Sliding Mode Control of the Modular Multilevel

Converter
Qichen Yang Maryam Saeedifard
School of Electrical and Computer Engineering School of Electrical and Computer Engineering
Georgia Institute of Technology Georgia Institute of Technology
Atlanta, Georgia 30332 Atlanta, Georgia 30332
Email: qyang45@gatech.edu Email: maryam@ece.gatech.edu

Abstract—The modular multilevel converter (MMC) has be-

+ + + i +
come the most promising converter topology for medium- and dc
SM 1 SM 1 SM 1
high-power applications. In this paper, a sliding mode control +
(SMC)-based method is proposed for the MMC. Based on C
_
vCx,j v p,a v p,b v p,c
analysis of the system dynamics under the proposed control
method, control parameter relations and their validity conditions SM N SM N SM N
are obtained, which provide guidance for systematic controller _ _ _
design. Compared to the conventional control methods, the i p,a
i p,b i p,c 1
proposed SMC-based method provides a faster dynamic response. V
Performance and effectiveness of the proposed control method are r r r 2 DC
demonstrated based on simulation studies in the PSCAD/EMTDC
software environment as well as experiments conducted on a l l l
laboratory prototype. a
vAC ia
I. I NTRODUCTION vAC
b _
ib O
The modular multilevel converter (MMC) has become the c
vAC ic +
most promising converter topology for medium- and high-
L R
power applications [1]–[4]. In the technical literatures, control
methods based on linear controllers, e.g., proportional-integral l l l
(PI) and proportional-resonant (PR) controllers, have been 1 V
proposed to control the MMC [1], [5]–[10]. In addition, r r r
2 DC
control methods based on nonlinear controllers are also pro- i n,a_ i n,b_ i n,c_
posed [11]–[15], which suffer from large computation loads
SM 1 SM 1 SM 1
and/or dependence on exact system models. In this paper, a
sliding mode control (SMC)-based method is proposed for the vn,a v n,b vn,c
MMC, which provides a faster dynamic response and better
robustness compared with the conventional methods. Based SM N SM N SM N
_
on analysis of a wide range of the system dynamics under + + +
the proposed control method, control parameter relations and
Fig. 1. Circuit diagram of the DC-AC MMC.
their validity conditions are obtained, which provide guidance
for systematic controller design. In addition, the SMC-based
control method can be incorporated into the conventional half-bridge (HB) SMs and a series-connected arm inductor.
control method to combine the advantages of both the SMC- Dynamics of the MMC can be summarized by [16], [17]
and conventional PI-based control methods. Effectiveness of 𝑝,𝑗 𝑗
1 𝑝,𝑗
the SMC-based control method is verified based on simulation 2 𝑉DC − 𝑛𝑝,𝑗 𝑣𝐶 −( 𝑟𝑖𝑝,𝑗 − 𝑙 𝑑𝑖𝑑𝑡 − 𝑅𝑖𝑗 − 𝐿 𝑑𝑖
)
𝑑𝑡
studies in the PSCAD/EMTDC software environment as well − 𝑉AC sin(𝜃 − 𝜙𝑗 ) + 𝑣ac,com = 0,
𝑛,𝑗 𝑑𝑖𝑛,𝑗 𝑑𝑖𝑗
as experiments conducted on a laboratory prototype. − 12 𝑉DC − 𝑛𝑛,𝑗 𝑣𝐶 𝑛,𝑗 𝑗
(− 𝑟𝑖 − 𝑙 𝑑𝑡 𝑗 − 𝑅𝑖 − 𝐿 )𝑑𝑡
− 𝑉AC sin(𝜃 − 𝜙 ) + 𝑣ac,com = 0, (1)
𝑝,𝑗
II. DYNAMICS OF THE M ODULAR M ULTILEVEL 𝑑𝑣𝐶 𝑛𝑝,𝑗 𝑝,𝑗
𝐶 𝑑𝑡 = 𝑁 𝑖 ,
C ONVERTER 𝑛,𝑗
𝑑𝑣𝐶 𝑛𝑛,𝑗 𝑛,𝑗
𝐶 𝑑𝑡 = − 𝑁 𝑖 ,
The circuit diagram of a three-phase MMC is shown in
𝑥,𝑗
Fig. 1. The MMC consists of two arms per phase where each where 𝑛𝑥,𝑗 , 𝑣𝐶 , and 𝑖𝑥,𝑗 represent the number of inserted
arm is comprised of 𝑁 series-connected, nominally identical SMs, averaged SM capacitor voltage, and arm current in the

