IET Renewable Power Generation

Research Article

ISSN 1752-1416
Application of modular multilevel converter Received on 29th September 2015
Revised on 1st March 2016
in medium voltage high power permanent Accepted on 15th March 2016
doi: 10.1049/iet-rpg.2015.0444
magnet synchronous generator wind energy www.ietdl.org

conversion systems
Mian Wang ✉, Yaowei Hu, Wenjian Zhao, Yue Wang, Guozhu Chen
College of Electrical Engineering, Zhejiang University, Teaching Building 2, No. 38 Zheda Road, Hangzhou, People’s Republic of China
✉ E-mail: [email protected]

Abstract: Traditional low voltage high power wind energy conversion system (WECS) requires heavy cables and large
step-up grid interfacing transformers. In recent years, power rating of offshore wind turbines already exceeds 10 MW,
and medium voltage (MV) converters have become strongly demanded. Modular multilevel converter (MMC) is
suitable for MV high power applications. However, there are two technical difficulties for MMC when applied in MV
permanent magnet synchronous generator (PMSG) WECS. The first one is the large sub module (SM) voltage
fluctuation caused by low frequency and high amplitude PMSG phase current. The other one is the low voltage ride
through (LVRT) problem. Overall control strategy based on carrier-phase-shifted -sinusoidal pulse width modulation
technique of the MMC WECS is proposed. For the first difficulty mentioned, a second-order circulating current injection
method based on closed-loop control is proposed. For the LVRT problem, a distributed braking chopper solution is
provided. The effectiveness of the injection method and the distributed braking chopper solution are validated by
simulation of a 5 MW/3.3 kV MMC WECS. Influences of the current injection amplitude on SM voltages and arm
currents are discussed and a trade-off method is given to optimise system design, which is also verified by simulation.

1 Introduction cable length, 5L-HB suffers extra cost, power losses and cable
inductance [2]. In a word, NPC type converters cannot handle all
With the rapid consumption of fossil fuels and common concerns for MV high power WECS, especially for voltage ratings higher than
environmental problems, various renewable energy sources are 4 kV.
considered as substitutions for traditional sources. Wind energy is On the other hand, promising MMCC contains the following
one of the fastest growing renewable sources in recent years [1–3]. topologies: single-star bridge-cells (SSBC), single-delta
Low speed direct-drive permanent magnetic synchronous bridge-cells (SDBC), double-star chopper-cells (DSCC) and
generators (DD-PMSGs) with full-scale power converters have double-star bridge-cells [11]. SSBC and SDBC are not suitable for
been widely adopted in multi-megawatt wind energy conversion WECS because they do not have a common dc link. The
systems (WECS) [4, 5]. Traditional multi-megawatt low voltage DSCC-MMCC, which is usually simplified as modular multilevel
DD-PMSG requires heavy cables and large step-up grid interfacing converter (MMC), is a newly introduced topology [14]. MMC is
transformers, which leads to high cost and low efficiency. Medium widely considered as the suitable converter of high-voltage direct
voltage (MV) WECS has been popular because it omits the current (HVDC), MV motor drive, reactive power compensation
step-up transformer and reduces the current rating of the generator and renewable power generation [15, 16]. Its advantages include:
and converter [6, 7]. This trend leads to the research of new (i) low output harmonics, (ii) high modularity and extend-ability,
full-scale power converters [1, 8, 9]. Multilevel converter is of (iii) suitable for all voltage rating with regular power devices and
lower harmonics and electromagnetic interference noises compared (iv) high equivalent switching frequency. However, there are some
with two-level converter. Regular power devices with low current technical difficulties for the application of MMC in MV
ratings can be adopted in multilevel converters. Therefore, PMSG-based WECS. First, the PMSG phase currents cause large
multilevel converter is the most promising topology for the grid voltage fluctuation in sub module (SM) capacitors [17, 18],
connection of large MV wind turbines, especially for offshore because the fluctuation magnitude is proportional to PMSG current
wind turbines [10]. magnitude while inversely proportional to the current frequency.
MV multilevel converters can be basically divided into two types, Large voltage fluctuation put up high requirement for power
namely neutral point clamped (NPC) converters and modular devices and capacitance, which is not acceptable for WECS
multilevel cascade converters (MMCCs) [11]. In back-to-back application as the space for the converter in a wind tower is
WECS, promising NPC topologies include three-level NPC limited. The other technical difficulty is the low voltage ride
converter (3L-NPC), three-level active NPC converter (3L-ANPC), through (LVRT) capability. Traditional LVRT schemes [5] based
nested NPC converter (NNPC) [8], five-level ANPC converter on dc-link chopper is no longer suitable in back-to-back MMCs
(5L-ANPC) [12], five-level H-bridge converter (5L-HB) and the because of the instant voltage problem.
combination of these topologies [2, 13]. On the basis of existing Few researches have been done to solve the SM voltage
high power devices with a maximum voltage rating of 6.5 kV, fluctuation problem in MMC WECS. Possible solutions have been
3L-NPC and 3L-ANPC are only suitable for wind turbines with proposed in MV motor drive systems [17–24] because the start-up
rated line-to-line voltage lower than 4 kV. NNPC, 5L-ANPC and procedure of MV motor suffers the same problem. All the
other similar topologies have the complexity and instability proposed techniques are based on common mode voltage and
problem since the floating capacitors should be pre-charged and circulating current injection, which aim to attenuate the fluctuation
their voltage should be kept stable through closed-loop control. without laying impact on the ac side of the converter. The
Meanwhile, in the need of open winding transformers and doubled techniques can be basically divided into open-loop injection

