonstrated that the disturbance voltage generated duringthe the dead time causes
The proposed
proposed disturbance
disturbance observer
observer uses uses thethe DCDC voltage
voltage and and the three-phase
three-phase current current
distortion of
configured in the
the voltage
VSI output
without the from
need the
to VSI.a It
add was alsocircuit.
hardware confirmed In the that the distorted
previous sections,
configured in the VSI without the need to add a hardware circuit. In the previous sections,
voltagedemonstrated
it also causes distortion in the three-phase current of the motor,the resultingtime in distor-
it was
was demonstrated that that the
the disturbance
disturbance voltagevoltage generated
generated duringduring the dead dead time causescauses
tion in the stator
distortion of the flux estimated
voltage output byfrom
the controller.
the VSI. The
It wasproposed
also disturbance
confirmed thatobserver
the uses
distorted
distortion of the voltage output from the VSI. It was also confirmed that the distorted
the command
voltage voltage output from the DTC-SVM controller and the distorted voltage and
voltage also alsocauses
causesdistortion
distortionininthe thethree-phase
three-phase current
current of ofthethe
motor,
motor, resulting
resultingin distortion
in distor-
current
in values
theinstator generated
flux flux
estimated by the
by the dead time situation to estimate the distorted voltages
tion the stator estimated bycontroller.
the controller.The Theproposed
proposed disturbance
disturbance observer
observer usesuses
the
Vdr_ dead andvoltage
command Vqr_ dead output
generatedfrominthe d − q −axiscontroller
theDTC-SVM coordinate system,
and which rotate
the distorted in syn-
voltage and
the command voltage output from the DTC-SVM controller and the distorted voltager and
current
chronization
current
valueswith
r values
generated by the
the rotor.
generated by The
dead time situation
the distortion
dead timevoltages
to estimate
situationVto
r the
anddistorted
estimate
d _ dead
r
Vthe voltages dVd_dead
on the voltages
distorted
q _ dead −q
andr V generated
r in the d-q-axis coordinate system, which rotate in synchronization
Vd _ deadcoordinate
−axis and Vq _ dead
q_dead
system estimated
generated through
in the dr − qthe disturbance
−axis coordinate observer
system,feed which the rotate
three-phase
in syn-
with the rotor. The distortion voltages Vd_dead and V r on the d-q-axis coordinate system
command voltage output to the controller, thereby q_dead
compensating
r for the r distortion volt-
chronization
estimated with the
through the disturbance
rotor. The distortion
observer feed voltages V _ dead and
the three-phase Vq _ dead voltage
command d −q
on the output
age generated during the dead time. The configuration ddiagram of the proposed algo-
to the controller, system
−axis thereby compensating forthethedisturbance
distortion voltage
observergenerated during the
rithmcoordinate
is shown in Figureestimated 12. through feed the three-phase
dead
command voltage output to the controller, thereby compensating for the distortion 12.
time. The configuration diagram of the proposed algorithm is shown in Figure volt-
age generated during the dead time. The configuration diagram of the proposed algo-
rithm is shown in Figure 12.

Figure 12. DTC-SVM

Figure 12. control block
DTC-SVM control block diagram
diagram applying
applyingthe
theproposed
proposedextended
extendedKalman
Kalmanfilter-based
filter-based
disturbance observer.
disturbance

In12.
Figure theDTC-SVM
case of the proposed
control blockdisturbance observer,
diagram applying the voltage
the proposed distortion
extended caused
Kalman by the
filter-based
dead time has nonlinearity;
disturbance observer. therefore, it was designed using the extended Kalman filter.
Electronics 2023, 12, 966 10 of 17

Equation (10) shows the distorted voltage value generated by the dead time added to the
d-q-axis model equation that rotates synchronously with the PMSM rotor.
 ∗  d
 " r #
Vd_dead
    
Vd Vd + Vd_comp Rs + dt Ld − wr L q id 0
= = + + (10)
Vq∗ Vq + Vq_comp wr L d d
Rs + dt Lq iq wr φ f r
Vq_dead