978-1-5386-1180-7/18/$31.00 ©2018 IEEE 1036

𝑥 (𝑥 = 𝑝, 𝑛) arm of phase 𝑗 (𝑗 = 𝑎, 𝑏, 𝑐), respectively, 𝑣ac,com III. T HE P ROPOSED S LIDING M ODE C ONTROL
is the common-mode component of the AC-side terminal
In the proposed SMC-based control method, common- and
voltages of the MMC, 𝑉DC is the DC-link voltage, 𝑉AC is
𝑗 differential-mode currents are actively controlled and two
the peak value of 𝑣AC , 𝑖𝑗 is the phase 𝑗 current, 𝜃 is the
𝑗 control structures are designed for each actively-controlled
phase angle of 𝑣AC , and 𝜙𝑗 is the initial phase angle of 𝑣AC
𝑎
.
variable. The two control structures are applied alternatively
Generally, the control inputs are obtained based on
for a certain time length, i.e., Δ𝑡, to force the corresponding
𝑝,𝑗 𝑛,𝑗 ( ) variable to slide along the boundary of the two structures
𝑛𝑝,𝑗 𝑣𝐶 − 𝑛𝑛,𝑗 𝑣𝐶 ≈ 𝑛𝑝,𝑗 − 𝑛𝑛,𝑗 𝑉𝐶ref
Δ within a hysteresis range, i.e., ±Δ𝑖𝑗com or ±Δ𝑖𝑗dif .
( = 𝑓)(⋅) + ℎfw,com (⋅) ,
𝑝,𝑗 𝑛,𝑗 (2)
𝑛𝑝,𝑗 𝑣𝐶 + 𝑛𝑛,𝑗 𝑣𝐶 ≈ 𝑛𝑝,𝑗 + 𝑛𝑛,𝑗 𝑉𝐶ref A. Single-Phase MMC
Δ
= 𝑔 (⋅) + ℎfw,dif , Since single-phase MMC has only one phase, the superscript
𝑗 of variables are removed in this sub-section III-A. To control
where 𝑉𝐶ref is the SM capacitor voltage reference, 𝑓 (⋅) and 𝑖com and 𝑖dif in a single-phase MMC, each of 𝑓 (⋅) and 𝑔(⋅)
𝑔(⋅) are functions to control the ac-side currents and the in (2) corresponds to two structures, i.e.,
circulating currents of the MMC, respectively, ℎfw,com (⋅) and {
ℎfw,dif are feed-forward components to control the ac-side 𝐹, if 𝑖com > 𝑖ref com
𝑓 (⋅) = , (7a)
currents and the circulating currents of the MMC. Common- −𝐹, if 𝑖com < 𝑖ref com
{
and differential-mode components are defined by: 𝐺, if 𝑖dif > 𝑖ref dif ,
𝑔 (⋅) = (7b)
[ 𝑗 ] [ ] [ 𝑝,𝑗 ] −𝐺, if 𝑖dif < 𝑖ref dif
𝑧com Δ 1 1 1 𝑧
𝑗 = 𝑛,𝑗 , (3) 𝑗
𝑧dif 2 1 −1 𝑧 where 𝑖ref ref 𝑗
com and 𝑖dif are references of 𝑖com and 𝑖dif , 𝐹
is commanded magnitude of voltage to control the ac-side
𝑥,𝑗
where 𝑧 𝑥,𝑗 is a general designation for 𝑖𝑥,𝑗 , 𝑣 𝑥,𝑗 , 𝑣𝐶 and currents, and 𝐺 is commanded magnitude of voltage to control
𝑥,𝑗 𝑗 𝑥,𝑗 𝑗
𝑛 , 𝑧com is common-mode component of 𝑧 , and 𝑧dif is the circulating currents. Based on (7), the two structures within
differential-mode component of 𝑧 𝑥,𝑗 . Based on (3), 𝑖𝑗com is each pair alternate based on comparison between the actively-
𝑥,𝑗
half of 𝑖𝑗 . Substituting (2) and (3) for 𝑖𝑥,𝑗 , 𝑛𝑥,𝑗 , and 𝑣𝐶 in controlled current and its reference. Besides, to generate 𝑛𝑗com
(1) yields and 𝑛𝑗dif , the feed-forward components in (2) are
⎡ ⎤ ℎfw,com (⋅) = −𝑉AC sin (𝜃) ,
𝑖𝑗com (8a)
𝑑
⎢
⎢ 𝑗 𝑖𝑗dif
⎥
⎥ ℎfw,dif = −𝑉DC . (8b)
𝑑𝜃 ⎣ 𝑣𝐶,com ⎦
𝑗 In a well-designed MMC, the magnitude of ripple com-
𝑣𝐶,dif ⎡ / ⎤
⎡ ⎤ ponents of each SM capacitor voltage are regulated within a
𝑖𝑗com 𝑗
𝑣AC 𝐿AC
[ ] ⎢ ⎥ limited range, i.e.,
𝐴11 𝐴12 ⎢ 𝑖𝑗dif ⎥ ⎢ ⎥
⎢ 𝑗 ⎥ 1 −𝑉dc /𝑙
=
𝐴21 𝐴22 ⎣ 𝑣𝐶,com ⎦ − 2𝜔 ⎢
⎣
⎥.
⎦ −Δ𝑣𝐶,com ≤ 𝑣𝐶,com − 𝑉𝐶ref ≤ Δ𝑣𝐶,com ,
0 (9)
𝑗
𝑣𝐶,dif 0 −Δ𝑣𝐶,dif ≤ 𝑣𝐶,dif ≤ Δ𝑣𝐶,dif ,
(4) where Δ𝑣𝐶,com and Δ𝑣𝐶,dif are ripple magnitudes of 𝑣𝐶,com
Matrices in (4) are and 𝑣𝐶,dif , respectively. Assuming 𝑖com is larger than 𝑖refcom and
[ 𝑅AC
] substituting (6a), (7a), (8), and (9) for 𝑛com , 𝑛dif , 𝑣𝐶,com , and
− 𝜔𝐿 0
𝐴11 = AC
𝑟 , 𝑣𝐶,dif in (4) yields
0 − 𝜔𝑙
[ ]
𝑛𝑗dif
− 𝜔𝐿AC
𝑛𝑗com 𝑑𝑖com 𝑅AC 1
𝐴12 = 𝜔𝐿AC , =− 𝑖com − 𝑈com , (10)
𝑛𝑗com 𝑛𝑗dif (5) 𝑑𝜃 𝜔𝐿AC 2𝜔𝐿AC
− 𝜔𝑙
[
𝜔𝑙 ] where 𝑈com is the actual magnitude of voltage to control the
1 𝑛𝑗dif −𝑛𝑗com
𝐴21 = , ac-side currents and is
𝜔𝑁 𝐶 −𝑛𝑗com 𝑛𝑗dif
𝐴22 = 0, 𝑈com =
Δ𝑣 𝐹 Δ𝑣𝐶,com 𝑣AC Δ𝑣𝐶,dif (±𝐺+𝑉DC ) (11)
𝐹 ± 𝑉𝐶,com
ref ∓ 𝑉𝐶ref
∓ 𝑉𝐶ref
.
where 𝐶

