Modular Multilevel Converters Recent Achievements and Challenges
Modular Multilevel Converters Recent Achievements and Challenges
ABSTRACT The modular multilevel converter (MMC) is currently one of the power converter topologies
which has attracted more research and development worldwide. Its features, such as high quality of voltages
and currents, high modularity and high voltage rating, have made the MMC a very good option for several
applications including high-voltage dc (HVdc) transmission, static compensators (STATCOMs), and motor
drives. However, its unique features such as the large number of submodules, floating capacitor voltages, and
circulating currents require a dedicated control system able to manage the terminal variables, as well as the
internal variables with high dynamical performance. In this paper, a review of the research and development
achieved during the last years on MMCs is shown, focusing on the challenges and proposed solutions for this
power converter still faces in terms of modeling, control, reliability, power topologies, and new applications.
INDEX TERMS Modeling, modular multilevel converters, modulation and control, multilevel converters,
power electronics.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
224 VOLUME 2, 2021
FIGURE 1. Dc to three-phase MMC topology.
[23] while the IGCT based submodule, shown in Fig. 1(g), valid for power system stability studies. They develop approx-
features higher efficiency, voltage rating and reliability [24]. imated models that neglect the effect of the capacitor volt-
age ripple in the dynamics of the system. As a consequence,
III. MODELING AND DESIGN the linearization process required to obtain the small-signal
The increasing integration of the MMC in HVdc systems re- ac-side impedance is simplified. However, the internal dy-
quires improved models of the converter to reduce simulation namics may have an impact on the ac-side impedance. On
time, considering that the power system must be simulated this basis, the multi-harmonic coupling behavior of the MMC
for several seconds or minutes. A simple and reliable model is represented and an impedance model for power system
also facilitates the theoretical analysis and consequently, the stability studies is obtained. [42] makes use of several dq-
design of the MMC. Models for theoretical analysis and sim- transformations at different frequencies to separate the mul-
ulation and design guidelines are discussed in this section. tiple frequency components within the MMC. State-variables
that represent the dynamic behavior of the system settle at an
A. MODELING equilibrium point. Consequently, a small-signal linear state-
The initial approach to model the MMC for control purposes space representation of the MMC can be calculated and the
uses a static model in which the ac and dc voltages and cur- state space matrix together with the eigenvalues and eigenvec-
rents are modeled without dynamical components [25], gener- tors that define the small-signal system stability are defined.
ating the ac and dc modulating indices directly [26]. A math- A complete HVdc system including converter stations, dc
ematical model of the MMC, including the capacitor voltage line and ac systems, as shown in Fig. 3, usually requires sev-
dynamics and balancing algorithm was introduced later [27]. eral seconds, or even minutes, of computational simulation.
The natural dynamic response of the circulating currents was These simulation times are not achievable if the model is
showed and the need for circulating current control to reduce based on switching devices [43]. Several models have been
capacitor voltage ripples demonstrated at this point. The dy- proposed to speed up the simulation, such as the linearized
namic model of the differential and summative arm currents, operating curve of insulated-gate bipolar transistors (IGBTs)
i.e., ac and dc currents, was introduced in [28] facilitating the to avoid the lengthy nonlinear solving [44], a model based
control design and improving its performance. The dynamic on dynamic phasors to directly obtain the steady-state opera-
model of the circulating current was also added to further tion [45], and a reduced-order small-signal model that sim-
improve the performance of the converter control [29]. In the plifies the model and enhances simulation efficiency [46].
particular case when the ac and dc sides are both grounded, Alternative MMC topologies, such as the alternate arm con-
a common-mode current appears through the converter arms. verter (AAC), have also been modeled in order to make it
The dynamic behavior of this common-mode current was in- compatible with multi-terminal dc (MTdc) simulation [47].
troduced in [30] completing a mathematical model with six Current-source models of the MMC, based on Norton equiv-
degrees of freedom corresponding to the six controlled arm alent to study a back-to-back system, can also be found in the
voltages. The extension of the MMC model using sequence technical literature [48].
decomposition was proposed to work with unbalanced grid The converter design has been improved during the last
conditions [31]. This model has been recently improved us- years, incorporating into the control strategy the thermal bal-
ing small-signal dynamics [32] and obtaining a steady-state ance among submodules in order to avoid overheating of a
model to improve the design [33]. submodule. To provide a simple thermal modeling, several
The effect of voltage imbalances and signal quantization on stages of thermal resistances and capacitances are used [49].
the accuracy of reduced simulation models is analyzed, and However, the main advances shown in this area are related to
improved results when using a low number of submodules the simplified power loss model for analytic purposes, both at
per arm are obtained [34]. Models to study the harmonic converter level [50] and at semiconductor level [51] using data
performance of the MMC are introduced such as a steady- from datasheet and linearized energy calculations. Thanks to
state mathematical model which relates harmonics in the arms the simplicity of these models, it is possible to accurately sim-
and line currents [35], and a simplified approach based on ulate the complete converter in terms of electrical variables,
harmonic linearization to facilitate the analysis of the MMC including the thermal behavior.