IET Renew. Power Gener., pp. 1–10

& The Institution of Engineering and Technology 2016 1
method [21, 23] and closed-loop injection method [17, 19, 20, 24]. 2 Modelling of back-to-back MMC-based WECS
Open-loop injection method estimates the circulating current needed
to eliminate the second-order harmonic of the SM voltage ripple. Its The circuit diagram of back-to-back MMC-based PMSG WECS is
drawback is that there are steady-state estimation errors. Closed-loop shown in Fig. 1. Generator side MMC and grid side MMC are of
injection method can eliminate the steady-state error through symmetrical structure. Each MMC consists of two arms per leg
circulating current feedback control. In [17], a sine-wave-based and each arm contains N cascaded half-bridge SMs. CSM is the
common mode voltage and circulating current injection method is SM capacitor. The upper arm (‘p’) and lower arm (‘n’) are
proposed but detailed control scheme is not given. In [19], a connected by two arm inductors L which absorbs the high-order
square-wave-based injection method is proposed. Theoretically, the harmonics of arm current and limits the peak value of phase
square-wave method can reduce the amplitude of circulating current. ac Side nodes are connected in the midpoint of two arm
current injected. However, it is hard to be realised because the inductors. Normally all the SMs are identical. Common dc-link
zero crossing point of the square-wave requires infinite instant capacitor is not needed because the SM capacitors are in
circulating current. In [24], two modified strategies are proposed substitution. Udc is the dc-link voltage and O is the virtual
based on the sine-wave injection method which aims to decrease midpoint of Udc. Ucjpx, Ucjnx are the capacitor voltages of SMjpx
the amplitude of the injection circulating current. However, the and SMjnx respectively, where phase j ∈ (phase u, v, w, a, b, c)
modified injection circulating current could result in over and x is an integer (1 ≤ x ≤ N ). ij is the phase current and uj is the
modulation. phase voltage. ipj and inj are the arm currents. izj is the circulating
As to the LVRT capability, the braking chopper solution of current. upj and unj are the cascaded module voltages of upper and
back-to-back MMC is discussed in [25, 26]. Conventional braking lower arms, respectively.
chopper in MV and high voltage systems consists of a number of
series connected devices and braking resistors, which is quite
difficult to realise. In [26], the authors propose a new chopper 2.1 PMSG modelling
based on cascaded half-bridges to eliminate the problem of device
On the basis of the synchronous rotating frame transformation,
series connection, but it requires additional control of the
excitation flux component Ψ and electrical magnetic torque
half-bridge capacitor voltages.
component Tem are decoupled. The flux and torque model is given
This paper focuses on the two critical problems in the application
in [5], where magnetic saturation, eddy loss and slot effect and so
of MMC in DD-PMSG WECS. On the basis of [17] a new
on of the generator are ignored
closed-loop circulating current injection method combined with
carrier-phase-shifted sinusoidal pulse width modulation
(CPS-SPWM) technique is proposed. The fluctuation of SM 3
Tem = p (C i − Cq id ) (1)
voltage is suppressed through a given negative q-axis reference of 2 o dq
circulating current in the negative sequence frame. The relation

between reference amplitude and the magnitude suppression is Cd = Ld id + Cf
(2)
discussed. For the LVRT solution, this paper proposes a new Cq = Lq iq
distributed braking chopper strategy. Braking choppers are
paralleled with all the SM capacitors to help riding through grid where, po is the number of pole-pairs. Ψd and Ψq are the dq flux
faults. Simulation model of a 5 MW–3.3 kV PMSG and the components. Ψf is the rotor flux, which is a constant. id and iq are
back-to-back MMC is built in the MATLAB/SIMULINK tool to the dq component of phase current. Torque component Tem can be
verify the correctness and effectiveness of the proposed strategies. regulated by iq with id = 0 control strategy [5].

Fig. 1 Circuit diagram of back-to-back MMC-based PMSG wind power system

IET Renew. Power Gener., pp. 1–10

2 & The Institution of Engineering and Technology 2016
 ⎧
2.2 MMC modelling ⎪ DE pj = D p pj dt
⎪
⎪
   
⎪
⎪ ij L dij
⎪
⎪ Udc L dij
Under symmetrical load and source, each phase of MMC works ⎪
⎨ = 2 2 dt − uj + 2 + Az 2 dt − uj dt
independently. Therefore, modelling three-phase MMC can be (10)
simplified as single phase. For example, assuming the voltage and ⎪
⎪ DEnj = D pnj dt
⎪
⎪
   
current of phase j are ⎪
⎪ ij L dij L dij
⎪
⎪= U

⎩ − uj − dc − Az − uj dt
uj = AU sin (vt + wu ) 2 2 dt 2 2 dt
(3)
ij = AI sin vt
where the harmonics of izj and common mode voltage uO′′ O are
AU and AI are the amplitude of uj and ij. ω is the angle frequency and j ignored. Az is the dc component of izj. Δppj and Δpnj are the ac
is the phase angle. According to the Kirchhoff’s Current Law (KCL) fluctuation of branch powers. ΔEpj and ΔEnj are the ac fluctuation

of branch energy. It can be concluded from (3) and (10) that the
i pj = izj + i psj peak-to-peak amplitude of the branch energy is proportional to
(4)
inj = izj − insj phase current magnitude AI while inversely proportional to the
current frequency. As the DD-PMSG operates at very low
ij mechanical and electrical angular speed, SM voltages of the
= i psj = insj (5)
2 generator side MMC suffers critical fluctuation.