𝑓 (⋅)+ℎfw,com (⋅) Based on (11), the range of 𝑈com is

𝑛𝑗dif = 2𝑉𝐶ref
, (6a)
𝑔(⋅)+ℎfw,dif
𝑈com ≤ 𝑈com,max
𝑛𝑗com = , (6b) Δ Δ𝑣𝐶,com (𝐹 +𝑉AC )+Δ𝑣𝐶,dif (𝐺+𝑉DC )
2𝑉𝐶ref =𝐹 + 𝑉𝐶ref
,
(12)
𝜔 is angular frequency of the AC source, 𝐿AC is 2𝐿 + 𝑙, and 𝑈com ≥ 𝑈com,min
Δ Δ𝑣𝐶,com (𝐹 +𝑉AC )+Δ𝑣𝐶,dif (𝐺+𝑉DC )
𝑅AC is 2𝑅 + 𝑟. =𝐹 − 𝑉𝐶ref
.

1037
Since Δ𝑡 is much smaller than the line period, 𝑈com can be Similarly, relation among 𝐺, Δ𝑡, and ripple magnitude of
regarded as a constant when a control structure is applied. 𝑖dif , i.e., Δ𝑖dif is
Solving (10) for 𝑖com yields 2𝑙Δ𝑖dif
Δ𝑡 = 𝑟 (𝑖ref
( ) dif +Δ𝑖dif )+𝑈dif
𝑅AC 2𝑙Δ𝑖dif
𝑖com (𝜃) = 𝑖com (𝜃0 ) exp − 𝜔𝐿 (𝜃 − 𝜃 0 ) ≈ 𝑈dif
(21)
( ( AC ) ) 2𝑙Δ𝑖dif
+ 𝑈𝑅com exp − 𝑅AC
(𝜃 − 𝜃 ) − 1 , ≈ 𝐺 .
AC 𝜔𝐿AC 0
(13) The validity conditions of (21) are
where 𝜃0 is the phase angle 𝜃 at the moment of alternating (
1 𝑟
)2
control structure. At the beginning and end of a period of 2 𝜔𝑙 Δ𝜃 ≪ 1,
applying a control structure, 𝑈dif,min ≫ 𝑟𝑖ref
dif
(22)
𝑉𝐶ref ≫ Δ𝑣𝐶,com (𝐺 + 𝑉DC ) + Δ𝑣𝐶,dif (𝐹 + 𝑉AC ) ,
𝑖com (𝜃0 + Δ𝜃) = 𝑖refcom (𝜃0 + Δ𝜃) − Δ𝑖com ,
(14) where 𝑈dif is actual magnitude of voltage to control the
𝑖com (𝜃0 ) = 𝑖ref
com (𝜃0 ) + Δ𝑖com ,
circulating currents and its minimum value is
where Δ𝑖com is the ripple magnitude of 𝑖com and Δ𝜃 is 𝜔Δ𝑡. 𝑈dif,min
By substituting (14) for 𝑖com (𝜃) and 𝑖com (𝜃0 ) into (13), and Δ𝑣𝐶,com (𝐺+𝑉DC )+Δ𝑣𝐶,dif (𝐹 +𝑉AC ) (23)
=𝐺− 𝑉𝐶ref
.
reorganizing the equation, (13) can be transformed into
( ) 𝑅 𝑖ref (𝜃 +Δ𝜃)−Δ𝑖 B. Three-Phase MMC
𝑅AC AC ( 0 com )+𝑈com
exp − 𝜔𝐿 Δ𝜃 = 𝑅ACcom (𝑖ref (𝜃0 )+Δ𝑖com )+𝑈com
. (15)
AC com In a three-phase MMC, the three-phase circulating currents
are independent. Therefore, 𝑔(⋅) and ℎfw,dif (⋅) of a three-phase
In (15), 𝑖ref
com is MMC are the same with those of a single-phase MMC in (7).
( ) However, dynamics of three-phase 𝑖𝑗com is different from that
𝑖ref ref ref
com (𝜃) = 𝐼com sin 𝜃 + 𝜙𝐼,com , (16)
of a single-phase MMC as the three-phase currents are not
ref
where 𝐼com and 𝜙ref