in the frequency domain [36]. More sophisticated methods
able to capture the internal harmonic dynamic behavior of the
MMC have been recently presented. Harmonic linearization B. DESIGN
theory is used to develop an ac-side impedance model of Industrial MMCs for HVdc systems have rated powers rang-
the MMC [37]. Alternatively, [38] and [39] use the harmonic ing from hundreds of MWs to a few GWs. Therefore, the
steady-state analysis to capture the linear time-periodic char- design of such systems involves many technical challenges,
acteristic of the MMC. including sizing reactive components, selecting proper power
From the standpoint of power system stability, impedance devices, designing the mechanical structure and cooling sys-
and state-space modeling of MMCs have been developed tem, etc. Since each submodule has a dc bus and there
during the last years. Ref. [40] and [41] are two of the first are hundreds of submodules, the MMC integrates many ca-
attempts to develop impedance models of three-phase MMCs pacitors. Minimizing the capacitance value required in each
submodule of the converter is a key target that has been exten- provide the required ac, dc and common-mode voltages [55].
sively addressed in the literature. Capacitor voltage ripples are This control system was improved using dynamical models
intimately linked to the circulating currents in the converter. to drive the ac and dc currents to their references and adding
The information regarding normalized capacitor voltage rip- an average capacitor voltage control and a capacitor voltage
ple amplitudes allows the sizing of capacitors for different cir- balance strategy [56]. Since then, a control structure using
culating current strategies considering all the possible operat- three stages: an internal capacitor voltage balance, a current
ing points of the converter [52]. When half of the submodules control stage, and a voltage, or energy, control stage, as shown
of each arm have half-bridge topology and the remaining have in Fig. 4, has commonly been employed [57], [58].
H-bridge topology the capacitor size can be reduced. Adding a At the innermost control level, the capacitor voltage bal-
zero sequence to the reference voltages demonstrated a further ance is usually implemented using a sorting algorithm, which
reduction in the capacitor size [53]. requires the value of each capacitor voltage and the current in
The design of a fully rated MMC with several hundred each arm. A version of this algorithm has been implemented
submodules where each valve has to meet all conditions using a current estimator to reduce the number of sensors
in service, the values of the transformer turns ratio, the and the wiring complexity [59]. It is possible to modify the
transformer reactance, and the valve reactance, considering sorting algorithm considering the specific topology of each
how the ripple that appears on the submodule voltage are submodule, for example, for flying capacitor submodules, in-
calculated [54]. tegrating the capacitor voltage balance into the modulation
algorithm [60].
IV. CONTROL AND MODULATION The current control generates the modulation indexes in
To benefit from the high power quality featured by the MMC, order to synthesize the required voltages in the converter
a suitable modulation technique and controller must be em- arms. This control stage is usually designed using the dynamic
ployed. The structure of the control systems, different control models of the MMC, but they can be adjusted including the
methods, control objectives and modulation techniques are positive and negative sequences in the model to work with un-
described in this section. balanced voltages [61]. It is also possible to take into account
the effect of the capacitor voltage ripple in the arm currents
A. CONTROL SYSTEM STRUCTURE using a transformation to obtain the fundamental component
Early attempts to control the MMC were focused on the di- of these voltages and feedforward it into the control loop,
rect generation of arm voltages, i.e. modulation indices, to improving the dynamic performance [62].
The outermost control loop is usually focused on regulating for voltage balancing and linearizing the objective function
the average arm capacitor voltages [63], or energies [64]. reduces the computational burden [77]. It is also possible to
Both alternatives have shown to provide proper regulation, but reduce the processing time by separating the optimization
the energy approach has the advantage of providing a linear process into two stages: one to calculate the insertion index,
relationship between power and currents, keeping the same and a different stage to calculate the inserted submodules [78].
dynamic behavior even if the operating point changes. Recent MPC algorithms proposed in the literature employ
Increasing the number of submodules often necessitates three different stages to control the grid side current, circu-
the use of distributed architectures for control methods and lating current and voltage balancing [79]. The first two stages
hardware design, which include capacitor grouping and inter- do not depend on the number of submodules, and only the
group balancing on top of all other balancing and control last one has a processing time that increases with the number
structures [65]. of submodules. Unbalanced grid voltages can be controlled
with MPC using a per-phase approach in which each phase
B. CONTROL METHODS is controlled separately [80], or using a three-phase approach
Most of the control methods proposed in the literature work that requires more calculations but improves the converter
well with a fixed ac frequency, such as in the case of HVdc performance [81].
systems. However, in motor drive applications, where the ac With the increased interest on the development of con-
frequency changes, the capacitor voltage control strategy has trol techniques to improve the inertial response and primary
to be modified. For example, it can be separated into three frequency control in grids with high penetration of power
different current and voltage injection strategies depending on converters, several grid-forming control algorithms have been
the motor speed [66] or controlling the ac and dc component adapted to operate together with MMCs [82], [83]. A virtual
of the arms independently using an unsymmetrical arm con- synchronous machine control technique for MMCs with iner-
trol [67]. The former has a simpler implementation, but suf- tia response has been introduced later [84]. Different inertia
fers from operating discontinuity produced when the control emulation strategies for an MMC-based HVdc transmission
strategy changes, while the latter has better performance, but link have been evaluated taking into consideration the energy
increased complexity. stored in the submodule capacitances [85]. A droop control for
To improve the control performance when the ac frequency MMC performs active power support as well as parallel oper-
changes, sliding mode control has been proposed for this ation with other voltage sources and islanding operation [86].
converter, providing faster dynamical behavior than linear It has been demonstrated that the internal energy control of
proportional-integral (PI) controllers without increasing the the MMC impacts the dynamic behavior while operated with
processing time [68]. grid-forming control [87].
