International Journal of Power Electronics and Drive Systems (IJPEDS)

Vol. 12, No. 4, December 2021, pp. 2251~2260

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v12.i4.pp2251-2260  2251

Modified multistep model predictive control for three-phase

induction motor drive system considering the common-mode
voltage minimization

Bao Binh Pho1, Nguyen Van Cao2, Tran Minh Hoan3, Phuong Vu4
1-4School of Electrical Engineering, Hanoi University of Science and Technology, Ha Noi, Vietnam
1National University of Civil Engineering, Ha Noi, Vietnam

Article Info ABSTRACT

Article history: It is acknowledged that the common-mode voltage may have detrimental
effects on an induction motor (IM) drive system if not properly addressed.
Received Jun 12, 2021 Therefore, in this paper, a modified multistep model predictive control
Revised Sep 13, 2021 method for IM drive system considering the common-mode voltage
Accepted Sep 20, 2021 minimization is proposed. This research uses a multi-objective cost function,
before applying the Sphere Decoding Algorithm to find the optimal control
input. The results show that the proposed control method not only reduces
Keywords: the common-mode voltage significantly but also mitigates the computational
burden of the microprocessor without affecting the system performance. The
Cascaded H-bridge inveter proposed control method is simulated by MATLAB/Simulink for an IM
Common-mode voltage drive system with an 11-level cascaded H-bridge inverter.
Field oriented control
Multistep predictive current
control This is an open access article under the CC BY-SA license.
Sphere decoder

Corresponding Author:
Phuong Vu
School of Electrical Engineering
Hanoi University of Science and Technology
No.1, Dai Co Viet Road, Hai Ba Trung, Hanoi, Vietnam
Email: [email protected]

1. INTRODUCTION
Recently, multilevel inverters (MLI) are widely used in high-power industrial applications because
of these advantages: low output voltage distortion, reducing the voltage level on semiconductor devices and
lowering the du/dt changing speed [1]. Especially, in comparison with other MLI topologies such as neutral-
point clamped [2], flying capacitors [3], multi-modular converters (MMC), the cascaded h-bridge (CHB) [4]-
[6] has emerged as a prominent one for driving IM system due to its high degree of modularity, allowing
upgrade and replacement easily. Furthermore, for the applications of induction motor drives, the connection
of CHB-MLI with multi-phase rectifier transformer ensures the current drawn from the grid have a sinusoidal
waveform, decreases the total harmonic distortion injected into the grid.
On the other hand, a considerable drawback of CHB-MLI that produces the common-mode voltage
(CMV) between the neutral points of the load and the converter [7], which can be calculated by (1) and
illustrated by Figure 1. Due to this CMV, high-frequency electromagnetic interference (EMI) is introduced,
causing problems regarding measurement and system monitoring [8]. Furthermore, in IM driving system, the
CMV generates the common-mode current making detrimental effects on the motor’s productivity and
endurance [9]. There have been many undergoing researches to minimize the CMV, for instance the addition
of the EMI active or passive filter [10], [11]. However, this hardware solution can lead to an increase in the

Journal homepage: http://ijpeds.iaescore.com

2252  ISSN: 2088-8694

size and cost of the system. Therefore, mitigation of CMV based on control method approaches like SVM
[12], [13], MPC [14], [15] is preferred.

1
𝑣𝑍𝑁 = 𝑣𝐶𝑀𝑉 = 3 (𝑣𝐴𝑁 + 𝑣𝐵𝑁 + 𝑣𝐶𝑁 ) (1)

Among these control methods, multistep MPC has significant benefits since it takes the switching
nature of the power converter. It is also shown that multistep MPC can improve the steady-state performance
of the system better than single-step MPC [16]-[20]. However, in [21] multistep model predictive control
(MPC) with CMV minimization target applies to RL load as the case study only. Regarding IM driving
system case study, in [22], the multistep MPC controller does not consider CMV minimization target since it
is designed toward optimizing the switching effort.
Therefore, this paper proposes a modified multistep MPC compared to [22]. To account for CMV,
an additional term that penalizes the CMV is added to the cost function before using the SDA algorithm to
solve the optimization problem. As evidence from the simulation results, this approach also cuts down the
computational burden, confirms the feasibility of this proposal in practical adoption.

IM

C N
B
A
Phase C
Phase B Sxi,3
Phase A

Sxi,1

with xϵ{A,B,C}
CHB-MLI iϵ{1,2…,n}

Figure 1. IM drive system with CHB-MLI.

2. MULTISTEP MODEL PRECDICTIVE CONTROL METHOD

In general, the proposed control method for IM drive system follows the field oriented controller
(FOC) approach, as depicted in Figure 2 [23]-[25]. There are two major control loops. The classical PI
controller takes care of the outer speed and flux control loop, whereas the inner current loop is controlled by
the multistep MPC. Notice that two different weighting factors which account for CMV minimization and
reduce switching effort are applied to the cost function.

