Model Predictive Current Control for Three-Level

Inverter-Fed PMSM Drives Based on Variable

2023 IEEE International Conference on Predictive Control of Electrical Drives and Power Electronics (PRECEDE) | 979-8-3503-9686-7/23/$31.00 ©2023 IEEE | DOI: 10.1109/PRECEDE57319.2023.10174435

Sequence Space Vector Modulation

Yubin Wang, Yongchang Zhang, Wenyue Zheng
School of Electrical and Electronic Engineering
North China Electric Power University
Beijing, China
[email protected]

Abstract—In this paper, a variable sequence space-vector- states of the converter, which is easy to implement online.
modulation-based model predictive current control (VSSVM- However, it applies only one voltage vector over the whole
MPCC) is proposed for permanent magnet synchronous motor control period, so the steady-state performance remains poor
(PMSM) supplied by a neutral-point clamped three-level voltage
source inverter (NPC-3LVSI). Unlike the conventional space- and the switching frequency is not constant.
vector-modulation-based model predictive current control (SVM- To improve the steady-state performance, some schemes
MPCC), which uses conventional switching sequence to synthesize have been proposed, which apply multiple vectors during
voltage reference vector, the proposed VSSVM-MPCC exploits one control period [3], [5] or extend control sets [6], [7].
novel switching sequences to reduce total harmonic distortion
Though these methods presents better performance, the error
(THD) in the fundamental line current. After obtaining the
voltage reference vector, it is synthesized by the optimal switching of reference values and controlled variables still exists. To ad-
sequence based on the principle of minimizing current harmonics. dress this problem, the space-vector-modulation-based model
Moreover, the fluctuation of neutral point voltage (NPV) is predictive current control (SVM-MPCC) applies three adjacent
addressed by adjusting slightly the selected switching sequence. voltage vectors to synthesize the voltage reference vector, and
Compared with conventional SVM-MPCC, the proposed method
meanwhile the switching frequency remains constant [8].
is studied by theoretical analysis and simulation over the whole
range of modulation. The results confirm that the proposed The space vector modulation (SVM) plays a critical role
method reveals better steady-state performance, reducing THD in SVM-MPCC. The conventional SVM generates switching
by more than 25% at high modulation indices at a given average pulses by fixed switching sequence, leading to high current
switching frequency. harmonics at high modulation indices. It is shown in the
Index Terms—Model predictive control, space vector modula-
tion, permanent magnet synchronous motor, current harmonics. existing research that using different switching sequences is
a viable scheme to improve the steady-state performance for a
two-level inverter [9]–[11]. For a three level inverter, there are
I. I NTRODUCTION some relevant schemes [12]–[14]. Ref. [12] introduces five
In recent years, permanent magnet synchronous motor switching sequences, two of which can reduce the current
(PMSM) has been widely applied in the industry applica- ripple at low and high modulation indices respectively. Ref.
tions due to simple structure, high torque density and strong [13] proposes a hybrid PWM methods, involving the use
overload capacity. There are two main control strategies for of five sequences in five different spatial regions. However,
electric drives, vector control (VC) and direct torque control this method causes additional switching losses when some
(DTC) [1]. VC relies on a cascade configuration of propor- sequences are switched. Additionally, these methods are only
tional integral (PI) controllers, and DTC selects the voltage evaluated with constant volts per hertz (V /f ) control.
vector through predefined switching tables and hysteresis loop In this paper, a variable sequence space-vector-modulation-
comparators. based model predictive current control (VSSVM-MPCC) is
As a high performance closed-loop control strategy, model developed for the NPC-3LVSI fed PMSM drive system to
predictive control (MPC) has received extensive attention and solve the above issues. The main advantages of the proposed
research in the field of power electronics and motor control method are summarized as follows: 1) The proposed VSSVM-
in recent years, which has the advantages of simple principle, MPCC selects the optimal switching sequence among three
fast dynamic response, multivariable control and easy handling switching sequences based on the principle of minimizing
of nonlinear constraints [2]. Compared with VC, MPC does current ripple, presenting the superior overall performance; 2)
not need complex parameter tuning. Compared with DTC, The average switching frequency is constant; 3) The principle
MPC optimizes the selection of the best voltage vector by is simple and universal, which makes it easy to implement.
predicting the motor state. Therefore, MPC is regarded as a Simulation results have validated the effectiveness of the
new generation of AC motor control theory after VC and DTC. proposed method.
Depending on the controlled variables, MPC can be divided This paper is organized as follows. Section II introduces the
into model predictive current control (MPCC) [3] and model mathematical model of NPC-3LVSI fed PMSM drive system.
predictive torque control (MPTC) [4]. Conventional MPC Section III presents conventional SVM-MPCC. Section IV
directly combines the control target and the discrete switching analyzes the current ripple of three switching sequences and