Vd , Vq , id , and iq represent the voltage and current values of the d-q-axis rotor coordi-
nate system, rotating in synchronization with the rotor. RS , Ld , Lq , wr , and φ f represent the
stator resistance, d-q-axis inductance values, rotor angular velocity, and rotor flux values,
respectively. Vd_comp and Vq_comp represent the voltage compensation value on the d-q-axis
coordinate system, and Vd_dead r r
and Vq_dead represent the distortion voltage on the d-q-axis
coordinate system caused by dead time. The voltage compensation values Vd_comp and
Vq_comp on the d-q-axis coordinate system are the same as the dead time distortion voltage
values. The disturbance observer using the EKF (extended Kalman filter) needs to be
modified by a state space model expressed as Equation (11).
x k = f ( x k −1 ) + w k −1
(11)
zk = Hxk−1 + νk−1
In order to modify in the form of the EKF equation of state, the state variable must
first be selected. In this paper, the DTC-SVM algorithm is performed, and the d-q-axis
r
stator flux and dead time distortion voltages φd , φq , Vd_dead r
, and Vq_dead are selected as
state variables to compensate for the distorted voltage generated during the dead time.
Substituting Equation (2) into Equation (10) to apply the selected state variable to the EKF
can be expressed as Equation (12).
 d 0  " Rs #  
− Ld wr φd0 1 0 Vd∗
      
φ
dt d = −1 0 Vd_dead 0
+ + + (12)
d
dt φq
−wr − RL s φq q
0 1 Vq∗ 0 −1 Vq_dead wr φ f

After that, if the state equation is organized into a system model and a measurement
model is used in EKF, it can be expressed as Equations (13) and (14). Assuming that the
distortion voltage caused by the dead time among the state variables does not change
rapidly, it can be expressed as Equation (13).
d 0
− RLds φd0
    
dt φd wr −1 0
d
dt φq  −wr − RLqs 0 −1  
    φ 
q
= +w
  
d r  V r
dt Vd_dead (13)
 
   0 0 0 0   d_dead 
d r r
Vq_dead
dt Vq_dead 0 0 0 0
= f (x) + w

φd0
 

φd0
   
1 0 0 0 φq 
= +ν
 
 Vr
φq 0 1 0 0  d_dead  (14)
r
Vq_dead
= Hx + ν
The observer design was carried out using the system model and measurement model
of the EKF organized in this way. Since the corresponding system model is nonlinear,
the nonlinear system model was linearized through a linearization method using the
Jacobian determinant. The matrix equation organized by the Jacobian matrix is shown in
Equation (15).
 ∂ f1 ∂ f1 ∂ f1 ∂ f1 
r r
∂φ0 ∂φq ∂Vd_dead ∂Vq_dead
 ∂ fd ∂ f2 ∂ f2 ∂ f2 

 2 r r
 ∂φd0 ∂φq ∂Vd_dead ∂Vq_dead 
A=
 ∂ f3 ∂ f3 ∂ f3 ∂ f3 
 (15)
 ∂φ0 ∂φq r
∂Vd_dead r
∂Vq_dead 
 d 
∂ f4 ∂ f4 ∂ f4 ∂ f4
r r
∂φd0 ∂φq ∂Vd_dead ∂Vq_dead
Electronics 2023, 12, 966 11 of 17

In addition, since the current system is a discretization model, the EKF system model
and the Jacobian matrix must be discretized. If the sampling time Ts is sufficiently small,
discretization is possible using the Euler integral. Therefore, the system model of this paper
was also discretized using the Euler integral and expressed as Equation (16).

φd0 (k) − RLs dTs φd0 (k − 1) φd0 (k − 1)

      
Ts wr − Ts 0
− RLs qTs
φq (k)   φ ( k − 1) φq (k − 1)
− Ts wr 0 − Ts 
     
q
= + +Q
     
r  V r r
Vd_dead (k) d_dead ( k − 1) Vd_dead ( k − 1)
 
   0 0 0 0     
r r r
Vq_dead (k) 0 0 0 0 Vq_dead ( k − 1) Vq_dead ( k − 1)

= ( Ts dtd f ( xk−1 ) + xk−1 ) + Q

(16)
φd0 (k )
 

φd0 (k)  φq (k)

   
1 0 0 0

= +R

 Vr
φq (k) 0 1 0 0  d_dead (k ) 
r
Vq_dead (k)

= Hxk + R
As an input factor to the EKF algorithm, the d-q-axis command voltage values, includ-
ing compensation voltages Vd_comp (k − 1) and Vq_comp (k − 1), were configured. It is also
assumed that the rotor magnetic flux of the motor is constant. wr is updated at each step of
the algorithm. In addition, matrices Q and R represent the covariance matrix of process
noise and the diagonal matrix of covariance of measurement noise, respectively, and were
selected as shown in Equation (17), assuming that the noise of the current system is very
small. Additionally, the system model matrix A used in the EKF is discretized and shown
in Equation (18), where I is a 4 × 4 identity matrix.
 