Microcontrollers Multistep MPC CHB-MLI

PI isd U*(k) λdc, λCMV
’*
Ѱrd (k)
* (k) i*sαβ(k) IM
dq Y*(k) Cost
_ ωs(k) Reference Uopt(k) Switching
* (k)
isq
’
function
Ѱrd(k) αβ Design
U(k) state table E
JN(k)
PI Y(k)
ω*(k)
_ Current predictive Sphere Decoding
θs(k) ѱrαβ’ (k) model Algorithm
ω(k)

αβ isabc
Sampling
Rotor flux model isαβ(k) abc
ω(k) ω
Sampling

Figure 2. The control scheme for induction motor with inner multistep MPC current control loop

Int J Pow Elec & Dri Syst, Vol. 12, No. 4, December 2021 : 2251 – 2260
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2253

2.1. Current predictive model

First, the physical model of the IM drive system is formulated as differential equations, representing
the relationship between system parameters. After that, a state-space representation of the system can be
obtained. The state-space representation of the continuous-time prediction model of the IM drive system
powered by a CHB-MLI can be written as (2):

𝑑𝒙(𝑡) = 𝑨′(𝑡). 𝒙(𝑡) + 𝑩′𝒖(𝑡)

{ (2)
𝒚(𝑡 + 1) = 𝑪. 𝒙(𝑡 + 1)

With

1 1−𝜎 1−𝜎 1−𝜎

−( + ) 0 𝜔(𝑡)
𝜎𝑇𝑠 𝜎𝑇𝑟 𝜎𝑇𝑟 𝜎
1 1−𝜎 1−𝜎 1−𝜎
0 −( + ) − 𝜔(𝑡)
𝜎𝑇𝑠 𝜎𝑇𝑟 𝜎 𝜎𝑇𝑟
𝑨′ (𝑡) = 1 1
;
0 − −𝜔(𝑡)
𝑇𝑟 𝑇𝑟
1 −1
[ 0 𝜔(𝑡) ]
𝑇𝑟 𝑇𝑟

2 −1 −1
𝑉𝑑𝑐 0 √3 −√3] ;
𝑩’ = [
3𝜎𝐿𝑠 0 0 0
0 0 0
1 0 0 0
𝑪=[ ]
0 1 0 0

′ 𝑇
Where the state vector 𝒙 = [𝑖𝑠𝛼 𝑖𝑠𝛽 𝜓𝑟𝛼 ′
𝜓𝑟𝛽 ] ∈ ℝ4 , the control input vector 𝒖 = 𝑺 =
𝑇
[𝑆𝐴 𝑆𝐵 𝑆𝐶 ] ∈ 𝕌 (the set U is the set of voltage level combinations in three phase), output vector 𝒚 =
𝒊𝑠𝛼𝛽 = [𝑖𝑠𝛼 𝑖𝑠𝛽 ]𝑇 ∈ ℝ2 , notice that 𝑨′(𝑡) is a time-variant matrix because it consists of time-variant
variables  (t ) . Specifically, 𝑖𝑠𝛼 , 𝑖𝑠𝛽 , 𝜓𝑟𝛼
′ ′
, 𝜓𝑟𝛽 stand for stator currents and rotor flux linkages in the
stationary 𝛼𝛽 -framework. System parameters 𝜎, 𝑇𝑠 ,𝑇𝑟 and 𝜔 are defined as total leakage factor, stator time
constant, rotor time constant and rotor speed respectively. After applying the Forward-Euler discretization
with a sampling period of 𝑇, the discrete-time prediction model can be expressed as (3).

𝑥(𝑘 + 1) = 𝐴. 𝑥(𝑘) + 𝐵𝑢(𝑘) 𝐴 = 𝐼4 + 𝑇 . 𝐴′ ∈ ℝ4𝑥4

{ with { (3)
𝑦(𝑘 + 1) = 𝐶. 𝑥(𝑘 + 1) 𝐵 = 𝑇. 𝐵′ ∈ ℝ4x3

Based on the prediction model in (4), the control input and output sequence over the prediction horizon N>1
can be constructed as (5) with relationships in (6).

T
U(k)  (u(k ))T (u(k  1))T ... (u(k  N  1))T   U N (4)

T
Y(k)  (y(k  1))T (y(k  2))T ... (y(k  N ))T   2N
(5)

𝐶𝐴 𝐶𝐵 0 ⋯ 0 0
𝐶𝐴2 𝐶𝐴𝐵 𝐶𝐵 ⋯ 0 0
𝑌(𝑘) = 𝛤𝑥(𝑘) + ϒ𝑈(𝑘) Where 𝛤 = [ ],ϒ = [ ] (6)
⋮ ⋮ ⋮ ⋱ ⋮ ⋮
𝑁
𝐶𝐴 𝐶𝐴𝑁−1 𝐵 𝐶𝐴𝑁−2 𝐵 ⋯ 𝐶𝐴𝐵 𝐶𝐵

2.2. Reference design

Because the stator current and rotor flux signals are slower than electrical ones, so that over the
finite prediction horizon N, these parameters are assumed to be constant. The output reference sequence for
the current controller is denoted by (7).