where ipsj and insj are the fundamental component of arm current. ipsj =
insj because of the symmetrical structure of phase legs 3 Control strategy
⎧
⎨ ij = i pj − inj The CPS-SPWM [27] method modulates each SM separately and
1 (6) produces more voltage level number than other modulation
⎩ izj = (inj + i pj )
2 methods, realising lower harmonics [28]. By comparing
modulation voltage of each SM with its triangular carrier, the
vj ij A A
izj = = U A cos wu − U AI cos (2vt + wu ) (7) PWM signals can be generated. The cascaded number of SMs is
Udc 2Udc I 2Udc small in MV WECS compared with other high voltage
applications, e.g. HVDC. Therefore the well-known drawback of
Equation (7) shows that circulating current of phase j consists of dc CPS-SPWM, not appropriate for MMCs with big number of SMs,
component and second-order harmonic component. is no longer a concern. The CPS-SPWM is a very appropriate
According to the Kirchhoff’s Voltage Law (KVL) in phase j modulation technique in MV wind power applications. The overall
⎧ control diagram of the full power back-to-back MMC WECS
⎪
⎪ U di based on CPS-SPWM is shown in Fig. 2.
⎨ u pj = dc + L pj − uj − uO′′ O
2 dt Parameters in Fig. 2 are listed in Table 1, where j ∈ (phase u, v, w,
(8)
⎪
⎪ Udc dinj a, b, c) and x is the module number (1 ≤ x ≤ N). The upper and lower
⎩ unj = +L + uj + uO′′ O arms are represented by subscript p and n, respectively. PI1–PI7 are
2 dt
proportional–integral regulators and P1, P2 are proportional
Hence the branch powers ppj and pnj are regulators.
The PMSG-side MMC is under Maximum power point tracking
⎧    (MPPT) control, which determines the active power flow in the
⎪ Udc di pj ij
⎪
⎨ pj
p = u i
pj pj = + L − u j − u ′′
O O i zj + back-to-back system. Pgmax can be calculated by detecting ωg at
2 dt 2
   (9) certain β and v. The MPPT controller tracks Pgmax through
⎪
⎪ Udc di ij
⎩ pnj = unj inj = +L
nj
+ uj + uO′′ O izj − feedback control of the grid-side active power Ps.
2 dt 2 The grid-side MMC maintains dc-link voltage stable via the
dc-link controller. Furthermore, reactive power injected into the
Equation (9) reveals that the dominant fluctuation of branch energy is grid is regulated by grid-side MMC too.

Fig. 2 Overall control diagram of the full-power back-to-back MMC PMSG wind power system

IET Renew. Power Gener., pp. 1–10

& The Institution of Engineering and Technology 2016 3
Table 1 Parameters of the overall control diagram
Parameters Description Controller

Pgmax maximum power point of PMSG MPPT controller

Ps active power of grid-side MMC
θg electrical rotating angle of PMSG
ωg electrical angle speed of PMSG
po pole-pairs of PMSG
ß pitch angle
v wind speed
iuvw, uuvw current and voltage of phase u, v, w phase current controller of PMSG-side MMC
igd, igq dq-axis currents of iuvw
igdref, igqref references of igd, igq
vu, vv, vw modulation voltages of phase u, v, w current controller
ugd, ugq dq-axis voltages of uuvw
izuvw circulating currents of phase u, v, w circulating current controller of PMSG-side MMC
igzd, igzq dq-axis circulating currents of izuvw
igzdref, igzqref references of igzdref, igzqref
vzu, vzv, vzw modulation voltages of circulating current controller of phase u, v, w
ipuvw, inuvw upper ( p) and lower (n) arms current of phase u, v, w SM dc voltage balancing and averaging controller
ipabc, inabc upper ( p) and lower (n) arms current of phase a, b, c
ucuvwpx, ucuvwnx xth SM voltage in upper ( p) and lower (n) arms of phase u, v, w
ucabcpx, ucabcnx xth SM voltage in upper ( p) and lower (n) arms of phase a, b, c
ucjpxref, ucjnxref reference voltages of ucjpx, ucjnx
vcjpx, vcjnx modulation voltages of SM voltage balancing controller
ucjavg average SM voltage of phase j
ucjref reference voltage of ucjavg
vjavg modulation voltage of SM voltage averaging controller of phase j
Udcref reference voltage of Udc dc-link voltage controller
iabc, uabc currents and voltages of phase a, b, c phase current controller of grid-side MMC
θs electrical rotating angle of grid voltage
isd, isq dq-axis currents of iabc
isdref, isqref references of isd, isq
va, vb, vc modulation voltages of phase a, b, c current controller
usd, usq dq-axis voltages of uabc
izabc circulating currents of phase a, b, c circulating current controller of grid-side MMC
iszd, iszq dq-axis circulating currents of izabc
iszdref, iszqref references of iszd, iszq
vza, vzb, vzc modulation voltages of circulating current controller of phase a, b, c
mjpx, mjnx modulation voltages of upper ( p) and lower (n) arms of phase j modulation
PWMuvw, pnx PWM signals of the xth SM in upper ( p) and lower (n) arms of phase u, v, w
PWMabc, pnx PWM signals of the xth SM in upper ( p) and lower (n) arms of phase a, b, c