Model predictive control (MPC) is a simple and powerful
alternative to control power converters which can be also C. ADDITIONAL CONTROL OBJECTIVES
applied when the ac frequency changes. One of the first ex- The circulating current is one of the distinctive features of the
amples of this type of control applied to MMCs was on a MMC, which represents the current that circulates among the
single-phase to single-phase MMC topology, in which the arms of the converter, but it does not circulate through the
input, output and circulating currents are controlled simulta- converter terminals. An early approach to control and elimi-
neously [69], and a back-to-back HVdc system [70]. Despite nate the circulating currents in PWM-based MMCs obtaining
the fast performance of MPC, the large number of possible the circulating current reference from a control loop based
switching states becomes one of the main weakness of this on PI controllers [77], [88], using a current controller based
control method in MMC applications due to the high pro- on double-frequency rotational synchronous frame [89]. The
cessing power required. To overcome this problem, several low-frequency harmonics of the circulating currents concen-
alternatives have been proposed such as modulated MPC, in trate the losses; hence the effort has been focused on design a
which a continuous optimization is employed and the result- controller to eliminate them [90]. Although the main control
ing modulation index is then modulated with phase-shifted objective regarding circulating currents initially was to sup-
pulse-width modulation (PWM) [71]. In addition to reduc- press them completely, it has been recently proven that the
ing the processing time, the modulated MPC approach also injection of a specific circulating current can be used to reduce
reduces the steady-state error [72]. A predefined switching losses [91] and also capacitor voltage ripples [92].
sequence can be used to reduce the circulating current [73]and The combination of arm currents and modulation signals
pulse-pattern control in order to reduce the harmonic content generates several harmonic components in the capacitor volt-
on the currents [74]. age of each submodule. This capacitor voltage ripple can usu-
It is possible to reduce the universe of possible switching ally be reduced by a proper design of submodule capacitance.
states by preselecting the available states [75], or grouping However, in a motor drive, where the ac operating frequency
them depending on the operation conditions [76]. Both al- can be lower than the grid, the capacitor voltage ripple in-
ternatives reduce the processing time by evaluating only a creases and impacts the performance of the whole control sys-
fraction of the total switching states, but the result corresponds tem [93]. There are several approaches from the topology per-
only to a local optimization. Also, clustering the submodules spective to deal with the capacitor voltage ripple, for example
generate different thermal stress among phases, which can be TABLE 1. Classification of Faults in MMC
avoided using the neutral-point shifting with minimal impact
on the circulating currents [123]. Moreover, using the shifted
neutral point combined with a dc shift strategy, it is possible
to extend the fault response capability maximizing the line
voltages [124]. Is important to have a proper ac fault manage-
ment strategy to avoid the large dc fault currents generated
by single-phase-to-ground failures that are not properly man-
aged [125].
B. SUBMODULE FAILURES
Faulty submodules can be dealt with different redundancy
strategies such as: (i) not including any additional submodules
and using the remaining ones with an overrated voltage ratio,
(ii) including redundant submodules which are permanently
working or (iii) using spare submodules that are connected
when some submodules fail. A reliability model is used to
assess the different redundancy strategies resulting on a large
dependency on the control system reliability [126]. Addi-
tionally, these strategies also perform differently in terms of
voltage and currents stress, resulting in competing behavior or with series-connected IGBTs. Case studies have been con-
between losses and dynamic [127]. ducted to compare the reliability characteristics of converters
The operation of redundant submodules which are acti- constructed using the two-submodule configurations [137]. It
vated when a submodule fails can be easily implemented is found that the reliability of the MMC with series-connected
using a phase-disposition PWM (PD-PWM) with a selection IGBTs is higher for the first few years but then decreases
algorithm to evaluate which and when the submodule has to rapidly. By assigning a reduced nominal voltage to the series
be activated [128]. On the other hand, phase-shifted PWM valve submodule upon IGBT module failure, the need to in-
(PS-PWM), can also be used with the same purpose, but it stall redundant submodules is greatly reduced.
requires carrier rearrangements [129] or rotative carrier selec- A precise reliability model is useful to employ statistical
tion [130]. tools, combined with information of submodules and com-
An MMC typically requires one voltage sensor per sub- ponents, to extend the availability of the MMC by properly
module and such a large number of voltage sensors impacts designing the number of redundant submodules and defining
the converter reliability. A capacitor voltage measuring sys- management schedules [138]. There is a trade-off between
tem with reduced number of sensors can be implemented, submodule redundancy and cost, which should be evaluated
where the minimum number can be as low as two voltage for each project to define redundancy design and operation
sensors per arm [131]. This technique can be used as a com- strategy [139].
plementary measuring system to further improve converter A reliability analysis performed on the circulating current
reliability by providing fault tolerance to voltage sensor fail- and common-mode voltage control strategies shows the ther-
ure. The proposed measurement technique can also be used mal stress of semiconductors depends on where they are phys-
to detect submodule failures, including both, open- and short- ically located in the circuit hence the cooling design can be
circuit semiconductor faults [132]. Other solutions to detect optimized [140].
faults without additional sensors have been previously pro-
posed in the literature [133]–[135]; however these solutions VI. TOPOLOGIES AND APPLICATIONS
can only detect open-circuit semiconductor faults. The most extended topology of the MMC is the one shown
In Table 1 a comparison of different types of faults and the in Fig. 1, where the submodules can be implemented by half-
causes identified is shown. or full-bridge converters. Many topology variations have been
presented in the literature, each of them targeting the improve-
C. RELIABILITY ment of certain features or addressing some specific applica-
A reliability model assuming periodic preventive maintenance tions. Several of these converter topologies and applications
is proposed in [136]. The model is derived by the reliability are discussed next.