𝐼,com are the magnitude and phase independent. Therefore, 𝑓 (⋅) and ℎfw,com (⋅) of a three-phase
( ) angle
ref 𝑅AC MMC need to be modified based on the 𝑑𝑞 transformation by
of 𝑖com , respectively. Linearizing exp − 𝜔𝐿AC Δ𝜃 around
⎡ 𝑎 ⎤ ⎡ 𝑎 ⎤ ⎡ 𝑎 ⎤
Δ𝜃 ≈ 0 in the left side of (15) yields 𝑛dif 𝑓 (⋅) ℎfw,com
( ) ⎣ 𝑛𝑏dif ⎦ = 1 ⎣ 𝑓 𝑏 (⋅) ⎦ + 1 ⎣ ℎ𝑏fw,com ⎦ , (24)
𝑅AC 𝑅AC 𝑛𝑐dif 2𝑉𝐶ref 𝑓 𝑐 (⋅) 2𝑉𝐶ref
exp − Δ𝜃 ≈ 1 − Δ𝜃. (17) ℎ𝑐fw,com
𝜔𝐿AC 𝜔𝐿AC
( ) where
Linearizing sin 𝜃 + 𝜙ref ⎡ ⎤
𝐼,com in (16) around 𝜃 ≈ 𝜃0 yields 𝑓 𝑎 (⋅) [ 𝑞 ]
( ) ⎣ 𝑓 𝑏 (⋅) ⎦ = −𝑇 −1 𝑓 𝑑 (⋅) ,
𝑖ref ref ref ref
com (𝜃0 + Δ𝜃) ≈ 𝑖com (𝜃0 ) + 𝐼com cos 𝜃0 + 𝜙𝐼,com Δ𝜃. 𝑐 𝑓 (⋅)
(18) ⎡ 𝑓 𝑎 (⋅) ⎤ ⎡ 𝑎 ⎤
ℎfw,com 𝑣AC
Substituting (16), (17), and (18) in (15) and simplifying,
⎣ ℎ𝑏fw,com ⎦ = −2 ⎣ 𝑣AC𝑏 ⎦ (25)
relation among 𝐹 , Δ𝑡, and Δ𝑖com can be obtained as: 𝑐 𝑐
2Δ𝑖com
ℎfw,com ⎡ 𝑣AC ⎤⎡ 𝑎 ⎤
Δ𝑡 = √ 0 −1 1 𝑖com
cos(𝜃0 +𝜙ref
𝐼,com )+ 𝐿
𝑈com ref 𝑅AC ref
𝐿AC +𝜔𝐼com 𝑖com (𝜃0 ) AC ⎣
2𝐿AC Δ𝑖com
AC
(19) − 4 3𝜔𝐿
3
1 0 −1 ⎦ ⎣ 𝑖𝑏com ⎦ .
≈ 𝑈com
2𝐿AC Δ𝑖com
−1 1 0 𝑖𝑐com
≈ 𝐹 .
In (25), {
Considering the validity conditions of all the approximations 𝐹, 𝑖𝑞 > 𝑖𝑞,ref
𝑓 𝑞 (⋅) = 𝑞 𝑞,ref ,
{ −𝐹, 𝑖𝑑 < 𝑖𝑑,ref
from (10) to (19), the validity conditions of (19) are derived
(26)
as: 𝐹, 𝑖 > 𝑖
𝑓 𝑑 (⋅) = ,
1 3
−𝐹, 𝑖𝑑 < 𝑖𝑑,ref
6 Δ𝜃
≪ 1, (20a)
( )2 where 𝑖𝜉com is the 𝜉-axis component of three-phase 𝑖𝑗com (𝜉 =
1 𝑅AC
Δ𝜃 ≪ 1, (20b) 𝑞, 𝑑) and 𝑖𝜉,ref 𝜉
2 𝜔𝐿AC com is the reference of 𝑖com .
√
ref 2 2 Common- and differential-mode ripple components, i.e.,
𝑈com,min ≫ 𝐼com (𝜔𝐿AC ) + (𝑅AC ) , (20c) 𝑗
Δ𝑣𝐶,com 𝑗
and Δ𝑣𝐶,dif are significantly smaller than 𝑉𝐶ref in
𝑉𝐶ref a well-designed MMC. Therefore, based on (4) and (26), if
(20d)
≫ Δ𝑣𝐶,com (𝐹 + 𝑉AC ) + Δ𝑣𝐶,dif (𝐺 + 𝑉DC ), 𝑖𝑞com and 𝑖𝑑com are both larger than their references, i.e., 𝐼com𝑞,ref
𝑑,ref 𝑎
and 𝐼com , the dynamics of 𝑖com becomes
where (20a) and (20b) keep the validity of linearization from √ ( )
𝑑 𝑎 𝑅AC 𝑎 3
(15) to (19) and (20c) and (20d) support the simplification of 𝑑𝜃 𝑖com = − 𝜔𝐿AC 𝑖com + 3 −𝑖𝑏com + 𝑖𝑐com
𝐹 (27)
the denominator of (19). − 2𝜔𝐿 AC
(cos (𝜃) + sin (𝜃)) ,