The phase current controller and circulating current controller of where m is the modulation index. Uvj and j are the amplitude and
two MMCs are in the same structure. However, they have different phase angle of vj, respectively.
references and phase locked loops (PLLs). According to (7) the
main ac component of the circulating current is the second-order
harmonic. As a result the circulating current is controlled in the 4 Proposed circulating current injection method
second-order synchronous frame in order to suppress the
corresponding ac component. The peak-to-peak amplitude of the phase branch energy is
Besides, the SM voltage controllers, including voltage balancing proportional to phase current magnitude AI and inversely
and voltage averaging controller, and the CPS-SPWM modulation proportional to the current frequency f ( fg for PMSG-side MMC
technique are shared by two MMCs in Fig. 3. The voltage and fs for grid-side MMC) according to (10). For low speed
balancing controller adjusts each individual SM voltage and the DD-PMSG the phase frequency fg = ωg/2π varies from several
voltage averaging controller aims at averaging SM voltage in each hertz to several hundred hertz. The rated fg is generally lower than
phase. 20 Hz, which means PMSG-side MMC has to operate at full-rated
According to Fig. 2, the modulation voltages of upper ( p) and current with very low frequency. As a result, SM voltages of the
lower (n) arms of phase j are PMSG-side MMC suffer critical ac fluctuation. In [17], the
⎧
 sine-wave-based circulating current injection method is proposed,
⎪ 1 2(vj + vzj − Nvcjpx − Nv javg )
⎪
⎨ jpx 2
m = 1 − but description of detailed control algorithm, the modulation
Udc technique and the mathematical expression of injected current is

 (11)
⎪
⎪ 1 2(vj − vzj + Nvcjnx + Nv javg ) not given. This article proposes a new detection method based on
⎩ m jnx = 1 + second-order harmonic circulating current injection. On the basis
2 Udc
of negative synchronous rotating frame transformation with a
The modulation voltage of the circulating current controller and SM rotating angle of −2θg, a negative q-axis reference is given in the
voltage controller, namely vzj, vcjpx, vcjnx and vjavg, are relatively circulating current controller. In this way, additional second-order
small compared with vj, the output of phase current controller. circulating current is injected to the arm current, which can
Therefore, the modulation voltage of upper and lower arms, suppress the SM voltage fluctuation and minimise the SM
namely mjp and mjn, can be simplified as capacitance.
According to (6) and (12), when there is no circulating current
⎧   injection, the upper arm SM voltages are
⎪ 1 2vj 1
⎪
⎪ jp
m = 1 − = [1 − m sin (vt + w)]
⎪
⎪ 2 U 2  
⎪
⎪  dc
 dUcjpx 1 1
⎨ 1 2vj 1 CSM = i pj m jpx = [1 − m sin (vg t + w)] Az + AI sin vg t
m jn = 1+ = [1 + m sin (vt + w)] (12) dt 2 2
⎪
⎪ 2 U 2
⎪
⎪
dc (13)
⎪
⎪ 2Uvj
⎪
⎩m =
Udc Thus, the ac component of Ucjpx is

IET Renew. Power Gener., pp. 1–10

4 & The Institution of Engineering and Technology 2016
 Equation (17) shows the minimum peak-to-peak value of ac
fundamental SM voltage when no circulating current is injected.
When second-order circulating current is injected

izj injected = kAI cos 2vg t (18)

where k is the injection coefficient and −kAI is the q-axis reference of

circulating current controller. The upper arm SM voltages are

dUcjpx 1
CSM = [1 − m sin (vg t + w)]
dt 2 
1
× Iz + kAI cos 2vg t + AI sin vg t (19)
2

It can be conducted from (19) that

dUcjpx
CSM = A sin vg t + B cos vg t
dt
+ C sin 2vg t + D cos 2vg t + E sin (3vg t + w) (20)
⎧
⎪
⎪ 2A + 2km cos wAI − m2 AI cos2 w
⎪
⎪ A= I
⎪
⎪ 8
⎪
⎪
⎪
⎪ 2km sin wA + m2 AI cos w sin w
⎪
⎪ B=− I
⎪
⎪
⎨ 8
mAI sin w (21)
⎪ C=−
⎪
⎪ 4
⎪
⎪ +
⎪
⎪ 4kA mAI cos w
⎪
⎪ D= I
⎪
⎪ 4
⎪
⎪
⎪
⎩ E = kmAI
2

where A, B, C, D and E are relevant coefficients of the fundamental,

second-order and third-order harmonics. The fundamental
component is the dominant component. The decrease of
fundamental voltage leads to the decrease of fluctuation
magnitude. The fundamental voltage Ucjpx_1 is

dUcjpx 1

CSM = A sin vg t + B cos vg t = A2 + B2 sin (vg t + g)
dt
(22)