𝒀∗ (𝑘) = [(𝒚∗ (𝑘 + 1))𝑇 (𝒚∗ (𝑘 + 2))𝑇 . . . (𝒚∗ (𝑘 + 𝑁))𝑇 ]𝑇 ∈ ℝ2𝑁 (7)

Modified multistep model predictive control for three-phase induction motor drive system … (Bao Binh Pho)
2254  ISSN: 2088-8694

𝑇
𝜋
Where𝒚∗ (ℓ) = [√𝑖𝑠𝛼
2 (𝑘) + 𝑖 2 (𝑘). 𝑠𝑖𝑛( 𝜔 ℓ𝑇 + 𝜑 )
𝑠𝛽 𝑠 0
2 (𝑘) + 𝑖 2 (𝑘). 𝑠𝑖𝑛( 𝜔 ℓ𝑇 + 𝜑 + )] ℓ ∈
√𝑖𝑠𝛼 𝑠𝛽 𝑠 0 2
𝑖𝑠𝛽 (𝑘)
{𝑘 + 1; . . . ; 𝑘 + 𝑁}, 𝜑0 = 𝑎𝑟𝑐𝑡𝑎𝑛( ) is the initial angle between 𝑖𝑠𝛼 and 𝑖𝑠𝛽 .
𝑖𝑠𝛼 (𝑘)
The control input reference sequence (8) is designed while considering a null CMV (𝑣𝐶𝑀𝑉 = 0).

𝑼∗ (𝑘) = [(𝒖∗ (𝑘))𝑇 (𝒖∗ (𝑘 + 1))𝑇 ... (𝒖∗ (𝑘 + 𝑁 − 1))𝑇 ]𝑇 (8)

1 0 1−𝜎 1−𝜎
−1 √3 𝑑𝒊∗𝑠𝛼𝛽 (ℓ)
𝜔
𝜎𝐿𝑠 1 1−𝜎 𝜎𝑇𝑟 𝜎
∗
Where 𝒖 (ℓ) = 2 2 .( +( + ) 𝒊∗𝑠𝛼𝛽 (ℓ) −[ 1−𝜎 1−𝜎 ] 𝝍′𝑟𝛼𝛽 (𝑘))
𝑉𝑑𝑐 𝑑𝑡 𝜎𝑇𝑠 𝜎𝑇𝑟
3 −√3 − 𝜔
𝜎 𝜎𝑇𝑟
[2 2 ]

2.3. Cost function

To achieve the desired CMV minimization target, the cost function (9) is modified match up to [22].
In addition to output reference tracking, the proposed method which is designed to minimize the CMV also
tracks the input references.

𝐽𝑁 (𝑘) = ∑𝑘+𝑁−1
𝑙=𝑘 (‖𝒚(ℓ + 1) − 𝒚∗ (ℓ + 1)‖22 + 𝜆𝑑𝑐 ‖𝛥𝒖(ℓ)‖22 + 𝜆𝐶𝑀𝑉 ‖𝒖(ℓ) − 𝒖∗ (ℓ)‖22 ) (9)

In (9),𝒖(ℓ) is the candidate control-input that generates the output current prediction𝒚(ℓ + 1) =
𝒊𝛼𝛽 (ℓ + 1); 𝛥𝑢(ℓ) = 𝑢(ℓ) − 𝑢(ℓ − 1); 𝜆𝑑𝑐 is the weighting factor for lessening switching effort, 𝜆𝐶𝑀𝑉 is the
weighting factor for CMV minimization. Representation of (9) in matrix form can be written as (10):

𝐽𝑁 (𝑘) = ‖𝒀(𝑘) − 𝒀∗ (𝑘)‖22 + 𝜆𝑑𝑐 ‖𝑺𝑼(𝑘) − 𝜠𝒖(𝑘 − 1)‖22 + 𝜆𝐶𝑀𝑉 ‖𝑼(𝑘) − 𝑼∗ (𝑘)‖22 (10)

𝑰3 𝟎3 ⋯ 𝟎3
𝑰3
−𝑰3 𝑰3 ⋯ 𝟎3
𝟎3
Where𝑺 = 𝟎3 −𝑰3 ⋯ 𝟎3 ,𝑬=[ ]
⋮
⋮ ⋮ ⋱ 𝟎3
𝟎3 [3𝑁𝑥3]
[ 𝟎3 𝟎3 ⋯ 𝑰3 ][3𝑁𝑥3𝑁]

2.4. Sphere decoding algorithm (SDA)

In this job, the SDA is adopted to solve the modified optimization problem, to obtain the input
sequence 𝑼(𝑘) = 𝑼𝑜𝑝𝑡 (𝑘) that minimizes the cost function𝐽𝑁 (𝑘) in detail. The computational process will be
presented concisely. Further detail about the SDA can be found in [18], [19]. By using (5), the cost function
(10) becomes:

𝐽𝑁 (𝑘) = 𝑼(𝑘)𝑇 𝑾𝑼(𝑘) +2𝐅(𝑘)𝑇 𝑼(𝑘) + 𝜺(𝑘) (11)

Where
𝑾 = ϒ𝑇 ϒ + 𝜆𝑑𝑐 𝑺𝑇 𝑺 + 𝜆𝐶𝑀𝑉 𝑰3𝑁
𝑭(𝑘) = ϒ𝑇 𝛤𝒙(𝑘) − ϒ𝑇 𝒀∗ (𝑘) − 𝜆𝑑𝑐 𝑺𝑇 𝑬𝒖(𝑘 − 1) − 𝜆𝐶𝑀𝑉 𝑼∗ (𝑘)
𝜺(𝑘) = [‖𝛤𝒙(𝑘) − 𝒀∗ (𝑘)‖22 + 𝜆𝑑𝑐 ‖𝑬𝒖(𝑘 − 1)‖22 + 𝜆𝐶𝑀𝑉 ‖𝑼∗ (𝑘)‖22 ]

𝜕𝐽
By solving 𝑁 = 0, the optimal solution 𝑼𝑢𝑐 is yielded in (12), which does not necessary belong to
𝜕𝑼
the finite control-input set U .

𝑼𝑢𝑐 = −𝑾−1 𝑭(𝑘) (12)

Because W is a symmetric and positive definite matrix for 𝜆𝑑𝑐 , 𝜆𝐶𝑀𝑉 >0, there exists only one unique
invertible lower triangular matrix H, which can be obtained by performing the Cholesky decomposition to
𝑊 −1 :𝑊 −1 = 𝑯−1 𝑯−𝑇 . Hence, the matrix 𝑯 will satisfy the equation: 𝑊 = 𝑯𝑇 𝑯. Consequently, the cost
function (11) can be rewritten as (13). Figure 3 illustrates the graphical representation of a MPC problem in
two-dimensional space. To simplify, control input vectors referring to the same ellipse will have the same
cost value. By transforming the cost function (11) into (13), the original ellipses are transformed into circles.
This procedure plays an important role in SDA.

Int J Pow Elec & Dri Syst, Vol. 12, No. 4, December 2021 : 2251 – 2260
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2255

𝐽𝑁 (𝑘) = ‖𝑯𝑼(𝑘) − 𝑼̄𝑢𝑐 (𝑘)‖22 where Uuc (k )  HUuc (k ) (13)

H ρ2
HU2

HU3
Uuc
ρ3 HUuc
HU1
ρ1

Figure 3. Representation of the meaning of the matrix H

The optimal control input for prediction horizon N can be presented as (14):

𝑼𝑜𝑝𝑡 (𝑘) = [𝒖𝑎𝑏𝑐 (𝑘) 𝒖𝑎𝑏𝑐 (𝑘 + 1) … 𝒖𝑎𝑏𝑐 (𝑘 + 𝑁 − 1)]𝑇 (14)

Now, reconstruct (14) into a sequence (element by element) as (15):

𝑼𝑜𝑝𝑡 = [𝑢1 𝑢2 … 𝑢3𝑁 ]𝑇 với i  1, 2,...,3N (15)

Figure 4 depicts the evaluation process of each element 𝑢𝑖 , so-called a node, of the control input
sequence U in (15). Figure 5 illustrates the flow diagram of SDA. First, an initial sphere 𝑆𝑖𝑛𝑖 is defined with
an initial center 𝑼̄𝑢𝑐 (𝑘) calculated by (16) and a selective initial radius 𝑝𝑖𝑛𝑖 . This initial sphere 𝑆𝑖𝑛𝑖 should be
small enough to contain at least one solution 𝑼(𝑘). Then during the evaluation process, the size of the sphere
is decreased until there is only one solution contained in it, and that is the optimal control input sequence.
Each time a node is visited, the radius 𝑝𝑖 is calculated by (17) before evaluating the condition (18). If 𝑝𝑖
violates condition (18), the node 𝑢𝑖 will be discarded and its following nodes (from 𝑢𝑖+1 to 𝑢3𝑁 ) will be
discarded without performing any computation to avoid unnecessary computation. The control input having
the smallest radius will be the optimal solution 𝑼(𝑘) = 𝑼𝑜𝑝𝑡 (𝑘).