function and described by the indices of equivalent failure rate
and forced outage rate. The reliability in terms of submodule A. MMC TOPOLOGIES
redundancy and maintenance interval is analyzed to provide Among the topologies recently proposed for MMCs, the star-
information for operation and planning, which is especially and delta-channel ones, shown in Figs. 5(a) and (b), respec-
valuable for offshore utilities. The submodules within the con- tively, are very interesting for motor drives due to the reduced
verter can be constructed with either individual IGBT modules capacitor voltage ripple they achieve. These topologies have
a set of additional submodules, shown as gray boxes, that The standard three-phase MMC topology can be used as a
are connected to the arms of different phases allowing the dc-dc converter connecting all the outputs to a common point,
exchange of power among the arms, and hence helping to as shown in Fig. 5(d). In this case, the control system has to
minimize the capacitor voltage ripples even when operating generate a controllable dc output voltage, while a circulating
at low frequency. The main drawback of these topologies is current is forced to circulate among the arms in order to
the large number of submodules [141], [142]. balance the capacitor voltages [147]. The previous topology
The Alternate Arm Converter (AAC) topology, in which can be improved by connecting three additional arms between
the arms are connected in series with a high voltage switch, the middle point of each phase and the output, as shown in
as shown in Fig. 5(c), allows the activation or deactivation of Fig. 5(e). By adding these submodules, the complexity of the
the complete arm. Therefore, it provides better performance control is reduced, while the effect of the internal circulat-
than the standard MMC in terms of circulating current, power ing current can be decoupled from the input and output dc
losses, and capacitor voltage ripple [143]. Additionally, it is currents [148]. The tolerance to faults can be increased as
possible to define an overlap period in which both arms in well because the output power can be redirected if one of the
the same phase are active at the same time to help capacitor submodules fails.
voltage balance and control [144]. It is also possible to inject Using a single-phase MMC in a half-bridge configuration, a
circulating current to achieve zero current switching in the high-frequency transformer and a single-phase diode rectifier,
high-voltage switch reducing the losses [145]. as shown in Fig. 5(f), a high step ratio dc-dc converter to en-
The high performance, in terms of efficiency and control- able HV and MV grids interconnection is achieved [149]. The
lability, showed by the MMC on HVdc systems, have pushed arms operate with square wave, which further increases the ef-
the development of medium-voltage dc (MVdc) systems for ficiency. Using only one-phase MMC, two auto-transformers,
power distribution in which the MMC becomes and essential LC filters and a diode rectifier, as shown in Fig. 5(g), pro-
element [146]. Several topologies providing dc-dc conversion vides a low step-up ratio dc-dc conversion [150]. The auto-
are described in the following. transformer leakage inductances are used as arm inductances,
FIGURE 6. MMC applications: (a) MVdc for shipboard, (b) three-phase motor drive, (c) modular multilevel STATCOM, (d) distributed BESS, and
(e) solid-state transformer, (f) high-voltage pulse generator.
providing resonant operation with soft switching while the cir- currents, but instead integrating full-bridge submodules in the
cuit interconnection between the upper and lower arms allows MMC and implementing a proper control strategy to operate
to circulate current through the arms to balance the capacitor the converter itself during the fault [155].
voltages. One of the early applications of the MMC was as a motor
A very simple alternative to implement dc-dc conver- drive, as shown in Fig. 6(b). However, the large increment
sion is using one MMC phase with the load connected be- of capacitor voltage ripple and the reduced controllability at
tween the upper and lower arms, as shown in Fig. 5(h). low speeds, slowing down the development of this applica-
This topology mimics the operation of a dc-dc buck or tion [66]. These disadvantages have recently been addressed
boost converter working with square-wave modulation for by several researchers using for example discontinuous modu-
high efficiency, achieving a drastic reduction in submodule lation to reduce the capacitor voltage ripples [98], or eliminat-
capacitance [151]. ing the common-mode voltage to extend the machine lifetime
A dc-dc converter using only one single MMC arm, in by using a flying capacitor submodule [156]. Decreasing the
which each capacitor submodule is connected to a high- operating capacitor voltage also reduces the voltage ripple
frequency isolated dual active bridge (DAB), as shown in allowing to reduce the capacitance [157]. A recent trend to
Fig. 5(i), has been proposed to connect a photovoltaic plant improve the performance of the MMC as a motor drive is the
to a high voltage dc grid [152]. This topology provides mod- incorporation of the machine model into the control system
ular interconnection, ground isolation of each submodule and and using vector control to manage both, the converter and the
independent maximum power point tracking (MPPT). machine at the same time [158]. Additionally, a back-to-back
MMC structure implementing a control strategy to provide
B. MMC APPLICATIONS a controllable current source in the grid converter, allows to
The MVdc systems mentioned in the previous section can be drive the machine at very low frequencies [159].
found in shipboards, as shown in Fig. 6(a), where the reliabil- The MMC enables direct connection of a static compen-
ity of the MMC is an important factor to consider. Designing sator (STATCOM) at distribution level and even at trans-
the on-board converter with redundant submodules, the con- mission level avoiding the use of transformers, as shown in
verter reliability is improved [153]. Moreover, this converter Fig. 6(c). Recent investigations have determined that negative
can be used in combination with energy storage and active fil- sequence compensation capability is a key performance indi-
tering functionalities in order to improve the availability of the cator for STATCOM devices; hence the diode-clamped [160]
complete on-board system [154]. Regarding fault tolerance, it and delta-connected MMC show better performance in terms
is not recommended to use dc circuit breakers to clear the fault of efficiency [161].