where

Fig. 3 Braking chopper solution in back-to-back MMCs 1

a PWM signal generation method A2 + B2 = [4m2 k 2 + (8mAI cos w − 4m3 AI cos w cos 2w)k
b Single phase circuit of conventional braking chopper solution
64
c Proposed distributed braking chopper solution + (2AI − m2 cos2 wAI )2 + m4 A2I cos2 wsin2 w]
11 mAI (23)
Ũ cjpx = cos (vg t + w1 ) + sin (2vg t + w) (14)
CSM vg 8CSM vg
A 2 + B 2 is a function of k whose√curve is an upward opening

parabola, so the minimum value of A2 + B2 is
where
  AI sin w 

AI 4 + m4 cos2 w − 4m2 cos2 w A2 + B2min = App
1 min = 4 + 2m4 cos2 w − 8m2 cos2 w
11 = (15) 32CSM vg
8
(24)
and
When coefficient k is
m2 cos w sin w
sin w1 = −  (16)
4 + m4 cos2 w − 4m2 cos2 w 8 cos w − 4m2 cos w cos 2w
k = k0 = − (25)
8m
According to (14) and (15), the minimum peak-to-peak value of
fundamental SM voltage is Equation (24) shows the minimum peak-to-peak value of
fundamental SM voltage fluctuation when second-order circulating
 current k0AIcos2ωgt is injected. Compared with (17), the
AI 4 + m4 cos2 w − 4m2 cos2 w
App
1 min = (17) peak-to-peak value of fundamental SM voltage has been greatly
4CSM vg decreased. Thus, the SM capacitor can be minimised with a

IET Renew. Power Gener., pp. 1–10

& The Institution of Engineering and Technology 2016 5
circulating current injection whose q-axis reference igzqref is −k0AI at Table 2 Parameters of PMSG
the negative sequence frame.
Quantity Value
Though arm currents increases when the second-order harmonics
is injected, it is quite worthy for MV high power wind generation to nominal power Pg 5000 kW
adopt the strategy, especially for offshore wind applications. As in rated angular speed ωg 1.1 rad/s
offshore wind applications, the most critical problem for MMC cut-in angular speed ωg0 0.5 rad/s
topology is the large volume of the SMs which is mainly decided rated frequency fg 16.8 Hz
pole-pairs po 96
by the SM capacitance. By regulating coefficient k, PMSG-side rated line-line voltage ugLL 3300 V
MMC could reach a trade-off between SM voltage fluctuation and rotor flux Ψf 25.518 Wb
arm current increment.

5 LVRT strategy based on distributed braking Table 3 Parameters of back-to-back MMCs

choppers Quantity Value

Braking chopper of the dc link is important for back-to-back power nominal power Ps 5500 kVA
conversion systems. For example, in WECS when grid voltage dip rated grid frequency fs 50 Hz
occurs, the braking chopper can absorb redundant wind energy arm inductor L 3 mH
nominal SM voltage 1000 V
and help the converters riding through fault conditions. Due to the SM capacitance 10 mF
high dc-link voltage level, conventional braking chopper in MV nominal dc-link voltage 6000 V
and high voltage systems consists of a number of series connected number of SMs per arm N 6
devices and braking resistors, which is quite difficult to realise. rated grid line-line voltage usLL 3300 V
carrier frequency fc 1 kHz
Furthermore conventional chopper would cause high instant
voltage peak on the arm inductor in back-to-back MMCs. In [26],
the authors propose a new chopper based on cascaded half-bridges
to eliminate the series connection problem of the devices. It can 6.1 Simulation under normal condition when k = 0
eliminate the voltage peak on arm inductors, but it requires
additional control of the half-bridge capacitor voltages. This paper Under normal condition all the choppers are disabled. When
proposes a new distributed braking chopper strategy for the amplitude coefficient of injected circulating current k = 0, there is
back-to-back MMCs. Braking choppers are distributed into all the no circulating current injection. The power increasing process and
SMs to help riding through grid faults. steady-state operation under rated power of the system (Fig. 1) are
Braking chopper solution and the pulse-width modulation (PWM) set as the normal simulation condition. PMSG works at cut-in
signal generation method in back-to-back MMCs is shown in Fig. 3. angular speed from t = 0 s to t = 0.1 s. Ps is close to zero during
The devices of conventional solution and proposed solution share the
same PWM signal generated in Fig. 3a. Where, PI8 is a
proportional–integral regulator. ENABLE is the enable signal of
all SM braking choppers. ENABLE is set when needed.
In Fig. 3b, when the devices switch on or switch off, the current of
the arm inductor changes suddenly, resulting in an instant voltage
peak on the dc link. Conventional chopper solution cannot help
the system during fault conditions.
The diagram of the proposed distributed braking chopper solution
is shown in Fig. 3c. ISM is the chopper current. All the SMs with
distributed choppers share the PWM signal generated by the
dc-link voltage closed-loop control in Fig. 3a. RSM is the braking
resistor of each SM. y is the number of SMs which are paralleled
with distributed braking choppers.
To ride through zero voltage dip, the value of RSM can be
calculated by

u2cjref
RSM = y (26)
Ps rated

where Ps_rated is the rated active power of the grid-side MMC.

For the case of 5 MW WECS, the number y should not be <2N in
order to ride through faults as well as maintain the SM voltage stable.
If y is <2N, the SM voltages would suffer a large drop and the current
on the braking choppers is excessive for regular devices.