ρopt = ρini

u1 -n n

u2 -n n -n n

u3N-1 -n n

u3N -n n -n n

Update Uopt ,ρopt

( : Node is visited, : Node is discarded, : Node is discarded without performing any computation)

Figure 4. A tree-diagram of the control input sequence

Modified multistep model predictive control for three-phase induction motor drive system … (Bao Binh Pho)
2256  ISSN: 2088-8694

Start
Uuc(k), H, ρpast =ρini,
i=0, c0=0, ρ0=0, u(k-1)

ρi=(H[i,1:i]U[1:i]-Uuc[i])2+ρi-1 i=i+1, Ui=-n+c0

N
i=i-1,c0=c0+1 If ρi<ρpast Y
If -n ≤ Ui ≤ n
Y
N N
c0=0 If i=3N
i=i-2
Y
Update Uopt(k)=U, ρpast=ρi
Y
c0=Ui+1+n+1 If i ≥ 0
N
apply Uopt(k)

End

Figure 5. SDA flow diagram [20]

𝑼̄𝒖𝒄 (𝒌) = 𝑯(−𝑾−𝟏 )𝑭(𝒌) (16)

𝝆𝟐𝒊 = || ⏟ [𝒖𝟏 . . . 𝒖𝒊 ]𝑻 − 𝒖̄⏟

[𝒉𝒊𝟏 . . . 𝒉𝒊𝒊 ] ⏟ 𝟐 𝟐
𝒖𝒄,𝒊 ||𝟐 + 𝝆𝒊−𝟏 (17)
𝑯[𝒊,𝟏:𝒊] 𝑼𝒐𝒑𝒕[𝟏:𝒊] ̄ 𝒖𝒄[𝒊]
𝑼

Condition: 𝝆𝟐𝒊 ≤ 𝝆𝟐𝒐𝒑𝒕 (18)

2.5. Switching state table

When the optimal solution 𝑼(𝒌)is obtained, only the first element 𝒖𝒂𝒃𝒄 (𝒌) is employed to the find
the switching states for the inverter. Notice that 𝒖𝒂𝒃𝒄 (𝒌) is the phase voltage levels. Table 1 shows the
switching states selection strategy clearly. In Table 1, variable represent the phases, the index notation
denotes the position-number of a H-bridge in phase, is the voltage level of phase, is the voltage level of the
H-bridge in phase, are the switching states of valve 1 and 3 of the H-bridge in phase. Figure 1 illustrates this
switching state selection strategy.

Table 1. Switching state selection strategy

𝑆𝑥𝑖 (𝑆𝑥𝑖,1 ; 𝑆𝑥𝑖,3 )
𝑆𝑥
𝑆𝑥1 (𝑆𝑥1,1 ; 𝑆𝑥1,3 ) 𝑆𝑥2 (𝑆𝑥2,1 ; 𝑆𝑥2,3 ) 𝑆𝑥3 (𝑆𝑥3,1 ; 𝑆𝑥3,3 ) 𝑆𝑥4 (𝑆𝑥4,1 ; 𝑆𝑥4,3 ) 𝑆𝑥5 (𝑆𝑥5,1 ; 𝑆𝑥5,3 )
+5 1(1;0) 1(1;0) 1(1;0) 1(1;0) 1(1;0)
+4 1(1;0) 1(1;0) 1(1;0) 1(1;0) 0(0;0)
+3 1(1;0) 1(1;0) 1(1;0) 0(0;0) 0(0;0)
+2 1(1;0) 1(1;0) 0(0;0) 0(0;0) 0(0;0)
+1 1(1;0) 0(0;0) 0(0;0) 0(0;0) 0(0;0)
0 0(0;0) 0(0;0) 0(0;0) 0(0;0) 0(0;0)
-1 0(0;0) 0(0;0) 0(0;0) 0(0;0) -1(0;1)
-2 0(0;0) 0(0;0) 0(0;0) -1(0;1) -1(0;1)
-3 0(0;0) 0(0;0) -1(0;1) -1(0;1) -1(0;1)
-4 0(0;0) -1(0;1) -1(0;1) -1(0;1) -1(0;1)
-5 -1(0;1) -1(0;1) -1(0;1) -1(0;1) -1(0;1)

Int J Pow Elec & Dri Syst, Vol. 12, No. 4, December 2021 : 2251 – 2260
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2257

3. SIMULATION RESULTS
The proposed control scheme is simulated by MATLAB/Simulink for IM drive system with an 11-
level CHB, DC voltage supply per CHB 𝑉𝑑𝑐 = 600𝑉 and the prediction horizon N=3. The IM drive system
parameters and the simulation scenarios are shown in Tables 2 and 3 respectively. To definitely demonstrate
the benefits of the proposed control method in company with [22], the CMV weighting factors are applied to
the cost function from 1,35s to 1,45s. The sampling period 𝑇 = 50𝜇𝑠.