0.1 0 0 0
 0 0.1 0 0 
Q=  0

0 0.1 0 
0 0 0 0.1 (17)
 
0.1 0
R=
0 0.1

1 − RLs Ts
 
d
wr Ts − Ts 0
 −wr Ts 1 − RLs qTs 0 − Ts 
 
A(k) = ( I + Ts A) =   (18)
 0 0 1 0 
0 0 0 1
The finally proposed disturbance observer using the EKF is shown in
Equations (19)–(23) below.
− −
x (k) = f ( x (k − 1)) (19)
−
P(k) = A P̂(k − 1) A T + Q (20)
− − −1
K = P(k) H T ( H P(k) H T + R) (21)
− −
x̂ (k) = x (k) + K (z(k) − H x (k )) (22)
− −
P̂(k ) = P(k ) − KH P(k) (23)
where K is the gain of the Kalman filter and P is the covariance vector of the state variable.
−
Additionally, x̂ (k) and P̂(k ) denote the estimated state variable and covariance and x (k),
Electronics 2023, 12, 966 12 of 17

−
P(k) denote the predicted value of the state variable and covariance. The covariance vector
P(k) is composed of a 4 × 4 matrix, and the covariance vector matrix is a diagonal matrix.
The initial value of the covariance vector was designed to be 10. The dead time distortion
compensation voltage value on the d-q-axis coordinate system Vd_comp and Vq_comp , which
is the state variable of the disturbance voltage observer using the designed EKF, is added to
the three-phase output voltage command value of the DTC-SVM controller. The values of
r
Vd_dead r
, Vq_dead and Vd_comp , Vq_comp estimated through the designed observer are compared
Electronics 2023, 12, x FOR PEER REVIEW 14 of 19
in Figures 13 and 14, respectively. It can be seen from Figures 13 and 14 that the estimated
Electronics 2023, 12, x FOR PEER REVIEW
value of the designed disturbance observer follows the actual value well. Based on 14 of 19
this,
it can be seen that performance was improved by compensating the DTC-SVM controller
with the compensation value that was followed using simulation.

Figure 13. The actual disturbance voltage occurring on the d −axis during the dead time and the
waveform of the d −axis disturbance compensation voltage estimated through the disturbance ob-
Figure 13.
Figure The actual
13. The actualdisturbance
disturbancevoltage
voltageoccurring
occurringon the dd-axis
onthe −axisduring
during the
the dead
dead time
time and
and the
the
server.
waveform of the d-axis disturbance compensation voltage estimated through the disturbance
waveform of the d −axis disturbance compensation voltage estimated through the disturbance ob- observer.
server.

14. The
Figure 14. The actual
actualdisturbance
disturbancevoltage
voltageoccurring
occurringonon
thethe q −axis
q-axis during the dead
during the time
deadand
timethe q-axis
and the
q −axis disturbance
disturbance compensation voltage waveform estimated through the disturbance observer.
compensation voltage waveform estimated through the disturbance observer.
Figure 14. The actual disturbance voltage occurring on the q −axis during the dead time and the
q −axis disturbance compensation voltage waveform estimated through the disturbance observer.
5. Simulation and Experiment Result of Proposed Algorithm
To validate the performance of the proposed EKF-based disturbance observer, the
5. Simulation and Experiment Result of Proposed Algorithm
PMSM operating model in MATLAB/Simulink was used. The simulation was conducted
Toenvironments.
in two validate the performance of thethe
The first outputs proposed EKF-based
stator flux trajectory,disturbance observer,
torque ripple, the
and three-
PMSM operating
phase stator model
current in MATLAB/Simulink
waveforms waswithout
in the simulation used. Thethesimulation was conducted
voltage compensation of
in two environments. The first outputs the stator flux trajectory, torque ripple,
the disturbance observer. Then, in the simulation, when the voltage compensation and three-
of the
 Figure 15.

Table 2. Motor parameter values.