232 VOLUME 2, 2021
TABLE 2. MMC Topologies Comparison
The high-voltage rating of the MMC allows to connect Additionally, the structure of this modular SST can be config-
a dc battery energy storage system (BESS) directly to the ured as shown in Fig. 6(e), where all the primary stages are
distribution or transmission grid, and the modularity of the connected in series providing high voltage rating, and the sec-
MMC can also be used to interconnect distributed energy ondary stages in parallel providing high current rating [168],
storage [162], as shown in Fig. 6(d). In this configuration, mimicking the feature of a distribution transformer.
each storage module can be controlled independently in order Due to the high voltage capability of the MMC it can be
to work at different operating points, increasing the overall ef- used as a high voltage pulse generator for water disinfec-
ficiency [163]. Furthermore, the high power quality provided tion [169], as shown in Fig. 6(f). One MMC phase is used
by the MMC enables its use in distributed energy storage sys- to store energy from the low voltage input into the floating
tems of electric vehicles, where electromagnetic interferences capacitors and then release it in the form of a high-voltage
should be minimized [164]. A decoupled power control is short-time period pulse [170].
proposed to improve the overall stability of distributed energy
storage units [165]. VII. CONCLUSION
The MMC has been proposed to build a solid-state trans- The MMC is an interesting power converter topology cur-
former (SST) using a single transformer and employing the rently used in several industrial applications and commercial
MMC to generate the high-voltage at the primary side [166]. products. A large amount of literature has been published on
This configuration simplifies the operation of the MMC but the MMC addressing the main issues this topology presents,
it lost the modularity provided by the MMC. Using a high- such as current dynamical models, capacitor voltage balanc-
frequency dual active bridge with isolation transformers is ing using sorting algorithms, control structure, and topologies
possible to operate each submodule to provide power trans- for conventional applications such as HVdc, motor drives and
fer independently, and hence, improving the reliability [167]. STATCOM.
TABLE 3. Challenges in MMC and Prospective Solutions be charged and discharged. This sorting algorithm can be in-
tegrated with the modulation to simplified it. However, open-
loop balancing methods such as the rotation of the modulation
signals still have potential due to the reduced computation
time required compared with any sorting algorithm. Due to the
nature of circulating current, its control considering several
frequency components is an interesting tendency. Most of
the published control schemes are intended for HVdc with
fixed output frequency. However, for variable frequency, the
control scheme must be developed in rotating coordinates or
using a frequency-independent strategy such as sliding con-
trol or MPC. However, applying MPC requires grouping or
clustering to reduce the large number of possible states to be
evaluated. The minimization of capacitor voltage ripples leads
to reducing the submodule size, given that approximately
half of its volume is the capacitor. This minimization can be
performed by topological changes at the submodule and con-
verter level or by injecting a circulating current and common-
mode voltage. When the MMC is placed in a distribution grid,
the virtual inertia provided can be optimized, considering the
large amount of distributed capacitors, and particularly for
microgrids, a grid forming methodology must be developed.
The MMC itself can be used to manage the dc fault currents
simplifying the protection system. This alternative requires
further investigations in terms of practical results. The ac
faults can usually be handled by shifting the neutral point
to reduce the asymmetric currents. Although this method has
been studied previously, the implementation in MMC requires
more research. There are several alternatives to implement
mitigation strategies when a submodule fails, which must be
evaluated thoroughly. The large number of voltage sensors
In this review, several challenges that have not been needed for control purposes impacts the MMC cost and re-
completely solved are identified and the prospective solutions liability. These voltages can be reconstructed using the arm
have been summarized. A list of these challenges is shown voltage and modulation indexes, but these observers must be
in Table 3. To ensure the proper operation of the MMC in precisely tuned.
unbalanced grids, precise small-signal models are required Several dc-dc conversion alternatives have been proposed
and the development of current and energy controllers in terms of topologies, and it is expected this trend continues.
considering the additional harmonic components that appear Although the main application of the MMC is still HVdc
in this condition. To study the interaction between the MMC systems, recent research findings allow its use on standard
with the power system, a suitable model of the ac side applications such motor drives and STATCOM but also on
impedance is required. In this regard, small-signal models emerging applications including BESS systems, MVdc sys-
also provide simple results that can be used to study the tems, and high-voltage pulse generators.
system stability based on the eigenvalues analysis. Power Addressing these challenges can not only improve the reli-
systems require dynamical simulation with a time scope of ability of the MMC and its performance, reduce its cost and
several seconds or even minutes. MMC simulation models size, widening its use in HVdc and motor drives, but also
need to be accurate and simple to simulate the MMC into a pushing the development of MVdc grids and enabling its use
power system. A proper thermal model is also an important in non-conventional applications.
challenge due to the significant economic implications; hence
the use of linearized energy calculation using data from
datasheets can provide accurate results. REFERENCES
Due to the large number of submodules per arm, nearest [1] R. Marquardt, “Stromrichterschaltungen mit verteilten energiespeich-
ern,” German Patent DE 201 22 923 U1, 2001.
level modulation is still the preferred option, but it needs to be [2] M. Glinka and R. Marquardt, “A new AC/AC-multilevel converter
combined with PWM to eliminate low-frequency harmonics. family applied to a single-phase converter,” in Proc. 5th Int. Conf.
A future trend is to adapt NLM to work with a low number of Power Electron. Drive Syst., vol. 1, 2003, pp. 16–23.
[3] M. Glinka and R. Marquardt, “A new AC/AC multilevel converter
submodules for MV applications. Capacitor voltage balancing family,” IEEE Trans. Ind. Electron., vol. 52, no. 3, pp. 662–669,
usually uses a sorting algorithm to identify the cells that must Jun. 2005.