1038
which can be decoupled into To prevent over-modulation under a step-up change of 𝑖𝑞,ref com
𝑞,ref,pre 𝑞,ref,new
from a previous reference, i.e., 𝐼com to 𝐼com , parame-
𝑑 𝑑 𝑅AC 𝑑 𝐹
𝑑𝜃 𝑖com = − 𝜔𝐿 𝑖
AC com
− 2𝜔𝐿AC , ters in the PI-based control method and the SMC-based control
𝑑 𝑞 𝑅AC 𝑞 𝐹 (28)
𝑑𝜃 𝑖com = − 𝜔𝐿 𝑖
AC com
− 2𝜔𝐿AC . method are constrained by the following upper limits:
𝑞
Solving for 𝑖𝜉com based on (28) and simplifying the results 𝐹max
√
2 2 ( 𝑞 𝑑,ref
)
based on approximations result yields ≈ (0.575𝑉DC ) − (𝐹 𝑑 ) − 𝑉AC − 𝜔𝐿AC 𝐼com ,
2𝜔𝐿AC Δ𝑖𝜉com
(36a)
Δ𝜃 ≈ 𝐹 . (29) √
𝑑,ref 2
(0.575𝑉DC )2 −(𝑉tmn ) −(𝑉AC
𝑞 𝑑,ref
−𝜔𝐿AC 𝐼com )
𝐾P,max ≈ 𝑞,ref,new
𝐼com 𝑞,ref,pre
−𝐼com
,
The validity conditions of (29) are
(36b)
( )2
1 𝑅AC
2 𝜔𝐿AC Δ𝜃 ≪ 1, where 𝐹 𝜉 is the magnitude of voltage to control 𝑖𝜉com in the
( 𝜉,ref ) 𝜉
𝐹 ≫ 𝑅AC 𝐼com + Δ𝑖𝜉com , (30) SMC-based control method, 𝑉AC is the 𝜉-axis component of
𝜉
𝑉𝐶ref ≫ Δ𝑣𝐶,dif , three-phase AC source voltages, and 𝑉tmn is 𝜉-axis component
𝜉
𝑉𝐶ref ≫ Δ𝑣𝐶,com . of AC-side terminal voltages of the MMC. The value of 𝑉tmn
is the ( )
C. Fixed-Frequency Comparison 𝑑,ref 𝑑,ref 𝑞,ref,pre
𝑉tmn = 𝑅AC 𝐼com + 𝜔𝐿AC 𝐼com . (37)
To limit the frequency variation of gating signals, 𝑖𝜉com and Based on (34), as 𝑖𝜉com reaches 𝐼com 𝜉,ref,new
, the zero-
𝑖𝑗dif are compared with their references at a fixed frequency, approaching speed of the error between 𝑖𝜉com and 𝐼com 𝜉,ref,new
i.e., 𝑓S,en . Since the fixed-frequency comparison could create decreases under the PI-based control method. On the other
a time-variable asymmetrical hysteresis range, the hysteresis hand, based on (35), under the proposed SMC-based control
range can be twice of those in (21) and (29). Therefore, method, the error decreasing speed is a constant, which is not
considering (21) and (29) as well as the variation of AC source impacted by the difference between 𝑖𝜉com and its reference.
voltages, the fixed frequency is designed based on Meanwhile, based on (36b), 𝐾P,max is inversely proportional
{ } to the step change magnitude, which also limits the speed
𝜔𝑉AC 𝐹 𝐺
𝑓S,en ≥ max , , . (31) of the dynamic response of the PI-based control with large
𝑉𝐶ref 𝐿AC Δ𝑖𝜉com 𝑙Δ𝑖𝑗dif 𝜉
Δ𝐼com . On the contrary, 𝐹max is not impacted by the step
change magnitude. Therefore, the proposed SMC-based con-
Subsequent to determination of 𝑓S,en , Δ𝑖𝜉com and Δ𝑖𝑗dif are trol can provide a faster dynamic response speed over the
conventional PI-based control.
Δ𝑖𝜉com = 𝑓S,en𝐹𝐿AC ,
(32) In addition, in the PI-based control, effective pole cancella-
Δ𝑖𝑗dif = 𝑓S,en
𝐺
𝑙. tion at 𝑠 = −𝑅AC /𝐿AC and prevention from over-modulation
both depend on exact parameters and precise model of the
The implementation block diagram of the proposed SMC- system, which are hard to be obtained in practice. However,
based control method is summarized in Fig. 2. in the proposed SMC-based control, alternation of control
structures is based on comparison between actively-controlled
D. Comparison with the Conventional PI-based Control Strat- variables and their references, which does not require precise
egy system parameters.
In the conventional PI-based control strategy, the integral Although a large 𝐹 can potentially increase the speed of
gain, i.e., 𝐾I is designed based on 𝑅AC , 𝐿AC , and the response during transients, the large excitation causes large
proportional gain, i.e., 𝐾P : high-frequency harmonics during steady state. To solve the
problem, an enhanced SMC-based control method with vari-
𝑅AC /𝐿AC = 𝐾I /𝐾P , (33) able 𝐹 is proposed:
{  
𝐹max , 𝑖𝜉com − 𝑖𝜉,ref  𝜉,thd
com  > 𝑖com
to cancel a pole at 𝑠 = −𝑅AC /𝐿AC [18]. Based on this 𝐹 =  𝜉,ref  , (38)
principle, zero-state response of 𝑖𝜉com for a step change of its 𝐹steady , 𝑖com − 𝑖com ≤ 𝑖𝜉,thd
𝜉
com