Parameter Value Unit

Electronics 2023, 12, 966 Motor Pole 8 13 of 17
DC-Voltage 48 V
Max. Torque 2.0 Nm
Phase Resistance
5. Simulation and Experiment Result of0.295 Proposed Algorithm Ω
dTo−axis the performance of the 0.22
Inductance
validate proposed EKF-based disturbance mH observer, the
q −axis Inductance 0.29
PMSM operating model in MATLAB/Simulink was used. The simulation mH was conducted
in two Magnetic
environments.
Flux The first outputs the stator flux trajectory, torque
0.0273 Wb ripple, and three-
phase stator current waveforms in the simulation without the voltage compensation of
the disturbance
First, Figure observer.
16 shows Then, in the
the stator simulation,
flux trajectorywhen the voltage
waveform compensation
when controlling of the
a DTC
disturbance
torque of 1.5 Nm.observer proposed
As shown in this
in Figure paper
16a, is be
it can applied, the the
seen that stator fluxflux
stator trajectory, torque
trajectory is
ripple, and phase stator current waveform are output. Afterward, the
distorted by the distorted voltage generated during the dead time rather than being cir-output results of each
of theThis
cular. twoconfirms
simulations thatwere
therecompared in order
are six ripples to assessinthe
per rotation theimprovement in the case
stator flux trajectory, as of
compensation
well as confirming using a disturbance
that observer.
the disturbance voltageThe is motor parameters
six times usedangular
the electrical in the experiments
velocity
ofand
thesimulations
motor through are shown in Tablesection.
the previous 2. The experimental
Figure 16b showssetup the
is shown
stator in Figure
flux 15.
trajectory
when the voltage is compensated using the proposed disturbance observer. Compared to
Table 2. Motor parameter values.
Figure 16a, it can be confirmed that the ripple seen in the stator flux trajectory is clearly
reduced. AsParameter
mentioned above, the estimatedValue stator flux trajectory has a great Unitinfluence on
the performance of the DTC-SVM controller. Subsequently,
Motor Pole 8
Figure 17a,b shows the torque
ripple when the voltage is not compensated and
DC-Voltage 48 when the voltage is compensated V by ap-
plying the Max.
proposed
Torquedisturbance observer. As2.0 shown in Figure 17, it can be confirmed
Nm that
the torque ripple
Phase is significantly reduced. Finally,
Resistance 0.295 in Figure 18a,b, the current Ω waveforms
d-axis
of the three Inductance
phases of the stator are compared0.22 with and without the compensation mH for the
disturbanceq-axisvoltage
Inductance
generated during the dead 0.29 time. As can be seen through mH the two
Magnetic Flux 0.0273 Wb
waveforms, the distortion of the stator current waveform is significantly reduced.

(a) (b)
Figure 15. (a) Experimental setup (b) Motors used during the test.
Figure 15. (a) Experimental setup (b) Motors used during the test.

First, Figure 16 shows the stator flux trajectory waveform when controlling a DTC
torque of 1.5 Nm. As shown in Figure 16a, it can be seen that the stator flux trajectory
is distorted by the distorted voltage generated during the dead time rather than being
circular. This confirms that there are six ripples per rotation in the stator flux trajectory, as
well as confirming that the disturbance voltage is six times the electrical angular velocity
of the motor through the previous section. Figure 16b shows the stator flux trajectory
when the voltage is compensated using the proposed disturbance observer. Compared to
Figure 16a, it can be confirmed that the ripple seen in the stator flux trajectory is clearly
reduced. As mentioned above, the estimated stator flux trajectory has a great influence
on the performance of the DTC-SVM controller. Subsequently, Figure 17a,b shows the
torque ripple when the voltage is not compensated and when the voltage is compensated by
applying the proposed disturbance observer. As shown in Figure 17, it can be confirmed that
the torque ripple is significantly reduced. Finally, in Figure 18a,b, the current waveforms
of the three phases of the stator are compared with and without the compensation for
 Electronics 2023, 12, 966 14 of 17

nics 2023, 12, x FOR PEER REVIEW 16 of 19

Electronics 2023, 12, x FOR PEER REVIEW
the disturbance
voltage generated during the dead time. As can be seen through the 16 of 19
two
Electronics 2023, 12, x FOR PEER REVIEW 16 of 19
waveforms, the distortion of the stator current waveform is significantly reduced.

(a) (b)
(a)
(a) (b)
(b)
Figure 16. Trajectory of stator flux during DTC−SVM torque control. (a) Scenario with no compen-
Figure 16.
Figure 16. Trajectory
16. Trajectory
Trajectory of
of stator
stator flux during
fluxdead
during DTC −SVM torque
DTC−SVM torque control.
control. (a)
(a) Scenario
Scenario with
with no
no compen-
compen-
sation for disturbance
Figure voltage generated
of during
stator flux time.
during (b) Scenario
DTC−SVM with
torque disturbance
control. voltage
(a) Scenario with no compen-
sation for disturbance voltage generated during dead time. (b) Scenario with disturbance voltage
generated during deadfor
sation time, which is compensated
disturbance usingduring
voltage generated a disturbance observer.
dead time. (b) Scenario with disturbance voltage
generated during dead time, which is compensated
compensated using
using aaa disturbance
disturbance observer.
observer.
generated during dead time, which is compensated using disturbance observer.