6 Simulation results and analysis

Simulation of a 5 MW/3.3 kV back-to-back MMCs PMSG wind

conversion system (as shown in Fig. 1) is carried out in
MATLAB/SIMULINK. The aforementioned circulating current
injection method and distributed braking chopper solution are
studied. The simulation results provide support to the stability of
the overall control strategy, the effectiveness of the injection Fig. 4 Simulation under normal condition
method and the capability of LVRT. Parameters of PMSG and a Grid-side MMC active and reactive powers
MMCs are listed in Tables 2 and 3. Normal condition and grid b Line-to-line voltage of PMSG
fault condition are set to verify the proposed strategies. c PMSG phase current

IET Renew. Power Gener., pp. 1–10

6 & The Institution of Engineering and Technology 2016
this period. From t = 0.1 s to t = 0.6 s ωg increases linearly and Ps When k = k0, the SM voltage, circulating current and arm currents
increases from zero to the rated power. After t = 0.6 s the system of phase u are shown in Fig. 6. Second-order circulating current is
works at rated condition. introduced as shown in Figs. 6c and d. The SM voltage fluctuation
The power injected to the grid and the power generation of PMSG is suppressed, with an rms value of 108.1 V.
are shown in Fig. 4. From t = 0.1 s to t = 0.6 s Ps rises and reaches 5 Table 4 shows the comparison of SM voltage when k varies.
MW shortly after t = 0.6 s. The MPPT strategy is verified. Reactive When k increases, the arm current rms value increases linearly.
power Qs remains zero because reference current isqref of the Meanwhile the SM voltage peak-to-peak value decreases, but the
grid-side MMC phase current controller is zero. The PMSG output ac SM voltage rms value does not strictly obey the declining rule.
voltage and current maintained stable after t = 0.6 s. The trade-off coefficient η1 and η2 is defined as
When k = 0, which means no circulating current is injected, the
SM voltage, circulating current and arm currents of phase u on the ⎧
⎪ dk
PMSG-side MMC is shown in Fig. 5. ⎪
⎨ h1 = d
Figs. 5a and b validate the mathematical analysis of the SM 1
(27)
voltage in (14). Fundamental component (16.8 Hz) and ⎪
⎪ dk
⎩ h2 =
second-order component are the main ac fluctuation sources. The d2
fundamental component is much larger than the second-order one.
The root-mean-square (rms) value of the ac fluctuation is 186.5 V Coefficient η1 and η2 represent the suppression effect of SM voltage
and the fundamental voltage rms value is 174.1 V, which lead to fluctuation with respect to the arm current increment. The larger η1
high voltage stress on the SM capacitor and devices. Fig. 5c and η2 are, the larger arm current increment is needed at the same
shows that circulating current contains only dc component when k SM voltage decrease. δ1, δ2, η1 and η2 offer different criteria to
= 0. choose the optimum k value. In order to reach a trade-off between
Figs. 4 and 5 validate the correctness of the proposed overall SM voltage fluctuation and arm current increment, the value of k
control strategy. The SM voltage ac fluctuation problem described should vary from case to case according to the specific system
by (14) is also verified by Fig. 5a. design constraints. Here, for example, if the voltage peak-to-peak

6.2 Simulation under normal condition when k varies

Equation (23) reveals that when the coefficient k increases from zero
to k0, the amplitude of the SM voltage fluctuation decreases. In this
part, simulation under normal conditions when k = 0.2k0, 0.4k0,
0.6k0, 0.8k0 and k0 are carried out (in this model k0 = 0.5). The
wind system power flow, PMSG phase current and line-to-line
voltage are the same as those in Fig. 4 in these simulations.
Injection of second-order circulating current does not introduce
influence to the PMSG phase current and line-to-line voltage.

Fig. 6 Simulation under normal condition when k = k0

Fig. 5 Simulation under normal conditions when k = 0 a Upper arm SM voltages ucup of phase u
a Upper arm SM voltages ucup of phase u b FFT analysis of ucup
b Fast fourier transformation (FFT) analysis of ucup c Circulating current izu and arm currents of phase u
c Circulating current izu and arm currents of phase u d FFT analysis of izu

IET Renew. Power Gener., pp. 1–10

& The Institution of Engineering and Technology 2016 7
Table 4 Comparison of the SM voltage fluctuation suppression effect when k varies
Arm current RMS value increase δk, % 0 20k0 40k0 60k0 80k0 100k0

ac SM voltage rms ucjpx, V 186.5 153.7 124.3 105.6 99.3 108.1

voltage rms decrease coefficient δ1, % 0 17.6 33.4 43.4 46.8 42.0
coefficient η1 1.136k0 1.198k0 1.382k0 1.709k0 2.381k0
SM fundamental voltage rms value ucjpx_1, V 174.1 148.7 122.2 98.4 73.9 47.9
SM peak-to-peak voltage A pp, V 430 335 275 260 245 240
peak-to-peak voltage decrease coefficient δ2, % 0 22.1 36.0 39.5 43.0 44.2
coefficient η2 0.905k0 1.111k0 1.519k0 1.860k0 2.262k0