[46] Q. Hao, Z. Li, F. Gao, and J. Zhang, “Reduced-order small-signal [66] A. Antonopoulos, L. Ängquist, S. Norrga, K. Ilves, L. Harnefors, and
models of modular multilevel converter and MMC-based HVDC H. Nee, “Modular multilevel converter AC motor drives with constant
grid,” IEEE Trans. Ind. Electron., vol. 66, , no. 3, pp. 2257–2268, torque from zero to nominal speed,” IEEE Trans. Ind. Appl., vol. 50,
Mar. 2019. no. 3, pp. 1982–1993, May/Jun. 2014.
[47] H. R. Wickramasinghe, G. Konstantinou, Z. Li, and J. Pou, “Alter- [67] B. Tai, C. Gao, X. Liu, and Z. Chen, “A novel flexible capacitor volt-
nate arm converters-based HVDC model compatible with the CIGRE age control strategy for variable-speed drives with modular multilevel
B4 DC grid test system,” IEEE Trans. Power Del., vol. 34, no. 1, converters,” IEEE Trans. Power Electron., vol. 32, no. 1, pp. 128–141,
pp. 149–159, Feb. 2019. Jan. 2017.
[48] M. M. Bhesaniya and A. Shukla, “Norton equivalent modeling of cur- [68] Q. Yang, M. Saeedifard, and M. A. Perez, “Sliding mode control of
rent source MMC and its use for dynamic studies of back-to-back con- the modular multilevel converter,” IEEE Trans. Ind. Electron., vol. 66,
verter system,” IEEE Trans. Power Del., vol. 32, no. 4, pp. 1935–1945, no. 2, pp. 887–897, Feb. 2019.
Aug. 2017. [69] M. A. Perez, J. Rodriguez, E. J. Fuentes, and F. Kammerer, “Predictive
[49] F. Hahn, M. Andresen, G. Buticchi, and M. Liserre, “Thermal analysis control of AC-AC modular multilevel converters,” IEEE Trans. Ind.
and balancing for modular multilevel converters in HVDC applica- Electron., vol. 59, no. 7, pp. 2832–2839, Jul. 2012.
tions,” IEEE Trans. Power Electron., vol. 33, no. 3, pp. 1985–1996, [70] J. Qin and M. Saeedifard, “Predictive control of a modular multilevel
Mar. 2018. converter for a back-to-back HVDC system,” IEEE Trans. Power Del.,
[50] L. Yang, Y. Li, Z. Li, P. Wang, S. Xu, and R. Gou, “A simplified ana- vol. 27, no. 3, pp. 1538–1547, Jul. 2012.
lytical calculation model of average power loss for modular multilevel [71] D. Zhou, S. Yang, and Y. Tang, “Model-predictive current control of
converter,” IEEE Trans. Ind. Electron., vol. 66, no. 3, pp. 2313–2322, modular multilevel converters with phase-shifted pulsewidth modu-
Mar. 2019. lation,” IEEE Trans. Ind. Electron., vol. 66, no. 6, pp. 4368–4378,
[51] Y. Zhang, H. Wang, Z. Wang, Y. Yang, and F. Blaabjerg, “Simplified Jun. 2019.
thermal modeling for IGBT modules with periodic power loss profiles [72] J. Wang, X. Liu, Q. Xiao, D. Zhou, H. Qiu, and Y. Tang, “Modulated
in modular multilevel converters,” IEEE Trans. Ind. Electron., vol. 66, model predictive control for modular multilevel converters with easy
no. 3, pp. 2323–2332, Mar. 2019. implementation and enhanced steady-state performance,” IEEE Trans.
[52] J. Pou, S. Ceballos, G. Konstantinou, V. G. Agelidis, R. Picas, and J. Power Electron., vol. 35, no. 9, pp. 9107–9118, Sep. 2020.
Zaragoza, “Circulating current injection methods based on instanta- [73] H. Mahmoudi, M. Aleenejad, and R. Ahmadi, “Modulated model
neous information for the modular multilevel converter,” IEEE Trans. predictive control of modular multilevel converters in VSC-HVDC
Ind. Electron., vol. 62, no. 2, pp. 777–788, Feb. 2015. systems,” IEEE Trans. Power Del., vol. 33, no. 5, pp. 2115–2124,
[53] P. Dong, J. Lyu, and X. Cai, “Optimized design and control for hybrid Oct. 2018.
MMC with reduced capacitance requirements,” IEEE Access, vol. 6, [74] M. Vasiladiotis, A. Christe, and T. Geyer, “Model predictive pulse
pp. 51069–51083, 2018. pattern control for modular multilevel converters,” IEEE Trans. Ind.
[54] C. Oates, “Modular multilevel converter design for VSC HVDC ap- Electron., vol. 66, no. 3, pp. 2423–2431, Mar. 2019.
plications,” IEEE J. Emerg. Sel. Top. Power Electron., vol. 3, no. 2, [75] B. Gutierrez and S. Kwak, “Modular multilevel converters (MMCS)
pp. 505–515, Jun. 2015. controlled by model predictive control with reduced calculation bur-
[55] S. Allebrod, D. SchmiSchmitt, and R. Marquardt, “Control structure den,” IEEE Trans. Power Electron., vol. 33, no. 11, pp. 9176–9187,
for modular multilevel converters,” in Proc. Power Convers. Intell. Nov. 2018.