(a) (b)
(a)
(a) (b)
(b)
Figure 17. Torque ripple that occurs during DTC torque control. (a) Scenario with no compensation
Figure
for disturbance Figure 17.
voltage17. Torque ripple
generated ripple that occurs during DTC torque control. (a) Scenario
Scenariovoltage
with no
no compensation
compensation
Torque duringthat
Torque theoccurs
dead time (b)DTC
during Scenario when
torque the disturbance
control. (a) with
generated duringfor disturbance voltage generated during the dead time (b) Scenario when the disturbance voltage
forthe
for dead timevoltage
disturbance
disturbance is compensated
voltage generated
generated using a disturbance
during
during the
the dead observer.
dead time
time (b)
(b) Scenario
Scenario when
when the
the disturbance
disturbance voltage
voltage
generated
generated during the dead time is compensated using a disturbance observer.
generated during
during the
the dead
dead time
time is
is compensated using aa disturbance
compensated using disturbance observer.
observer.

(a) (b)
(a)
(a) (b)
(b)
Figure 18. Three−phase stator current waveform during DTC torque control. (a) Scenario with no
Figure
compensation for 18. Three−phase
Three−phase
disturbance
Figure 18. stator current
voltage generated
stator current
duringwaveform during
the deadduring
waveform DTC
time. (b)
DTC torquewhen
Scenario
torque control.
control. (a)
the(a)
dis-Scenario with
Scenario with no
no
Figure 18. Three
compensation for−disturbance
phase statorvoltage
currentgenerated
waveform during
during theDTC
deadtorque
time. control.
(b) (a) Scenario
Scenario when thewith
dis-
turbance voltage generated during
compensation the dead time
for disturbance is compensated
voltage using athe
generated during disturbance observer.
dead time. (b) Scenario when the dis-
no compensation
turbance for disturbance
voltage generated
generated duringvoltage generated
the dead
dead time is during the dead
is compensated
compensated usingtime. (b) Scenario when the
turbance voltage during the time using aa disturbance
disturbance observer.
observer.
disturbance voltage generated during the dead time is compensated using a disturbance observer.
In addition, the algorithm based on the extended Kalman Filter proposed in this pa-
per is compared with In addition,
In addition,
the the algorithm
the algorithm
conventional based on
based
compensation on the extended
the extended
algorithm. Kalman
Kalman
In the Filter proposed
Filter
conventional proposed
algo- in this
in this pa-
pa-
per is compared
per iswas
rithm, the voltage compared with the
with the
compensated conventional
byconventional compensation
considering compensation algorithm.
only the dead algorithm. In the conventional
In thetoconventional
time according the algo-
algo-
rithm,
rithm,
direction of the the applied
the
current voltage to
voltage was
was compensated
compensated
each bymotor.
by
phase of the considering
considering only the
only
Therefore, the dead
dead time
time according
conventional according to
to the
the
direction of the current applied to each phase of the motor. Therefore, the conventional
direction of the current applied to each phase of the motor. Therefore, the conventional
Electronics 2023, 12, 966 15 of 17

Electronics 2023, 12, x FOR PEER REVIEW In addition, the algorithm based on the extended Kalman Filter proposed in this17
paper
of 19
is compared with the conventional compensation algorithm. In the conventional algorithm,
the voltage was compensated by considering only the dead time according to the direction
of the current
algorithm applied
could to each phase
not compensate for of the motor.
torque ripple Therefore,
due to the the conventional
nonlinearity algorithm
of voltage drop
could not compensate for torque ripple due to the nonlinearity of voltage
that varies according to the amount of current through the power switch elementdrop that varies
and
according to the amount of current through the power switch element and diode. However,
diode. However, the proposed algorithm can compensate for the torque ripple by predict-
the proposed algorithm can compensate for the torque ripple by predicting the nonlinear
ing the nonlinear voltage drop component occurring in the diode and power switch ele-
voltage drop component occurring in the diode and power switch element. A comparison
ment. A comparison of the average torque ripple is shown in Figure 19a. In addition, The
of the average torque ripple is shown in Figure 19a. In addition, The average stator flux
average stator flux ripple when the conventional compensation algorithm is applied and
ripple when the conventional compensation algorithm is applied and the stator flux ripple
the stator flux ripple when the proposed algorithm is applied are shown in Figure 19b. As
when the proposed algorithm is applied are shown in Figure 19b. As shown in Figure 19, it
shown in Figure 19, it can be seen that the compensation algorithm proposed in this paper
can be seen that the compensation algorithm proposed in this paper is more stable in terms
is more stable in terms of torque ripple than the conventional algorithm when voltage
of torque ripple than the conventional algorithm when voltage drop components due to
drop components due to the diode and the power switch device occur. Also in terms of
the diode and the power switch device occur. Also in terms of stator flux, the proposed
stator flux, the proposed algorithm is more stable than the conventional algorithm. As a
algorithm is more stable than the conventional algorithm. As a result, it was confirmed
result, it was confirmed through the experimental results that the proposed algorithm is
through the experimental results that the proposed algorithm is more robust to external
more robustthan
disturbance to external disturbance
the conventional than the conventional algorithm.
algorithm.