979-8-3503-9686-7/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on October 27,2023 at 18:43:17 UTC from IEEE Xplore. Restrictions apply.
 TABLE I u5 u4 u3
S WITCHING STATES AND VOLTAGE LEVEL OF A NPC-3LVSI 020 120 220

Sx Sx1 Sx2 ux
u6 u18 u15
121 221 u2
2 1 1 Udc /2 021 210
1 0 1 0 010 110
0 0 0 −Udc /2 u17 u25 u16
u19 u13
000
u7 022
122
u26 111 100
000
200 u1
011 000
211
u20
222
u27 u u14
P u22
Sa1 Sb1 Sc1 23
+ 012
112 212
201
C1 U u8 001 101 u12
- c1 u21 u24
Sa2 Sb2 Sc2

Udc O PMSM 002 102 202

u9 u10 u11
Sa3 Sb3 Sc3
C2 +
U
- c2 ZVV P-Type N-Type MVV LVV
Sa4 Sb4 Sc4 SVV SVV
N

Fig. 2. Voltage vectors and switching positions of a NPC-3LVSI.

Fig. 1. Circuit topology of a NPC-3LVSI fed PMSM drive.

In the stationary αβ reference frame, the mathematical

introduces the principle of VSSVM-MPCC. Section V shows model of a surface-mounted PMSM (SPMSM) can be ex-
the simulation results. Finally, Section VI gives a conclusion. pressed as [15]:

II. M ATHEMATICAL M ODEL OF T HE S YSTEM

= Rs is + dψ

us dt
s
(2)
ψs = Lis + ψ f
The circuit topology of a NPC-3LVSI fed PMSM drive is
shown in Fig. 1, where Udc is the total DC-link voltage. Each
where ψ f = ψf ejθr ; us , is , ψ s , Rs , L, ψf and θr represent
phase can output three different levels of voltage with respect
stator voltage vector, stator current vector, stator flux vector,
to neutral point O, i.e., Udc /2, 0, −Udc /2.
stator resistance, stator inductance, permanent magnet flux
The switching position in phase x (x ∈ {a, b, c}) is defined linkage and electrical rotor position, respectively.
as Sx (x ∈ {a, b, c}). The two switching states Sx1 and Sx2 According to (2), the equation of state with the stator current
on the upper side of phase x and the two switching states Sx3 as the state variable can be expressed as:
and Sx4 on the lower side are complementary, i.e., Sx1 and
Sx3 , Sx2 and Sx4 operate in opposite states, and the switching dis us − Rs is − E
states Sx1 and Sx4 cannot be on at the same time. In practice, = (3)
dt L
only the two switching signals on the upper side of each phase
are generated directly by the controller, while the other two where E = jωr ψ f denotes the back electromagnetic force;
switching signals are generated using logic inversion and dead ωr is the electrical rotor speed.
time generator circuits. The switching table for each phase is The neutral point voltage (NPV) fluctuation is defined as:
shown in Table I.
Since each phase can output three different voltages, there ∆Uc = Uc1 − Uc2 (4)
are 33 = 27 voltage vectors, as shown in Fig. 2. They can
be divided into four main categories, namely large vectors, where Uc1 and Uc2 represent the upper and lower capacitor
medium vectors, small vectors and zero vectors. The small voltages.
vectors appear in pairs, representing the same voltage vector, Based on the circuit topology shown in Fig. 1, the derivative
so the pairs of small vectors are called redundant small vectors. of ∆Uc can be expressed as:
The redundant small vectors are divided into P-type small vec-
tors (connected to the positive bus) and N-type small vectors d∆Uc 1
(connected to the negative bus), having the opposite effects = (Ha ia + Hb ib + Hc ic ) (5)
dt C
on NPV fluctuation. The voltage vector can be expressed in
terms of Udc and Sx as: where C = C1 = C2 is the capacitor value; Hx (x ∈ {a, b, c})
2