Table 2. IM drive system parameters

Nominal Power 𝑃𝑑𝑚 1119 KW
Nominal Electromagnetic Torque 𝑀𝑑𝑚 6906 Nm
Nominal Current 𝐼𝑑𝑚 234.18 A
Nominal Voltage 𝑈𝑑𝑚 3.3 kV
Nominal Speed 𝜔𝑑𝑚 1470 rpm

Table 3. Simulation scenarios

Time 0-0.5 [s] 0.5-1 [s] 1-1.25 [s] 1.25-1.35 [s] 1.35-1.45 [s]
Reference speed Magnetization Gradually increase speed Speed Stabilization 𝜔ref = 1470𝑟𝑝𝑚
𝜔ref = 0 𝜔ref = 1470𝑟𝑝𝑚
Weighting Factors 𝜆𝑑𝑐 ≈ 0; 𝜆𝐶𝑀𝑉 ≈ 0 𝜆𝐶𝑀𝑉 ≈ 0𝜆𝑑𝑐 = 4 𝜆𝐶𝑀𝑉 = 4𝜆𝑑𝑐 = 4

It is presented in Figure 6 that the rotor flux, the rotor speed, and the electromagnetic torque are
smoothly ramped-up and quickly reach their reference values. Figure 7 illustrates the stator current and the
CMV at the steady-state. The effectiveness of the proposed control method is clearly demonstrated when the
CMV weighting factor is applied in the period of (1.35-1.45) [s]. Especially in (1.15-1.25) [s] and (1.25-1.35)
[s], the CMV is registered at 1800V (3𝑉𝑑𝑐 ), while in (1.35-1.45) [s], the CMV is drastically reduced to 200V
1
( . 𝑉𝑑𝑐 ). Moreover, Table 4 depicts the number of evaluated nodes during the SDA optimization process.
3
It can be seen that this number is significantly decreased from 144,969 nodes to 88,396 nodes,
indicates the reduction of the computational burden evidently. However, these benefits are traded for a small
increase of THD and switching effort, specifically from 0.58% to 0.61% for THD and from 2.43 to 2.78
times per cycle for switching effort. Finally, Figure 8 shows that the stator current in  -framework is
attained its reference for all simulation scenarios with the error below 5%. Hence, the additional CMV
minimization target does not affect the system performance.

1500 1472
1000
Speed
(rpm)

1470
500
0 1468
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.2 1.3 1.4

8000
Momen

6000 7000
(Nm)

4000
2000 6900
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 6800
1.2 1.3 1.4

10 7.635
(Wb)
Flux

5 7.63
7.625
0 7.62
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.2 1.3 1.4
Time (s)
Reference value Feedback value

Figure 6. Graphs of motor speed, torque and motor flux

Modified multistep model predictive control for three-phase induction motor drive system … (Bao Binh Pho)
2258  ISSN: 2088-8694

Reference [22] Proposal method

400

CMV (V) Current (A)

200
0
-200
-400
1.15 1.2 1.25 1.3 1.35 1.4 1.45
2000 Thời gian (s)
0
-2000
1.15 1.2 1.25 1.3 1.35 1.4 1.45
Time(s)
Phase A Phase B Phase C Common-mode Voltage

Figure 7. Graphs of stator currents and common-mode voltage

360 360 360

Current αβ (A)

340 340 340

320 320 320

1.165 1.17 1.175 1.18 1.265 1.27 1.275 1.28 1.365 1.37 1.375 1.38

400
200
0
-200
-400
1.15 1.2 1.25 1.3 1.35 1.4 1.45
Error (%)

5
0
-5
1.15 1.2 1.25 1.3 1.35 1.4 1.45
Time (s)
Isα ref Isβ ref Isα feedback Isβ feedback Error Isα Error Isβ

Figure 8. Graph of stator current and error value in the αβ- framework

Table 4. Results about THD, switching effort, and number of evaluated nodes
Time 1.15-1.25 [s] 1.25-1.35 [s] ( Reference [22]) 1.35-1.45 [s] (Proposal)
THD 0.46% 0.58% 0.61%
Switching Effort 3.26 (times/ cycle) 2.43 (times/ cycle) 2.78 (times/ cycle)
Number of evaluated nodes 314336 (nodes) 144969 (nodes) 88396 (nodes)

4. CONCLUSION
This paper proposes a modified multistep MPC for IM drive system with a CHB-MLI. This
proposed method achieves two main contributions: firstly, the reduction of CMV while retaining the current
quality and switching effort at the desired range. Secondly, the required execution time for microcontrollers
is decreased. Based on this potentiality, future research work will be focused on improving the controller
performance in advance for practical implementation.

ACKNOWLEDGEMENT
Our research is funded by the project of National University of Civil Engineering (NUCE)
numbered 30-2021/KHXD-TĐ.

REFERENCES
[1] K. Dhineshkumar, C. Subramani, A. Geetha, and C. Vimala, “Performance analysis of PV powered multilevel
inverter,” International Journal of Electrical and Computer Engineering (IJECE), vol. 9, no. 2, p. 753-760, April
2019, doi: 10.11591/ijece.v9i2.pp753-760.
[2] P. Barbosa, P. Steimer, L. Meysenc, M. Winkelnkemper, J. Steinke, and N. Celanovic, “Active Neutral-Point-
Clamped Multilevel Converters,” 2005 IEEE 36th Power Electronics Specialists Conference, 2005, pp. 2296-2301,
doi: 10.1109/PESC.2005.1581952.
[3] S. Du, B. Wu, N. R. Zargari, and Z. Cheng, “A Flying-Capacitor Modular Multilevel Converter for Medium-
Voltage Motor Drive,” IEEE Transactions on Power Electronics, vol. 32, no. 3, pp. 2081-2089, March 2017, doi:
10.1109/TPEL.2016.2565510.