(a) (b)
Figure 19. (a) Comparison of torque ripple for disturbance voltage generated by MOSFETs and di-
Figure 19. (a) Comparison of torque ripple for disturbance voltage generated by MOSFETs and
odes when applying the conventional algorithm and the proposed EKF-based control algorithm (b)
diodes when applying the conventional algorithm and the proposed EKF-based control algorithm
Comparison of stator flux ripple for disturbance voltage generated by MOSFETs and diodes when
(b) Comparison
applying of stator fluxalgorithm
the conventional ripple forand
disturbance voltage
the proposed generatedcontrol
EKF-based by MOSFETs and diodes when
algorithm.
applying the conventional algorithm and the proposed EKF-based control algorithm.
6. Conclusions
6. Conclusions
Inthis
In thispaper,
paper,an anextended
extendedKalman
Kalmanfilter-based
filter-baseddisturbance
disturbance voltage
voltage observer
observer is is pro-
pro-
posedto
posed to improve
improve the the performance
performance of ofstator
statorflux
fluxestimation
estimationand andtorque
torque ripple
ripple induced
induced by
the dead time. The performance of the proposed observer was
by the dead time. The performance of the proposed observer was demonstrated through demonstrated through ex-
periments with
experiments withMATLAB/Simulink.
MATLAB/Simulink. Through
Through the the
results of the
results experiment,
of the experiment, it was con-
it was
firmed that the distortion that exists in the trajectory of the estimated
confirmed that the distortion that exists in the trajectory of the estimated stator flux was stator flux was
clearly improved. It was also confirmed that the disturbance observer
clearly improved. It was also confirmed that the disturbance observer based on the ex- based on the ex-
tendedKalman
tended Kalmanfilter
filterestimated
estimatedthe thedisturbance
disturbancevoltage
voltagevery
verywell.
well.InInaddition,
addition,the thetorque
torque
ripple in the conventional algorithm increased by up to 3%
ripple in the conventional algorithm increased by up to 3% of the maximum torqueof the maximum torque as the
as
the disturbance voltage generated by the MOSFET and diode increased. In the case the
disturbance voltage generated by the MOSFET and diode increased. In the case of of
algorithm
the algorithm proposed
proposed in the paper,
in the thethe
paper, torque
torqueripple is maintained
ripple is maintained within 0.5%.
within It was
0.5%. also
It was
confirmed
also confirmedthat that
the magnitude
the magnitude of theoftorque ripple ripple
the torque when thewhen proposed algorithm
the proposed was ap-
algorithm
plied
was was improved
applied was improved by 20% bycompared
20% compared to thetotorque rippleripple
the torque whenwhenthe conventional
the conventional algo-
rithm waswas
algorithm applied.
applied.In Interms
termsof of
stator
statorflux
fluxripple,
ripple,when
whenthethe conventional algorithm was
conventional algorithm was
applied,the
applied, theripple
rippleofofthethestator
statorflux
fluxestimation
estimationvalue valueincreased
increasedby byup
uptoto15%
15%of ofthethestator
stator
fluxcommand
flux commanddue dueto tothethedisturbance
disturbancevoltage
voltagecaused
causedby bythe
thepower
powerswitch
switchelements
elementsand and
diodes. Conversely, when the proposed algorithm was applied, the
diodes. Conversely, when the proposed algorithm was applied, the ripple of the stator flux ripple of the stator
flux estimation value was maintained at 5% or less of the stator flux command. Through
the experimental results, the proposed algorithm is confirmed to be more stable than the
conventional algorithm. Finally, The algorithm proposed in this paper is robust to
Electronics 2023, 12, 966 16 of 17

estimation value was maintained at 5% or less of the stator flux command. Through the
experimental results, the proposed algorithm is confirmed to be more stable than the con-
ventional algorithm. Finally, The algorithm proposed in this paper is robust to disturbances
in terms of stator flux estimation and the torque ripple, and its performance improves when
performing PMSM control based on the DTC-SVM algorithm without using additional
hardware circuits or complex measurement methods.