is defined as:
us = 3 ua + ub a + uc a2

(1)
= 2
3 Sa + Sb a + Sc a2 U2dc 
1 if Sx = 1
Hx = (6)
j 23 π 2 0 otherwise
where a = e represents phase displacement of 3π among
a, b, c phases.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on October 27,2023 at 18:43:17 UTC from IEEE Xplore. Restrictions apply.
 U dc
ref
rref + iq Voltage 3
PI dq isref reference usref Pulses 2
− vector
SVM
r idref = 0  calculation 3L-NPC u2
VSI
ref
u
s
4 uˆ sref
isk +1   u1
u13,14
ia 1
One step delay
isk 
compensation
ib
abc
5
d 6
PMSM
dt e

Fig. 3. Control diagram of the conventional SVM-MPCC. Fig. 4. Simplification of a three-level space-vector diagram.

  k
III. P RINCIPLE OF C ONVENTIONAL SVM-MPCC where iref
s = iref
d + ji ref
q ej (θr +2ωr Tsc ) denotes the
current reference at (k + 2)th sampling instant; E k+1 =
The overall control diagram of the conventional SVM-
E k ejωr Tsc .
MPCC is shown in Fig. 3, which mainly consists of two
control loops. The outer PI controller is used to generate q-axis After obtaining the voltage reference vector uref
s , its synthe-
current reference value iref
q for speed regulation and the d-axis sis is achieved by SVM, which is based on the simplification
ref
current reference value id is set as zero. The inner controller of space-vector diagram [16]. To implement the simplification,
includes current prediction and vector synthesis, which forces the space-vector diagram of three-level inverter is considered
the actual current to track the reference value calculated by to be composed of six small hexagons that are the space-vector
the outer speed controller. The details will be explained in the diagrams of two-level inverter, as shown in Fig. 4. Taking the
following text. first hexagon as an example, when uref is located in the first
s
By discretizing equation (3), the stator current at the next sector (shaded part), the equivalent voltage reference vector
sampling instant can be predicted. To ensure the accuracy of ûref corresponding to the two-level inverter is obtained by
s
the prediction, the second-order Eulerian discretization method subtracting the center voltage vector of this hexagon u13,14
is used in this paper, and the discretized stator current state from uref
s . The duration time of three nearest voltage vectors
equation is: can be calculated as:

 k+1 Tsc (uk k k
s −Rs is −E )
isp = iks + L ûref
s sin 60° − β


k+1 k (7) T1 = Tsc (9a)
 ik+1 = ik+1 + 0.5Tsc (−Rs )(isp −is ) Udc /3 sin (60° )
s sp L