Motion Eur., 2009, pp. 12–14. [76] A. Rashwan, M. A. Sayed, Y. A. Mobarak, G. Shabib, and T. Senjyu,
[56] M. Hagiwara, R. Maeda, and H. Akagi, “Control and analysis of the “Predictive controller based on switching state grouping for a modular
modular multilevel cascade converter based on double-star chopper- multilevel converter with reduced computational time,” IEEE Trans.
cells (MMCC-DSCC),” IEEE Trans. Power Electron., vol. 26, no. 6, Power Del., vol. 32, no. 5, pp. 2189–2198, Oct. 2017.
pp. 1649–1658, Jun. 2011. [77] A. Mora et al., “Model-predictive-control-based capacitor voltage bal-
[57] M. Hagiwara and H. Akagi, “Control and experiment of pulsewidth- ancing strategies for modular multilevel converters,” IEEE Trans. Ind.
modulated modular multilevel converters,” IEEE Trans. Power Elec- Electron., vol. 66, no. 3, pp. 2432–2443, Mar. 2019.
tron., vol. 24, no. 7, pp. 1737–1746, Jul. 2009. [78] P. Guo et al., “A novel two-stage model predictive control for modu-
[58] R. Lizana, M. A. Perez, D. Arancibia, J. R. Espinoza, and J. Rodriguez, lar multilevel converter with reduced computation,” IEEE Trans. Ind.
“Decoupled current model and control of modular multilevel convert- Electron., vol. 66, no. 3, pp. 2410–2422, Mar. 2019.
ers,” IEEE Trans. Ind. Electron., vol. 62, no. 9, pp. 5382–5392, Sep. [79] J. Huang et al., “Priority sorting approach for modular mul-
2015. tilevel converter based on simplified model predictive control,”
[59] P. Hu, R. Teodorescu, S. Wang, S. Li, and J. M. Guerrero, “A current- IEEE Trans. Ind. Electron., vol. 65, no. 6, pp. 4819–4830,
less sorting and selection-based capacitor-voltage-balancing method Jun. 2018.
for modular multilevel converters,” IEEE Trans. Power Electron., [80] J. Xu, Y. Yu, and C. Zhao, “The predictive closed-loop averaging con-
vol. 34, no. 2, pp. 1022–1025, Feb. 2019. trol of MMC phase-unit losses under unbalanced conditions,” IEEE
[60] A. Dekka, B. Wu, R. L. Fuentes, M. Perez, and N. R. Zargari, “Voltage- Trans. Power Del., vol. 34, no. 1, pp. 198–207, Feb. 2019.
balancing approach with improved harmonic performance for modular [81] Z. Gong, X. Wu, P. Dai, and R. Zhu, “Modulated model predictive con-
multilevel converters,” IEEE Trans. Power Electron., vol. 32, no. 8, trol for MMC-based active front-end rectifiers under unbalanced grid
pp. 5878–5884, Aug. 2017. conditions,” IEEE Trans. Ind. Electron., vol. 66, no. 3, pp. 2398–2409,
[61] Y. Liang, J. Liu, T. Zhang, and Q. Yang, “Arm current control strategy Mar. 2019.
for MMC-HVDC under unbalanced conditions,” IEEE Trans. Power [82] C. Verdugo, J. I. Candela, and P. Rodriguez, “Grid support functional-
Del., vol. 32, no. 1, pp. 125–134, Feb. 2017. ities based on modular multilevel converters with synchronous power
[62] H. Yang, W. Li, L. Lin, and X. He, “Decoupled current control with control,” in Proc. IEEE Int. Conf. Renewable Energy Res. Appl., 2016,
synchronous frequency damping for MMC considering sub-module pp. 572–577.
capacitor voltage ripple,” IEEE Trans. Power Del., vol. 33, no. 1, [83] O. D. Adeuyi et al., “Frequency support from modular multilevel
pp. 419–428, Feb. 2018. converter based multi-terminal HVDC schemes,” in Proc. IEEE Power
[63] R. Lizana, M. A. Perez, S. Bernet, J. R. Espinoza, and J. Rodriguez, Energy Soc. Gen. Meeting, 2015, pp. 1–5.
“Control of arm capacitor voltages in modular multilevel convert- [84] S. D’Arco, G. Guidi, and J. A. Suul, “Operation of a modular multi-
ers,” IEEE Trans. Power Electron., vol. 31, no. 2, pp. 1774–1784, level converter controlled as a virtual synchronous machine,” in Proc.
Feb. 2016. Int. Power Electron. Conf., 2018, pp. 782–789.
[64] H. Fehr and A. Gensior, “Improved energy balancing of grid-side [85] S. D’Arco, T. T. Nguyen, and J. A. Suul, “Evaluation of virtual inertia
modular multilevel converters by optimized feedforward circulating control strategies for MMC-based HVDC terminals by P-HiL experi-
currents and common-mode voltage,” IEEE Trans. Power Electron., ments,” in Proc. IECON - 45th Annu. Conf. IEEE Ind. Electron. Soc.,
vol. 33, no. 12, pp. 10903–10913, Dec. 2018. 2019, vol. 1, pp. 4811–4818.
[65] B. Xia et al., “Decentralized control method for modular mul- [86] S. Henninger, M. Schroeder, and J. Jaeger, “Grid-forming droop con-
tilevel converters,” IEEE Trans. Power Electron., vol. 34, no. 6, trol of a modular multilevel converter in laboratory,” in Proc. IEEE
pp. 5117–5130, Jun. 2019. Power Energy Soc. Gen. Meeting, 2018, pp. 1–5.