voltage decrease coefficient δ2 is set between 35–40%, k could be value of k should be chosen according to the system design
chosen between 0.4k0–0.6k0, which supplies enough SM voltage constraints, with δ1, δ2, η1 and η2 as the criteria.
suppression and acceptable arm current increase.
Table 4 validates that the SM voltage fluctuation has been
suppressed with the proposed circulating current injection method. 6.3 Simulation under grid fault condition
When k = k0, the fundamental SM voltage is the smallest, which
validates the correctness of (23)–(25). Furthermore, the optimum The fault condition is set to verify the proposed distributed chopper
solution. The system works at rated condition before t = 0.35 s.
When t = 0.35 s, an 80% grid voltage dip occurs and the output
power of the system is limited. Redundant energy on the dc
capacitors would cause overvoltage. When t = 0.975 s the fault is
removed.
Fig. 7 shows the dc-link voltage response when voltage faults
occurs under conventional and proposed braking chopper
solutions. Figs. 7a and b are the grid phase voltage and current,
respectively. Under conventional solution Udc suffers instant
voltage peaks when the devices switch on and off in Fig. 7c. With
the proposed strategy, Udc is stable and has no instant voltage
peaks as shown in Fig. 7d.
The full power back-to-back converters ride through grid voltage
fault smoothly. Simulation result has validated the correctness and
effectiveness of the proposed distributed braking chopper solution.
Simulation results in Fig. 8 are performed under 80% three phase
grid voltage dip with the proposed distributed braking chopper
LVRT solution. Fig. 8a shows the active and reactive powers of
grid-side MMC. Ps is limited during fault because of the voltage
dip. Fig. 8b shows the braking chopper current ISM in single SM
of PMSG-side MMC. ISM does not suffer large instant current
peaks during 80% three phase voltage dip, which validates the
feasibility of the proposed strategy.

Fig. 7 Simulation under 80% three phase grid voltage dip

Fig. 8 Simulation under 80% three phase grid voltage dip with the
a Grid phase voltage uabc
b Grid current iabc
proposed LVRT solution
c dc-Link voltage under conventional strategy a Grid-side MMC active and reactive power
d dc-Link voltage under proposed strategy b Braking chopper current ISM in single SM of PMSG-side MMC

IET Renew. Power Gener., pp. 1–10

8 & The Institution of Engineering and Technology 2016
 method is designed to suppress the fundamental voltage
component on the SM capacitor. The amplitude of the injected
circulating current is studied to optimise system design. Second,
braking choppers are distributed into SMs to realise LVRT, which
avoids instant peaks of the dc-link voltage during grid fault.
Simulation results of a 5 MW/3.3 kV PMSG wind system under
normal condition validate the correctness of the overall control
strategy. The results also confirm that SM voltage fluctuation is
suppressed under the proposed injection method. Furthermore,
simulation results under three phase and single phase grid voltage
dip conditions validate the effectiveness of the distributed braking
chopper solution.