where iks and E k represent stator current and back electro- ûref
magnetic force at kth sampling instant; ik+1 is the predicted
s sin (β)
sp T2 = Tsc (9b)
k+1
correction value of stator current; is is the predicted value of Udc /3 sin (60° )
stator current at (k + 1)th sampling instant; Tsc is the control
period, i.e., sampling period; uks is the voltage vector applied TZ = Tsc − (T1 + T2 ) (9c)
during kth control period. where T1 , T2 and TZ represent the duration time of u1 , u2
In practical digital implementation, there exists one-step and u13,14 , respectively. The duration time TZ is divided into
digital delay, i.e., the control actions calculated during kth two equal parts for u13 and u14 . During a control period,
control period are updated at the beginning of (k + 1)th the switching sequence is conventionally u13 (100)-u1 (200)-
control period. To compensate for the negative effect of the u2 (210)-u14 (211) or its time-reversed version u14 (211)-
delay, the current value at (k + 1)th sampling instant is first u2 (210)-u1 (200)-u13 (100).
predicted according to (7), and then the current value at
(k + 2)th sampling instant is predicted by shifting (7) one-
IV. P ROPOSED VSSVM-MPCC
step ahead and thus not repeated here.
Since the mechanical time constant is much larger than the A. Analysis of Current Ripple about Novel Switching Se-
electromagnetic time constant, the electrical rotor speed ωr quences
can be assumed to be constant over several control periods.
Assuming that the stator current ik+2 reaches the current In conventional SVM-MPCC, the switching sequence be-
s
reference iref , the voltage reference vector uref applied gins from a N-type (P-type) small vector and ends with a P-
s s
during (k + 1)th control period can be expressed as: type (N-type) small vector during a control period, leading
to high current harmonics at high modulation indices. To
  improve steady-state performance over the whole range of
L2 iref
s − ik+1
s modulation, the novel switching sequences are investigated,
uref
s = + Rs ik+1
s + E k+1 (8) one of which applies only the N-type small vector, the other
Tsc (L − 0.5Tsc Rs )

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on October 27,2023 at 18:43:17 UTC from IEEE Xplore. Restrictions apply.
 TABLE II
T HE OPTIMAL SEQUENCE SELECTION
2 2
1 1
Modulation index M The optimal sequence
1 1
0 0 M < 0.89 Conventional sequence
1 0.89 ≤ M < 0.99 Clamping sequence
0 0 M ≥ 0.99 Double-switching clamping sequence
TZ TZ TZ T1 T2
T1 T2
2 2
Tsc Tsc

(a) (b)

2
1

1
0

0 T1 T
TZ T2 1
2 2
Tsc
(c)
Fig. 5. The switching states Sx of (a) conventional sequence, (b) clamping
sequence, (c) double-switching clamping sequence.

of which applies only the N-type small vector and uses one of
the other two vectors twice at the same time. The mentioned
switching sequences above are termed conventional sequence,
clamping sequence and double-switching clamping sequence, Fig. 6. Spatial division about the optimal sequence presenting the lowest
respectively. For the voltage reference vector uref
s shown in current ripple rms according to different M and α.
Fig. 4, the switching states Sx (x ∈ {a, b, c}) corresponding
to three switching sequences are presented in Fig. 5.
Owing to one phase clamped, the average switching fre- 2 4
√ 2 3
quency of clamping sequence is two thirds of the other two
2
IhIII = m2 12m cos (β)+121443√sin(β)m cos (β)
2 2
3 sin(β)m2 cos(β)
sequences. To remain the same average switching frequency, +m2 +3m cos (β)+15
√ 144
2
the control period needs to be multiplied by two thirds, namely −12 3m cos3 (β)+4 sin(β)m cos2 (β)
2
+m2 +21m
√ 144
3 Tsc , when applying clamping sequence. +m2 −36 3m cos(β)−58 sin(β)m+48
The modulation index M of uref s and the equivalent mod- 144
(14)
ulation index m of ûref
s can be calculated as: where IhI , IhII and IhIII denote the current ripple rms in a
√ control period of conventional sequence, clamping sequence
3 uref
s
M= (10) and double-switching clamping sequence, respectively [9].
Udc According to the relationship of uref s and ûref
s , the equiv-
√ alent modulation m and the phase angle β can be expressed
2 3 ûref
s with M and α. Thus IhI , IhII and IhIII can be also expressed
m= (11) with M and α. Fig. 6 shows spatial division about the optimal
Udc
sequence presenting the lowest current ripple rms according
The rms value of the total stator current ripple correspond- to different M and α. Due to symmetry of sectors, the rms
ing to three sequences in a control period can be obtained value of the total stator current ripple corresponding to three
as (12), (13) and (14), which are normalized with respect to sequences over a fundamental period can be calculated as (15),
U√
dc Tsc
2 3 L
. only related to M .
R π3 2
2 4
√ 2 3 IhI,II,III dα
2
IhI = m2 −24m cos (β)+24144 √
3 sin(β)m cos (β)
Ihf I,II,III = 0
2
π (15)
2 2
12m cos (β)−24 3 sin(β)m2 cos(β) 3
+m2 144 2
2 (12) Fig. 7 shows a comparison of current ripple rms over a
+m2 21m√+16 sin(β)m
144
cos (β)
fundamental period pertaining to three sequences. It shows that
+m2 −12 3m cos(β)−16144
sin(β)m+12
clamping sequence and double-switching clamping sequence
have lower current ripple rms than conventional sequence at
√
2
2 4 2
4 2 12m cos (β)+12 3 sin(β)m cos (β)
3 high modulation indices.
IhII = 9 m 36 √
2 2 2 By selecting the optimal sequence on the principle of
+ 94 m2 −15m cos √(β)−336 3 sin(β)m cos(β) minimizing current ripple according to M and α or only M ,
2 3
(β)+4 sin(β)m cos2 (β)
+ 49 m2 +12m −12 3m cos 36 the steady-state performance can be improved over the whole
+ 94 m2 −22 sin(β)m+12
36
range of modulation. Thus, a variable sequence space-vector-
(13) modulation (VSSVM) method is proposed. For simplicity, the