reference, i.e., 𝑖𝜉,ref 𝜉,ref,new

com from 0 to a new reference, i.e., 𝐼com where 𝐹max is maximum of 𝐹 , 𝐹steady is the magnitude of
𝜉
is illustrated in (34), while the response of 𝑖com under the voltage in steady state, and 𝑖𝜉,thd
com is a threshold to determine
proposed SMC-based control is illustrated in (35). the magnitude of excitation (𝐹max or 𝐹steady ). In (38), 𝑖𝜉,thd
com
( ( )) is designed by
𝜉 2𝐾P 𝑡 𝜉,ref,new 𝐹steady
𝑖com = 1 − exp − 𝐼com (34) 𝑖𝜉,thd
Δ
𝐿AC com = 𝜂 . (39)
𝐿AC 𝑓S,en

𝐹𝑡 Besides, 𝐹max is designed by (36a), 𝐹steady is designed based

𝑖𝜉com ≈ (35) on (32), and 𝜂 is slightly larger than 1.
2𝐿AC

1039
 j,ref
ncom np,j,ref
Equations (6b), (7b) & (8b) S/H +
+
vabc
AC

idifj idifj,ref Switching

Inner-Arm Signals
dq dq,ref fS,en Balance &
icom icom
Modulation
MMC
Equation F Equation Equation _+
(24) S/H + VDC _
(38) (26) ndifj,ref nn,j,ref
Fig. 2. Block diagram of the proposed SMC-based control method.

Since in the PI-based control method, the magnitude of TABLE I

ripple components during steady state is smaller, a hybrid PARAMETERS OF THE STUDY SYSTEM
control method based on combination of the SMC-based and Rating Value Value
the PI-based control methods is proposed, which is illustrated AC system nominal voltage 𝑉in,ll,rms 30 kV
in Fig. 3. In the hybrid control method, small steady state Nominal frequency of ac system 𝑓 60 Hz
DC link voltage 𝑉DC 60 kV
voltage ripple magnitude under the PI-based control method
Reference of SM capacitor voltage 𝑉𝐶ref 3 kV
and excellent dynamic response of the SMC-based control MMC Parameter Value
method are combined together. 𝐿 7 mH
𝑅 0.03 Ω
Series-connected inductance 𝑙 6 mH
start Resistance of series-connected inductor 𝑟 0.02 Ω
Number of SMs in each cluster 𝑁 20
ξ _ i ξ,ref |> i ξ,thd Submodule capacitor 𝐶 14000 𝜇F
|icom com com Control Parameter Value
Y N
𝐹 static in (32) 6 kV
𝐺 in (32) 6 kV
SMC-based PI-based 𝑓S,en in (32) 7.5 kHz
method method

end TABLE II
M AGNITUDE OF R IPPLE C OMPONENTS OF 𝑖𝜉com AND 𝑖𝑗dif
Fig. 3. Hybrid control of the MMC based on combination of PI- and SMC-
based strategies. 𝑖𝑑com (𝑖𝑑 /2) 𝑖𝑞com (𝑖𝑞 /2) 𝑖𝑎
dif
Simulation-based ∣Δ𝑖∣ (kA) 0.039 0.042 0.105
Theoretical ∣Δ𝑖∣ (kA) 0.04 0.04 0.133
IV. S IMULATION R ESULTS
To demonstrate performance and effectiveness of the pro-
posed control method, a three-phase MMC is simulated in (33) and (36). As illustrated in Fig. 6(a), the proposed SMC-
the PSCAD/EMTDC software environment. Parameters of the based control method can provide a faster dynamic response
study system are listed in Table I and the control system is than the conventional PI-based control method. In addition,
designed based on (31), (36), and (38). Steady-state waveforms Fig. 6(b) shows that 𝑖𝑑 remains to be decoupled from 𝑖𝑞 during
under the proposed SMC-based control method are provided the transient under the proposed SMC-based control method.
in Fig. 4, where the setpoints of real and reactive power are Therefore, the three-phase line currents are well-regulated, as
120 MW and 80 MVAr, respectively. Based on Figs. 4(a)- illustrated in Fig. 6(c). In addition, Figs. 6(d)-(f) show that
(c), three-phase line currents, circulating currents, and SM the currents of the six arms are controlled as expected, where
capacitor voltages are well regulated under the proposed the second-order components are suppressed. As illustrated in
method. Figs. 4(d)-(f) illustrate 𝑖𝜉com and 𝑖𝑗dif within one line Figs. 6(g)-(i), the average SM capacitor voltages of the six
cycle. The magnitudes of ripple components of 𝑖𝜉com and 𝑖𝑗dif arms are also well-regulated.
are closely matched with their theoretical values in (32), as
shown in Table II. Dynamic response of the MMC to a step V. E XPERIMENTAL V ERIFICATION
change of 𝑖𝑞,ref under the proposed SMC-based method is The proposed SMC-based control method is also verified
presented in Fig. 6, where 𝑖𝑞,ref steps up from 0 to 15 MW experimentally. The prototype is shown in Fig. 5 and the
at 𝑡 = 0.27 s and 𝑖𝑑,ref remains to be at -10 MVAr. Dynamic parameters are listed in Table III. In the experimental pro-
response of 𝑖𝑞 under the proposed SMC-based control method totype, the DC-side of the three-phase MMC is supplied
is compared with the one under PI-based control method in by a programmable DC power supply while its AC side is
Fig. 6(a), where the PI-based controller is designed based on connected to a three-phase RL load. The control method