8 References
1 Yuan, X., Chai, J., Li, Y.: ‘A transformer-less high-power converter for large
permanent magnet wind generator systems’, IEEE Trans. Sustain. Energy, 2012,
3, (3), pp. 318–329
2 Blaabjerg, F., Liserre, M., Ma, K.: ‘Power electronics converters for wind turbine
systems’, IEEE Trans. Ind. App., 2012, 48, (2), pp. 708–719
3 Yaramasu, V., Wu, B., Sen, P.C., et al.: ‘High-power wind energy conversion
systems: state-of-the-art and emerging technologies’, Proc. IEEE, 2015, 103, (5),
pp. 740–788
4 Muyeen, S.M., Takahashi, R., Murata, T., et al.: ‘Low voltage ride through
capability enhancement of wind turbine generator system during network
disturbance’, IET Renew. Power Gener., 2009, 3, (1), pp. 65–74
5 Mian, W., Ye, T., Xia, F., et al.: ‘A hybrid LVRT control scheme for PMSG wind
power system’. Power Electronics and Motion Control Conf. (IPEMC), Harbin,
China, 2012, pp. 1173–1177
6 Liserre, M., Cardenas, R., Molinas, M., et al.: ‘Overview of multi-MW wind
turbines and wind parks’, IEEE Trans. Ind. Electron., 2011, 58, (4), pp. 1081–1095
7 Lee, K., Jung, K., Suh, Y., et al.: ‘Comparison of high power semiconductor
devices losses in 5 MW PMSG MV wind turbines’. 2014 29th Annual IEEE
Applied Power Electronics Conf. and Exposition (APEC), 2014, pp. 2511–2518
8 Narimani, M., Wu, B., Cheng, Z., et al.: ‘A new nested neutral point-clamped
(NNPC) converter for medium-voltage (MV) power conversion’, IEEE Trans.
Power Electron., 2014, 29, (12), pp. 6375–6382
9 Chivite-Zabalza, J., Larrazabal, I., Zubimendi, I., et al.: ‘Multi-megawatt wind
turbine converter configurations suitable for off-shore applications, combining
3-L NPC PEBBs’. 2013 IEEE Energy Conversion Congress and Exposition
(ECCE), 2013, pp. 2635–2640
10 Islam, M.R., Youguang, G., Jianguo, Z.: ‘A high-frequency link multilevel
cascaded medium-voltage converter for direct grid integration of renewable
energy systems’, IEEE Trans. Power Electron., 2014, 29, (8), pp. 4167–4182
11 Akagi, H.: ‘Classification, terminology, and application of the modular multilevel
cascade converter (MMCC)’, IEEE Trans. Power Electron., 2011, 26, (11),
pp. 3119–3130
12 Geyer, T., Mastellone, S.: ‘Model predictive direct torque control of a five-level
ANPC converter drive system’, IEEE Trans. Ind. Appl., 2012, 48, (5), pp. 1565–1575
13 Ma, K., Blaabjerg, F., Xu, D.: ‘Power devices loading in multilevel converters for
10 MW wind turbines’. 2011 IEEE Int. Symp. on Industrial Electronics (ISIE), 2011
14 Lesnicar, A., Marquardt, R.: ‘An innovative modular multilevel converter topology
suitable for a wide power range’. 2003 IEEE Bologna PowerTech Conf., Bologna,
Italy, 2003, vol. 3, p. 1
15 Ilves, K., Antonopoulos, A., Norrga, S., et al.: ‘Steady-state analysis of interaction
Fig. 9 Simulation under 80% phase a voltage dip with the proposed LVRT
between harmonic components of arm and line quantities of modular multilevel
solution converters’, IEEE Trans. Power Electron., 2012, 27, (1), pp. 57–68
a Grid voltage uabc 16 Alexander, A., Thathan, M.: ‘Modelling and analysis of modular multilevel
b Grid voltage uabc extended from t = 0.3 s to t = 0.5 s converter for solar photovoltaic applications to improve power quality’, IET
c dc-Link voltage Udc Renew. Power Gener., 2015, 9, (1SI), pp. 78–88
d Braking chopper current ISM 17 Korn, A.J., Winkelnkemper, M., Steimer, P.: ‘Low output frequency operation of
the modular multi-level converter’. IEEE Energy Conversion Congress and
Exposition (ECCE 2010), Atlanta, USA, 2010, p. 5
18 Wang, C., Hao, Q., Ooi, B.T.: ‘Reduction of low-frequency harmonics in modular
Fig. 9 shows the responses of the system under single phase multilevel converters (MMCs) by harmonic function analysis’, IET Gener. Transm.
(phase a) 80% voltage dip. Voltage of phase a sags at t = 0.35 s Distrib., 2014, 8, (2), pp. 328–338
and when t = 0.975 s the fault is removed. During the fault Udc is 19 Hagiwara, M., Hasegawa, I., Akagi, H.: ‘Startup and low-speed operation of an
adjustable-speed motor driven by a modular multilevel cascade inverter
stable and ISM does not suffer large instant current peaks. The (MMCI)’. 2012 IEEE Energy Conversion Congress and Exposition (ECCE),
results show that the back-to-back MMC converters can ride 2012, pp. 718–725
through single phase grid voltage fault smoothly with the proposed 20 Kolb, J., Kammerer, F., Gommeringer, M., et al.: ‘Cascaded control system of the
distributed LVRT solution. modular multilevel converter for feeding variable-speed drives’, IEEE Trans.
Power Electron., 2015, 30, (1SI), pp. 349–357
21 Engel, S.P., De Doncker, R.W.: ‘Control of the modular multi-level converter for
minimized cell capacitance’. Proc. 2011 – 14th European Conf. on Power
7 Conclusions Electronics and Applications (EPE 2011), 2011
22 Ilves, K., Antonopoulos, A., Norrga, S., et al.: ‘A new modulation method for the
modular multilevel converter allowing fundamental switching frequency’, IEEE
Two technical difficulties for the application of MMC in MV PMSG Trans. Power Electron., 2012, 27, (8), pp. 3482–3494
wind systems are studied and reasonable solutions are proposed in 23 Picas, R., Pou, J., Ceballos, S., et al.: ‘Optimal injection of harmonics in circulating
this paper. First, overall control strategy based on CPS-SPWM currents of modular multilevel converters for capacitor voltage ripple
minimization’ (IEEE, New York, 2013), pp. 318–324
modulation technique for the back-to-back full power MMC 24 Debnath, S., Qin, J., Saeedifard, M.: ‘Control and stability analysis of modular
PMSG wind systems is proposed. On the basis of the overall multilevel converter under low-frequency operation’, IEEE Trans. Ind. Electron.,
control strategy, a second-order circulating current injection 2015, 62, (9), pp. 5329–5339

IET Renew. Power Gener., pp. 1–10

& The Institution of Engineering and Technology 2016 9
25 Debnath, S., Qin, J., Bahrani, B., et al.: ‘Operation, control, and applications of the 27 Hagiwara, M., Akagi, H.: ‘Control and experiment of pulsewidth-modulated
modular multilevel converter: a review’, IEEE Trans. Power Electron., 2015, 30, modular multilevel converters’, IEEE Trans. Power Electron., 2009, 24, (7),
(1SI), pp. 37–53 pp. 1737–1746
26 Schoening, S., Steimer, P.K., Kolar, J.W.: ‘Braking chopper solutions for modular 28 Li, Z., Wang, P., Zhu, H., et al.: ‘An improved pulse width modulation method for
multilevel converters’. Proc. 2011 – 14th Conf. on Power Electronics and chopper-cell-based modular multilevel converters’, IEEE Trans. Power Electron.,
Applications (EPE 2011), 2011 2012, 27, (8), pp. 3472–3481

IET Renew. Power Gener., pp. 1–10

10 & The Institution of Engineering and Technology 2016