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on October 27,2023 at 18:43:17 UTC from IEEE Xplore. Restrictions apply.
 0.16
TABLE III
M OTOR AND CONTROL PARAMETERS
0.14
DC voltage Udc 540 V
0.12 Rated power PN 2.5 kW
Rated voltage UN 380 V
Rated frequency fN 125 Hz
0.1
Rated torque TN 21 Nm
Pole pairs Np 5
Ihf

0.08 Stator resistance Rs 1.43 Ω

M=0.89 Inductance of d-axis Ld 18.8 mH
0.06
Inductance of q-axis Lq 18.8 mH
M=0.99 Rotor flux ψf 0.3126 Wb
Control period Tsc 200 µs
0.04

Conventional sequence
0.02
Clamping sequence
Double-switching clamping sequence for the two different control periods. Nevertheless, the simu-
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 lation results suggest that considering Tsck+1
= 23 Tsc has little
M effect on the control performance. In order to reduce the com-
Fig. 7. Analytically evaluated current ripple rms over a fundamental period putational burden, urefs is calculated assuming Tsck+1
= Tsc
corresponding to three sequences. in this paper.
For conventional sequence, NPV balance can be achieved
by adjusting the proportion of N-type small vector and P-type
U dc small vector during one control period when ∆Uc exceeds
VSSVM
ref 3 usref M
admissible limitation. For clamping sequence, a homogeneous
 ref i Sequence Sequence
r +
q Voltage
PI dq isref reference usref U dc selection T k +1
sc Pulse Pulses sequence that just applies P-type small vector is introduced,
− vector gen.
r i =0
ref  calculation u ref  Duty cycle
calculation 3L-NPC which is employed when ∆Uc is greater than zero. This
d s
T1 T2 TZ VSI
method can achieve NPV fluctuation control and maintain
nearly identical performance, and in the meanwhile, the
isk +1
switching frequency hardly increase. For double-switching
ia clamping sequence, there is a similar method.
One step delay isk 
compensation
ib
abc In summary, the main implementation steps of the proposed
method include:
d PMSM
1) Measure stator current iks , rotor speed ωr and DC-link
dt e capacitor voltages Uc1 , Uc2 . Calculate ∆Uc according to (4).
2) Predict ik+1
s based on (7) to compensate one-step delay.
Fig. 8. Control diagram of the proposed VSSVM-MPCC.
3) Calculate iref
s and uref
s according to (8) and (16).
4) Calculate M based on (10) and select the optimal se-
quence by Table II.
proposed VSSVM method in this paper selects the optimal
5) If necessary, control NPV fluctuation.
sequence only depending on M , summarized in Table II.
6) Generate switching pulses and apply to the 3L-NPC
inverter at (k + 1)th sampling instant.
B. Principle of VSSVM-MPCC