Author Contributions: Conceptualization, D.-I.S.; methodology, D.-I.S.; software, D.-I.S.; validation,

D.-I.S.; writing—original draft preparation, D.-I.S. and J.-S.H.; writing—review and editing, D.-I.S.,
J.-S.H., H.-S.L. and J.-S.P.; supervision, G.-H.L. All authors have read and agreed to the published
version of the manuscript.
Funding: This research was supported by the Korea Evaluation Institute of Industrial Technol-
ogy (KEIT, Korea) grant funded by the Government of Korea (MOTIE, Korea) (No. 1055001018,
Development of 50~150 kW electric drive system for construction machinery).
Data Availability Statement: Data sharing is not applicable to this article.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Takahashi, I.; Noguchi, T. A New Quick-Response and High-Efficiency Control Strategy of an Induction Motor. IEEE Trans. Ind.
Appl. 1986, IA-22, 820–827. [CrossRef]
2. Depenbrock, M. Direct self-control (DSC) of inverter-fed induction machine. IEEE Trans. Power Electron. 1988, 3, 420–429.
[CrossRef]
3. Kang, J.-K.; Chung, D.-W.; Sul, S.K. Direct torque control of induction machine with variable amplitude control of flux and torque
hysteresis bands. In Proceedings of the IEMDC’99—IEEE International Electric Machines and Drives Conference, Proceedings
(Cat. No.99EX272), Seattle, WA, USA, 9–12 May 1999; pp. 640–642.
4. Swierczynski, D.; Kazmierkowski, M.P. Direct torque control of permanent magnet synchronous motor (PMSM) using space
vector modulation (DTC-SVM)-simulation and experimental results. In Proceedings of the IECON 02—IEEE 2002 28th Annual
Conference of the Industrial Electronics Society, Seville, Spain, 5–8 November 2002; Volume 1, pp. 751–755.
5. Hiba, H.; Ali, H.; Othmen, H. DTC-SVM control for three phase induction motors. In Proceedings of the 2013 International
Conference on Electrical Engineering and Software Applications, Hammamet, Tunisia, 21–23 March 2013; pp. 1–7.
6. Swierczynski, D.; Kazmierkowski, M.P.; Blaabjerg, F. DSP based direct torque control of permanent magnet synchronous motor
(PMSM) using space vector modulation (DTC-SVM). In Proceedings of the 2002 IEEE International Symposium on, L’Ayuila,
Italy, 8–11 July 2002; Volume 3, pp. 723–727.
7. Zelechowski, M.; Kazmierkowski, M.P.; Blaabjerg, F. Controller design for direct torque controlled space vector modulated
(DTC-SVM) induction motor drives. In Proceedings of the ISIE 2005—IEEE International Symposium on Industrial Electronics,
2005, Dubrovnik, Croatia, 20–23 June 2005; Volume 3, pp. 951–956.
8. Sadat, A.R.; Ahmadian, S.; Vosoughi, N. A novel torque ripple reduction of switched reluctance motor based on DTC-SVM
method. In Proceedings of the 2018 IEEE Texas Power and Energy Conference (TPEC), Memorial Student Center, Texas A & M
University, College Station, TA, USA, 8–9 February 2018; pp. 1–6.
9. Shinohara, A.; Yamamoto, K. Estimation Error Analysis of Stator Flux Observer for DTC-Based PMSM Drives. In Proceedings
of the 2018 International Power Electronics Conference (IPEC-Niigata 2018-ECCE Asia), Niigata, Japan, 20–24 May 2018;
pp. 1308–1314.
10. Shinohara, A.; Inoue, Y.; Morimoto, S.; Sanada, M. Comparison of stator flux linkage estimators for PWM-based direct torque
controlled PMSM drives. In Proceedings of the 2015 IEEE 11th International Conference on Power Electronics and Drive Systems,
Sydney, Australia, 9–12 June 2015; pp. 1035–1040.
11. Rahman, M.F.; Zhong, L. Problems of stator flux oriented torque controllers for the interior permanent magnet motor.
In Proceedings of the IPEMC 2000—Third International Power Electronics and Motion Control Conference (IEEE Cat.
No.00EX435), Beijing, China, 15–18 August 2000; Volume 1, pp. 342–345.
12. Kadjoudj, M.; Golea, N.; Benbouzid, M.E. Problems of Stator Flux Estimation in DTC of PMSM Drives. J. Electr. Eng. Technol.
2007, 2, 468–477. [CrossRef]
13. Kim, S.Y.; Lee, W.; Rho, M.S.; Park, S.Y. Effective Dead-Time Compensation Using a Simple Vectorial Disturbance Estimator in
PMSM Drives. IEEE Trans. Ind. Electron. 2010, 57, 1609–1614.
14. Urasaki, N.; Senjyu, T.; Uezato, K.; Funabashi, T. Adaptive Dead-Time Compensation Strategy for Permanent Magnet Synchronous
Motor Drive. IEEE Trans. Energy Convers. 2007, 22, 271–280. [CrossRef]
15. Urasaki, N.; Senjyu, T.; Uezato, K.; Funabashi, T. An adaptive dead-time compensation strategy for voltage source inverter fed
motor drives. IEEE Trans. Power Electron. 2005, 20, 1150–1160. [CrossRef]
Electronics 2023, 12, 966 17 of 17