1040
 5 0.4

iq (kA)
ij (kA)
0.2 PI−based
0
SMC
0
−5 0.269 0.27 0.271 0.272 0.273
0.38 0.39 (a) 0.4 0.41 0.42 (a)
2
ix,a (kA)

id (kA)
0.3
0 0.25
−2 0.2
0.38 0.39 (b) 0.4 0.41 0.42 0.269 0.27 0.271
(b) 0.272 0.273
3.2 1
v̂Cx,a (kV)

ij (kA)
3 0

2.8 −1
0.38 0.39 (c) 0.4 0.41 0.42 0.24 0.26 0.28 (c) 0.3 0.32
0.5
id (kA)

ix,a (kA)
2.2
0
2.1
−0.5
0.4 0.405 (d) 0.41 0.415 0.24 0.26 0.28 (d) 0.3 0.32
0.5
iq (kA)

ix,b (kA)
3.3
0
3.2
−0.5
0.4 0.405 (e) 0.41 0.415 0.24 0.26 0.28 (e) 0.3 0.32
0.8 0.5
iadif (kA)

ix,c (kA)
0.7
0
0.6
0.4 0.405 (f) 0.41 0.415 −0.5
0.24 0.26 0.28 (f) 0.3 0.32
3.1
v̂Cx,a (kV)

Fig. 4. Steady-state MMC waveforms under the proposed sliding mode 3.05
control: (a) three-phase line currents, (b) arm current of phase 𝑎, (c) SM 3
capacitor voltages of phase 𝑎, (d) 𝑑-axis component of line currents and its 2.95
reference, (e) 𝑞-axis component of line currents and its reference, and (f)
circulating current of phase 𝑎 and its reference. 0.24 0.26 0.28 (g) 0.3 0.32
3.2
v̂Cx,b (kV)

2.8
0.24 0.26 0.28 (h) 0.3 0.32
3.1
v̂Cx,c (kV)

3.05
3
2.95
0.24 0.26 0.28 (i) 0.3 0.32

Fig. 6. Response of the MMC to a step change of 𝑖𝑞,ref under the proposed
sliding mode control: (a,b) 𝑞- and 𝑑-axis components of the line currents and
their references, (c) three-phase line currents, (d-f) arm currents of phases 𝑎,
𝑏, and 𝑐, and (g-i) SM capacitor voltages of phases 𝑎, 𝑏, and 𝑐.

is implemented in the RT-Lab and OPAL-RT rapid control

prototyping tool.
Fig. 7 presents the experimental waveforms under the pro-
posed sliding mode control, where 𝑖𝑞,ref steps up from 1 A to
4 A at 𝑡 = 50 ms while 𝑖𝑑,ref remains to be zero. As illustrated
in Figs. 7(a)-(d), the three-phase line currents closely track
their references under the sliding mode control. Based on
Fig. 7(d), the rise time of 𝑖𝑞 is about 1.4 ms, reaching 3.7 A
(90 % the Δ𝐼 𝑞,ref ) without any overshoot. Figs. 7(e) and (f)
Fig. 5. Experiment Setup of the three-phase MMC prototype. present the arm currents and SM capacitor voltages of phase

1041
 10 ms/div 10 ms/div

(a) (a)
10 ms/div 10 ms/div

(b) (b)
2 ms/div 2 ms/div

(c) (c)
2 ms/div 2 ms/div

(d) (d)
20 ms/div 20 ms/div

(e) (e)
20 ms/div 20 ms/div

(f) (f)

Fig. 7. Experimental MMC waveforms under the proposed SMC-based Fig. 8. Experimental MMC waveforms under the conventional PI-based
control method: (a) three-phase line currents, (b) 𝑞- and 𝑑-axis components control method: (a) three-phase line currents, (b) 𝑞- and 𝑑-axis components
of the line currents and their references, (c) zoomed-in three-phase line of the line currents and their references, (c) zoomed-in three-phase line
currents, (d) zoomed-in 𝑞- and 𝑑-axis components of the line currents and currents, (d) zoomed-in 𝑞- and 𝑑-axis components of the line currents and
their references, (e) arm currents of phase 𝑎, and (f) SM capacitor voltages their references, (e) arm currents of phase 𝑎, and (f) SM capacitor voltages
of phase 𝑎. of phase 𝑎.