The proposed VSSVM-MPCC method combines VSSVM V. S IMULATION R ESULTS

and MPCC to improve control performance. The overall con-
In order to verify the effectiveness of the proposed VSSVM-
trol diagram is shown in Fig. 8. Firstly, the voltage reference
MPCC method, simulation studies are implemented on a NPC-
vector is calculated according to the deadbeat control principle.
3LVSI fed PMSM drive. The relevant motor and control
Then, the corresponding modulation index is determined. After
parameters are given in Table III. For comparison, the con-
that, the optimal sequence is selected based on the principle
ventional SVM-MPCC and the proposed VSSVM-MPCC are
of minimizing current ripple. Finally, the optimal sequence is
tested.
generated by SVM. The more details are as follows.
The simulation tests are conducted at 32%, 64%, 94%, 97%
Owing to the adjustment of control period when applying
and 100% rated speed, the average modulation index Mave of
clamping sequence, the voltage reference vector uref s is cal-
that are 0.33, 0.63, 0.93, 0.96 and 0.985, respectively. Fig.
culated as (16) by modifying (8).
9 shows the harmonic spectrum of a-phase stator current at
  100% rated speed under rated load for conventional SVM-
L2 iref
s − ik+1
s
MPCC and VSSVM-MPCC. Comparison of current THD at
uref
s = k+1 k+1

+ Rs ik+1
s + E k+1 (16) other speeds under rated load is listed in Fig. 10. The proposed
Tsc L − 0.5Tsc Rs VSSVM-MPCC presents the superior overall performance.
The current THD is reduced by more than 25% at high
k+1
where Tsc represents (k + 1)th control period. Thus, the modulation indices, when compared with conventional SVM-
voltage reference vector uref
s should be calculated separately MPCC.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on October 27,2023 at 18:43:17 UTC from IEEE Xplore. Restrictions apply.
 10 clamping sequence and double-switching clamping sequence.
5
Thus, a VSSVM method considering the stator current ripple
rms and average switching frequency is proposed, with which
I a (A)

0
the optimal sequence can be selected according to modulation
-5
index M on the principle of minimizing current ripple. The
-10
0.000 0.004 0.008 0.012 0.016 0.020 0.024 0.028 0.032
VSSVM-MPCC generates switching pulses by VSSVM to
time(s) track the current reference after obtaining the voltage reference
Fundamental ( 125Hz ) = 8.9169A THD = 2.1133%
vector uref
s . The simulation results validate the effectiveness
1 of the proposed method. Compared with conventional SVM-
Hn /H1 (%)

MPCC, the proposed VSSVM-MPCC presents much better

0.5 steady-state performance without increasing the computational
burden. The current THD can be reduced by more than
0
0 2 4 6 8 10 12 14 16 18 20
25% at high modulation indices at a given average switching
Frequency (kHz) frequency.
(a)

10

R EFERENCES
I a (A)

-5 [1] D. Casadei, F. Profumo, G. Serra, and A. Tani, “Foc and dtc: two
viable schemes for induction motors torque control,” IEEE Trans. Power
-10
0.000 0.004 0.008 0.012 0.016 0.020 0.024 0.028 0.032
Electron., vol. 17, no. 5, pp. 779–787, Sep. 2002.
time(s) [2] S. Vazquez, J. Rodriguez, M. Rivera, L. G. Franquelo, and M. No-
Fundamental ( 125Hz ) = 8.8636A THD = 1.5738% rambuena, “Model predictive control for power converters and drives:
Advances and trends,” IEEE Trans. Ind. Electron., vol. 64, no. 2, pp.
0.8 935–947, Feb 2017.
Hn /H1 (%)