16. Urasaki, N.; Senjyu, T.; Uezato, K.; Funabashi, T. A dead-time compensation strategy for permanent magnet synchronous motor
drive suppressing current distortion. In Proceedings of the IECON’03—29th Annual Conference of the IEEE Industrial Electronics
Society (IEEE Cat. No.03CH37468), Roanoke, VA, USA, 2–6 November 2003; Volume 2, pp. 1255–1260.
17. Chen, H.; Yao, Z.; Liu, Y.; Lin, J.; Wang, P.; Gao, J.; Zhu, S.; Zhou, R. PMSM Adaptive Sliding Mode Controller Based on Improved
Linear Dead Time Compensation. Actuators 2022, 11, 267. [CrossRef]
18. Urasaki, N.; Senjyu, T.; Uezato, K.; Funabashi, T. On-line dead-time compensation method for permanent magnet synchronous
motor drive. In Proceedings of the 2002 IEEE International Conference on Industrial Technology (IEEE ICIT’02), Bangkok,
Thailand, 11–14 December 2002; Volume 1, pp. 268–273.
19. Kim, H.; Moon, H.; Youn, M. A Novel Dead-Time Compensation Method using Disturbance Observer. J. Power Electron. 2002, 2,
55–66.
20. Sayeef, S.; Rahman, M.F. Improved Flux and Torque Estimators of a Direct Torque Controlled Interior PM Machine with
Compensations for Dead-time Effects and Forward Voltage Drops. J. Power Electron. 2009, 9, 438–446.
21. Borrero, A.J. Assessing Kalman Filter in the Identification of Synchronous Machine Stability Parameters. Master’s Thesis, Ohio
University, Athens, OH, USA, 1983.
22. Bolognani, S.; Tubiana, L.; Zigliotto, M. Extended Kalman filter tuning in sensorless PMSM drives. IEEE Trans. Ind. Appl. 2003, 39,
1741–1747. [CrossRef]
23. Quang, N.K.; Hieu, N.T.; Ha, Q.P. FPGA-Based Sensorless PMSM Speed Control Using Reduced-Order Extended Kalman Filters.
IEEE Trans. Ind. Electron. 2014, 61, 6574–6582. [CrossRef]
24. Zedong, Z.; Yongdong, L.; Fadel, M.; Xi, X. A Rotor Speed and Load Torque Observer for PMSM Based on Extended Kalman
Filter. In Proceedings of the 2006 IEEE International Conference on Industrial Technology, Mumbai, India, 15–17 December 2006;
pp. 233–238.
25. Li, X.; Kennel, R. General Formulation of Kalman-Filter-Based Online Parameter Identification Methods for VSI-Fed PMSM. IEEE
Trans. Ind. Electron. 2021, 68, 2856–2864. [CrossRef]
26. Zitzelsberger, J.; Hofmann, W. Space Vector Modulation with Current Based Dead Time Compensation using Kalman-Filter.
In Proceedings of the IECON 2006—32nd Annual Conference on IEEE Industrial Electronics, Paris, France, 7–10 November 2006;
pp. 1533–1538.
27. Buchta, L.; Bartik, O. Dead-Time Compensation Strategies Based on Kalman Filter Algorithm for PMSM Drives. In Proceedings
of the IECON 2019—45th Annual Conference of the IEEE Industrial Electronics Society, Lisbon, Portugal, 14–17 October 2019; pp.
986–991.
28. Buchta, L.; Otava, L. Adaptive compensation of inverter non-linearities based on the Kalman filter. In Proceedings of the IECON
2016—42nd Annual Conference of the IEEE Industrial Electronics Society, Florence, Italy, 23–26 October 2016; pp. 4301–4306.
29. Bialkowski, W. Application of Kalman filters to the regulation of dead time processes. IEEE Trans. Autom. Control. 1983, 28,
400–406. [CrossRef]
30. Diallo, D.; Arias, A.; Cathelin, J. An inverter dead-time feedforward compensation scheme for PMSM sensorless drive operation. In
Proceedings of the 2014 First International Conference on Green Energy ICGE 2014, Sfax, Tunisia, 25–27 March 2014; pp. 296–301.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.