Int J Pow Elec & Dri Syst, Vol. 12, No. 4, December 2021 : 2251 – 2260
Int J Pow Elec & Dri Syst ISSN: 2088-8694  2259

[4] M. V. Chung, D. T. Anh, P. Vu, and M. L. Nguyen, “Hardware in the loop co-simulation of finite set-model
predictive control using fpga for a three level CHB inverter,” International Journal of Power Electronics and Drive
Systems (IJPEDS), vol. 11, no. 4, pp. 1719-1730, December 2020, doi: 10.11591/ijpeds.v11.i4.pp1719-1730.
[5] V. C. Mai, M. L. Nguyen, T. H. Vo, P. V. Hoang, and T. M. Tran, “Hardware In the Loop Simulation of Predictive
Current Control for IM Fed by Multi-Level Cascaded H-Bridge Inverters,” 2019 IEEE Vehicle Power and
Propulsion Conference (VPPC), 2019, pp. 1-5, doi: 10.1109/VPPC46532.2019.8952288.
[6] C. M. Van, T. N. Xuan, P. V. Hoang, M. T. Trong, S. P. Cong, and L. N. Van, “A Generalized Space Vector
Modulation for Cascaded H-bridge Multi-level Inverter,” 2019 International Conference on System Science and
Engineering (ICSSE), 2019, pp. 18-24, doi: 10.1109/ICSSE.2019.8823465.
[7] S. Wei, N. Zargari, B. Wu, and S. Rizzo, “Comparison and mitigation of common mode voltage in power converter
topologies,” Conference Record of the 2004 IEEE Industry Applications Conference, 2004. 39th IAS Annual
Meeting., 2004, pp. 1852-1857 vol.3, doi: 10.1109/IAS.2004.1348722.
[8] Yen-Shin Lai, Po-Sheng Chen, Hsiang-Kuo Lee, and J. Chou, “Optimal common-mode voltage reduction PWM
technique for inverter control with consideration of the dead-time effects-part II: applications to IM drives with
diode front end,” IEEE Transactions on Industry Applications, vol. 40, no. 6, pp. 1613-1620, Nov.-Dec. 2004,
doi: 10.1109/TIA.2004.836151.
[9] I. Seo, N. B. Belaynehn, C. Park, and J. Kim, “A Study of Common Mode Voltage Generation According to
Modulation Methods and Reduction Strategies on MMC System,” 2018 IEEE Energy Conversion Congress and
Exposition (ECCE), 2018, pp. 3988-3995, doi: 10.1109/ECCE.2018.8557920.
[10] J. Kalaiselvi and S. Srinivas, “Passive common mode filter for reducing shaft voltage, ground current, bearing
current in dual two level inverter fed open end winding induction motor,” 2014 International Conference on
Optimization of Electrical and Electronic Equipment (OPTIM), 2014, pp. 595-600, doi:
10.1109/OPTIM.2014.6850977.
[11] H. Akagi, H. Hasegawa, and T. Doumoto, “Design and performance of a passive EMI filter for use with a voltage-
source PWM inverter having sinusoidal output voltage and zero common-mode voltage,” IEEE Transactions on
Power Electronics, vol. 19, no. 4, pp. 1069-1076, July 2004, doi: 10.1109/TPEL.2004.830039.
[12] H. Truong, C. Mai, C. Nguyen, and P. Vu, “Modified Space Vector Modulation for Cascaded H-Bridge Multilevel
Inverter with Open-Circuit Power Cells,” Journal of Electrical and Computer Engineering, vol. 2021, pp. 1-14,
2021, doi: 10.1155/2021/6643589.
[13] L. G. G. P. de Castro, M. B. R. Corrêa, and C. B. Jacobina, “A fast space-vector algorithm for common-mode
voltage elimination in multilevel converters,” 2013 Brazilian Power Electronics Conference, 2013, pp. 243-247,
doi: 10.1109/COBEP.2013.6785122.
[14] C. Mai-Van, S. Duong-Minh, D. Tran-Huu, B. Binh-Pho, and P. Vu, “An improved method of model predictive
current control for multilevel cascaded H-bridge inverters,” Journal of Electrical Engineering, vol. 72, no. 1, pp. 1-
11, 2021, doi: 10.2478/jee-2021-0001.
[15] S. Kwak and S. Mun, “Model Predictive Control Methods to Reduce Common-Mode Voltage for Three-Phase
Voltage Source Inverters,” IEEE Transactions on Power Electronics, vol. 30, no. 9, pp. 5019-5035, Sept. 2015, doi:
10.1109/TPEL.2014.2362762.
[16] T. Geyer, P. Karamanakos, and R. Kennel, “On the benefit of long-horizon direct model predictive control for
drives with LC filters,” 2014 IEEE Energy Conversion Congress and Exposition (ECCE), 2014, pp. 3520-3527,
doi: 10.1109/ECCE.2014.6953879.
[17] T. Geyer and D. E. Quevedo, “Performance of Multistep Finite Control Set Model Predictive Control for Power
Electronics,” IEEE Transactions on Power Electronics, vol. 30, no. 3, pp. 1633-1644, March 2015, doi:
10.1109/TPEL.2014.2316173.
[18] F. Grimm, P. Kolahian, Z. Zhang, and M. Baghdadi, “A Sphere Decoding Algorithm for Multistep Sequential
Model-Predictive Control,” IEEE Transactions on Industry Applications, vol. 57, no. 3, pp. 2931-2940, May-June
2021, doi: 10.1109/TIA.2021.3060694.
[19] T. Geyer and D. E. Quevedo, “Multistep Finite Control Set Model Predictive Control for Power Electronics,” IEEE
Transactions on Power Electronics, vol. 29, no. 12, pp. 6836-6846, Dec. 2014, doi: 10.1109/TPEL.2014.2306939.
[20] P.Karamanakos, T. Geyer, N. Oikonomou, F. D. Kieferndorf, and S. Manias, “Direct model predictive control: A
review of strategies that achieve long prediction intervals for power electronics,” IEEE Industrial Electronics
Magazine. 2014, doi: 10.1109/MIE.2013.2290474.
[21] R. Baidya, R. P. Aguilera, P. Acuña, S. Vazquez and H. d. T. Mouton, "Multistep Model Predictive Control for
Cascaded H-Bridge Inverters: Formulation and Analysis," in IEEE Transactions on Power Electronics, vol. 33, no.
1, pp. 876-886, Jan. 2018, doi: 10.1109/TPEL.2017.2670567.
[22] R. Baidya, et al., “Dealing with Suboptimality in Multistep Model Predictive Control for Transient Operations,”
2019 IEEE Energy Conversion Congress and Exposition (ECCE), 2019, pp. 3780-3785, doi:
10.1109/ECCE.2019.8912815.
[23] F. Wang, et al., “An Encoderless Predictive Torque Control for an Induction Machine With a Revised Prediction
Model and EFOSMO,” IEEE Transactions on Industrial Electronics, vol. 61, no. 12, pp. 6635-6644, Dec. 2014,
doi: 10.1109/TIE.2014.2317140.
[24] M. M. Bech, J. K. Pedersen, and F. Blaabjerg, “Field-oriented control of an induction motor using random
pulsewidth modulation,” in IEEE Transactions on Industry Applications, vol. 37, no. 6, pp. 1777-1785, Nov.-Dec.
2001, doi: 10.1109/28.968191.