0.6 [3] Y. Zhang, D. Xu, and L. Huang, “Generalized multiple-vector-based

model predictive control for pmsm drives,” IEEE Trans. Ind. Electron.,
0.4
vol. 65, no. 12, pp. 9356–9366, Dec 2018.
0.2 [4] H. Miranda, P. Cortes, J. I. Yuz, and J. Rodriguez, “Predictive torque
0 control of induction machines based on state-space models,” IEEE Trans.
0 2 4 6 8 10 12 14 16 18 20 Ind. Electron., vol. 56, no. 6, pp. 1916–1924, June 2009.
Frequency (kHz)
[5] L. Yan, M. Dou, and Z. Hua, “Disturbance compensation-based model
(b) predictive flux control of spmsm with optimal duty cycle,” IEEE J.
Emerg. Sel. Topics Power Electron., vol. 7, no. 3, pp. 1872–1882, Sep.
Fig. 9. Harmonic spectrum of a-phase stator current at 100% rated speed
2019.
under rated load. (a) Conventional SVM-MPCC, (b) VSSVM-MPCC.
[6] Y. Zhang, H. Jiang, and H. Yang, “Model predictive control of pmsm
drives based on general discrete space vector modulation,” IEEE Trans.
2.5 Energy Convers., vol. 36, no. 2, pp. 1300–1307, June 2021.
Conventional SVM-MPCC [7] Z. Zhou, C. Xia, Y. Yan, Z. Wang, and T. Shi, “Torque ripple minimiza-
VSSVM-MPCC tion of predictive torque control for pmsm with extended control set,”
IEEE Trans. Ind. Electron., vol. 64, no. 9, pp. 6930–6939, Sep. 2017.
2
[8] Q. Wang, H. Yu, C. Li, X. Lang, S. S. Yeoh, T. Yang, M. Rivera,
S. Bozhko, and P. Wheeler, “A low-complexity optimal switching time-
Stator current THD(%)

modulated model-predictive control for pmsm with three-level npc

1.5 converter,” IEEE Trans. Transport. Electrific., vol. 6, no. 3, pp. 1188–
1198, Sep. 2020.
[9] G. Narayanan and V. Ranganathan, “Analytical evaluation of harmonic
distortion in pwm ac drives using the notion of stator flux ripple,” IEEE
1 Trans. Power Electron., vol. 20, no. 2, pp. 466–474, March 2005.
[10] D. Zhao, V. S. S. P. K. Hari, G. Narayanan, and R. Ayyanar, “Space-
vector-based hybrid pulsewidth modulation techniques for reduced har-
0.5
monic distortion and switching loss,” IEEE Trans. Power Electron.,
vol. 25, no. 3, pp. 760–774, March 2010.
[11] M. Gendrin, J.-Y. Gauthier, and X. Lin-Shi, “A predictive hybrid pulse-
width-modulation technique for active-front-end rectifiers,” IEEE Trans.
0 Power Electron., vol. 32, no. 7, pp. 5487–5496, July 2017.
Mave =0.33 Mave =0.63 Mave =0.93 Mave =0.96 Mave =0.985
[12] S. Das and G. Narayanan, “Novel switching sequences for a space-
Average modulation index vector-modulated three-level inverter,” IEEE Trans. Ind. Electron.,
vol. 59, no. 3, pp. 1477–1487, March 2012.
Fig. 10. Comparison of current THD at different modulation indices under [13] S. Das, G. Narayanan, and M. Pandey, “Space-vector-based hybrid
rated load. pulsewidth modulation techniques for a three-level inverter,” IEEE Trans.
Power Electron., vol. 29, no. 9, pp. 4580–4591, Sep. 2014.
VI. C ONCLUSION [14] K. Nanshikar, K. S., and S. Das, “Space vector based hybrid pwm
schemes for three-level inverters to reduce current ripple,” in 2021
National Power Electronics Conference (NPEC), Dec 2021, pp. 1–6.
This paper proposes a VSSVM-MPCC to improve the [15] Y. Zhang and J. Zhu, “Direct torque control of permanent magnet syn-
steady-state performance of NPC-3LVSI fed PMSM drives. On chronous motor with reduced torque ripple and commutation frequency,”
IEEE Trans. Power Electron., vol. 26, no. 1, pp. 235–248, Jan 2011.
the basis of theoretical analysis, the rms value of stator current [16] J. H. Seo, C. H. Choi, and D. S. Hyun, “A new simplified space-vector
ripple in a control period and a fundamental period is derived pwm method for three-level inverters,” IEEE Trans. Power Electron.,
about three different sequences, i.e., conventional sequence, vol. 16, no. 4, pp. 545–550, July 2001.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on October 27,2023 at 18:43:17 UTC from IEEE Xplore. Restrictions apply.