𝑎, which are well controlled during the transient. Fig. 8 SMC-based control method, control parameter relations and
presents the corresponding experimental waveforms under the their validity conditions are obtained. Compared with PI-
conventional PI-based control with the same scenario, where based control, the proposed control method is capable of
𝑖𝑞,ref steps up from 1 A to 4 A. As shown in Fig. 8(d), under providing a faster dynamic response without compromising the
the PI-based control, the rise time for 𝑖𝑞 is 2.2 ms, which is computational effort or requirement for precise system model.
57% longer than that under the proposed sliding mode control. Simulation and experimental results are presented to illustrate
This verifies the capability of the proposed SMC-based control the salient features of the proposed control method as opposed
method in providing a faster dynamic response over the PI- to the conventional PI-based control method.
based one.
VI. C ONCLUSION R EFERENCES

This paper proposes a sliding mode controller (SMC) for the [1] S. Debnath, J. Qin, B. Bahrani, M. Saeedifard, and P. Barbosa, “Oper-
MMC. Based on the analysis of the MMC under the proposed ation, control, and applications of the modular multilevel converter: A
review,” IEEE Trans. Power Electron., vol. 30, no. 1, pp. 37–53, Jan
2015.
TABLE III
PARAMETERS OF THE EXPERIMENTAL SETUP [2] M. A. Perez, S. Bernet, J. Rodriguez, S. Kouro, and R. Lizana, “Circuit
topologies, modeling, control schemes, and applications of modular
multilevel converters,” IEEE Trans. on Power Electron., vol. 30, no. 1,
Quantity Value
pp. 4–17, Jan 2015.
Nominal frequency 60 Hz
[3] S. Allebrod, R. Hamerski, and R. Marquardt, “New transformerless,
DC source voltage 𝑉DC 100 V
scalable modular multilevel converters for HVDC-transmission,” in
Load 𝑅𝑙𝑜𝑎𝑑 6.7 Ω
Proc. IEEE PESC, June 2008, pp. 174–179.
Load 𝐿𝑙𝑜𝑎𝑑 4.5 mH
[4] J. Dorn, H. Huang, and D. Retzmann, “A new multilevel voltage-sourced
Number of SMs per arm 𝑁 4
converter topology for HVDC applications,” in Cigre Session, 2008, pp.
Buffer inductance 𝑙 2.5 mH
B4–304.
Submodule capacitor 𝐶 1410 𝜇F
Switching frequency 𝑓sw 8 kHz [5] K. Sharifabadi, L. Harnefors, H.-P. Nee, S. Norrga, and R. Teodorescu,
Dynamics and Control. Wiley-IEEE Press, 2016, pp. 133–213.

1042
 [6] R. Teodorescu, F. Blaabjerg, M. Liserre, and P. C. Loh, “Proportional-
resonant controllers and filters for grid-connected voltage-source con-
verters,” Proc. IEE EPA, vol. 153, no. 5, pp. 750–762, September 2006.
[7] Q. Tu, Z. Xu, and L. Xu, “Reduced switching-frequency modulation
and circulating current suppression for modular multilevel converters,”
IEEE Trans. on Power Del., vol. 26, no. 3, pp. 2009–2017, July 2011.
[8] G. Konstantinou, J. Pou, S. Ceballos, R. Picas, J. Zaragoza, and
V. G. Agelidis, “Control of circulating currents in modular multilevel
converters through redundant voltage levels,” IEEE Trans. on Power
Electron., vol. 31, no. 11, pp. 7761–7769, Nov 2016.
[9] X. Shi, Z. Wang, B. Liu, Y. Liu, L. M. Tolbert, and F. Wang, “Char-
acteristic investigation and control of a modular multilevel converter-
based HVDC system under single-line-to-ground fault conditions,”
IEEE Trans. on Power Electron., vol. 30, no. 1, pp. 408–421, Jan 2015.
[10] X. She, A. Huang, X. Ni, and R. Burgos, “Ac circulating currents
suppression in modular multilevel converter,” in Proc. IEEE IECON,
Oct 2012, pp. 191–196.
[11] J. Qin and M. Saeedifard, “Predictive control of a modular multilevel
converter for a back-to-back HVDC system,” IEEE Trans. on Power
Del., vol. 27, no. 3, pp. 1538–1547, July 2012.
[12] M. Vatani, B. Bahrani, M. Saeedifard, and M. Hovd, “Indirect finite
control set model predictive control of modular multilevel converters,”
IEEE Trans. on Smart Grid, vol. 6, no. 3, pp. 1520–1529, May 2015.
[13] B. S. Riar, T. Geyer, and U. K. Madawala, “Model predictive direct
current control of modular multilevel converters: Modeling, analysis,
and experimental evaluation,” IEEE Trans. on Power Electron., vol. 30,
no. 1, pp. 431–439, Jan 2015.
[14] L. Ben-Brahim, A. Gastli, M. Trabelsi, K. A. Ghazi, M. Houchati, and
H. Abu-Rub, “Modular multilevel converter circulating current reduction
using model predictive control,” IEEE Trans. on Ind. Electron., vol. 63,
no. 6, pp. 3857–3866, June 2016.
[15] H. Brnklau, A. Gensior, and J. Rudolph, “A model-based control scheme
for modular multilevel converters,” IEEE Trans. on Ind. Electron.,
vol. 60, no. 12, pp. 5359–5375, Dec 2013.
[16] L. Harnefors, A. Antonopoulos, S. Norrga, L. Angquist, and H. P. Nee,
“Dynamic analysis of modular multilevel converters,” IEEE Trans. on
Ind. Electron., vol. 60, no. 7, pp. 2526–2537, July 2013.
[17] D. C. Ludois and G. Venkataramanan, “Simplified terminal behavioral
model for a modular multilevel converter,” IEEE Trans. on Power
Electron., vol. 29, no. 4, pp. 1622–1631, April 2014.
[18] A. Yazdani and R. Iravani, Grid-Imposed Frequency VSC System:
Control in dq-Frame. Wiley-IEEE Press, 2010, pp. 204–244.

1043