Modified multistep model predictive control for three-phase induction motor drive system … (Bao Binh Pho)
2260  ISSN: 2088-8694

[25] A. K. Ali and R. G. Omar, “Finite control set model predictive direct current control strategy with constraints
applying to drive three-phase induction motor,” International Journal of Electrical and Computer Engineering
(IJECE), vol. 11, no. 4, pp. 2916-2924, August 2021, doi: 10.11591/ijece.v11i4.pp2916-2924.

BIOGRAPHIES OF AUTHORS

Pho Bao Binh received her MS degree from Hanoi University of Science and
Technology, Vietnam in 2006 in Control Engineering and Automation. Since 2019, she
has started to study her PhD degree at Hanoi University of Science and Technology. Her
research interests include multilevel converter, model predictive current control.

Nguyen Van Cao has been a student at School of Electrical Engineering, Hanoi
University of Science and Technology, Vietnam since 2016. He is a member of Power
Electronic Laboratory managed by PhD Phuong Vu-Hoang. His research interests include
cascaded H-bridge multilevel converter and its applications in electrical machine drive and
photovoltaic system.

Tran Minh Hoan is a student at Hanoi University of Science and Technology. Since 2019,
he has been a member of Power Electronic Laboratory managed by PhD Phuong Vu-Hoang.
His current research interest are the high-power converters, especially the cascaded H-bridge
multilevel inverter.

Phuong Vu received his B.S., M.S., and Ph.D. degrees from Hanoi University of Science
and Technology, Vietnam, in 2006, 2008, and 2014, respectively, all in Control
Engineering and Automation. Since 2006 he has been employed at Hanoi University of
Science and Technology, where he is a lecturer and researcher at school of electrical
engineering. His research interests include modeling and controlling of power electronics
converters for applications such as photovoltaic, wind system, electrical machine drive.

Int J Pow Elec & Dri Syst, Vol. 12, No. 4, December 2021 : 2251 – 2260