[127] J. V. M. Farias, A. F. Cupertino, H. A. Pereira, S. I. S. Junior, and [147] H. Yang, M. Saeedifard, and A. Yazdani, “An enhanced closed-loop
R. Teodorescu, “On the redundancy strategies of modular multilevel control strategy with capacitor voltage elevation for the DC-DC mod-
converters,” IEEE Trans. Power Del., vol. 33, no. 2, pp. 851–860, ular multilevel converter,” IEEE Trans. Ind. Electron., vol. 66, no. 3,
Apr. 2018. pp. 2366–2375, Mar. 2019.
[128] G. Konstantinou, J. Pou, S. Ceballos, and V. G. Agelidis, “Active [148] F. Zhang, W. Li, and G. Joos, “A transformerless hybrid modu-
redundant submodule configuration in modular multilevel converters,” lar multilevel DC-DC converter with DC fault ride-through capa-
IEEE Trans. Power Del., vol. 28, no. 4, pp. 2333–2341, Oct. 2013. bility,” IEEE Trans. Ind. Electron., vol. 66, no. 3, pp. 2217–2226,
[129] B. Li, Z. Xu, J. Ding, and D. Xu, “Fault-tolerant control of Mar. 2019.
medium-voltage modular multilevel converters with minimum perfor- [149] X. Xiang, X. Zhang, T. Luth, M. M. C. Merlin, and T. C. Green,
mance degradation under submodule failures,” IEEE Access, vol. 6, “A compact modular multilevel DC-DC converter for high step-
pp. 11772–11781, 2018. ratio MV and HV use,” IEEE Trans. Ind. Electron., vol. 65, no. 9,
[130] K. Li, L. Yuan, Z. Zhao, S. Lu, and Y. Zhang, “Fault-tolerant control of pp. 7060–7071, Sep. 2018.
MMC with hot reserved submodules based on carrier phase shift mod- [150] X. Zhang, X. Xiang, T. C. Green, X. Yang, and F. Wang, “A push-
ulation,” IEEE Trans. Power Electron., vol. 32, no. 9, pp. 6778–6791, pull modular-multilevel-converter-based low step-up ratio DC trans-
Sep. 2017. former,” IEEE Trans. Ind. Electron., vol. 66, no. 3, pp. 2247–2256,
[131] R. Picas, J. Zaragoza, J. Pou, S. Ceballos, and J. Balcells, “New Mar. 2019.
measuring technique for reducing the number of voltage sensors in [151] Z. Xing, X. Ruan, and H. Xie, “PWM control scheme for a buck/boost
modular multilevel converters,” IEEE Trans. Power Electron., vol. 31, modular multilevel DC/DC converter with reduced submodule capac-
no. 1, pp. 177–187, Jan. 2016. itance,” IEEE Trans. Power Electron., vol. 33, no. 5, pp. 3911–3925,
[132] R. Picas, J. Zaragoza, J. Pou, and S. Ceballos, “Reliable modular May. 2018.
multilevel converter fault detection with redundant voltage sensor,” [152] C. A. Rojas, S. Kouro, M. A. Perez, and J. Echeverria, “DC-DC
IEEE Trans. Power Electron., vol. 32, no. 1, pp. 39–51, Jan. 2017. MMC for HVDC grid interface of utility-scale photovoltaic conversion
[133] B. Li, S. Shi, B. Wang, G. Wang, W. Wang, and D. Xu, “Fault diagnosis systems,” IEEE Trans. Ind. Electron., vol. 65, no. 1, pp. 352–362,
and tolerant control of single IGBT open-circuit failure in modular Jan. 2018.
multilevel converters,” IEEE Trans. Power Electron., vol. 31, no. 4, [153] Y. Chen, Z. Li, S. Zhao, X. Wei, and Y. Kang, “Design and implemen-
pp. 3165–3176, Apr. 2016. tation of a modular multilevel converter with hierarchical redundancy
[134] D. Zhou, S. Yang, and Y. Tang, “A voltage-based open-circuit fault ability for electric ship MVDC system,” IEEE J. Emerg. Sel. Top.
detection and isolation approach for modular multilevel converters Power Electron., vol. 5, no. 1, pp. 189–202, Mar. 2017.
with model-predictive control,” IEEE Trans. Power Electron., vol. 33, [154] R. Mo and H. Li, “Hybrid energy storage system with active filter
no. 11, pp. 9866–9874, Nov. 2018. function for shipboard MVDC system applications based on isolated
[135] W. Zhou, J. Sheng, H. Luo, W. Li, and X. He, “Detection and lo- modular multilevel DC/DC converter,” IEEE J. Emerg. Sel. Top. Power
calization of submodule open-circuit failures for modular multilevel Electron., vol. 5, no. 1, pp. 79–87, Mar. 2017.
converters with single ring theorem,” IEEE Trans. Power Electron., [155] V. Staudt, R. Bartelt, and C. Heising, “Fault scenarios in DC
vol. 34, no. 4, pp. 3729–3739, Apr. 2019. ship grids: The advantages and disadvantages of modular multi-
[136] B. Wang, X. Wang, Z. Bie, P. D. Judge, X. Wang, and T. C. Green, level converters,” IEEE Electrific. Mag., vol. 3, no. 2, pp. 40–48,
“Reliability model of MMC considering periodic preventive main- Jun. 2015.
tenance,” IEEE Trans. Power Del., vol. 32, no. 3, pp. 1535–1544, [156] S. Du, B. Wu, and N. R. Zargari, “Common-mode voltage elimina-
Jun. 2017. tion for variable-speed motor drive based on flying-capacitor modular
[137] J. Guo, J. Liang, X. Zhang, P. D. Judge, X. Wang, and T. C. Green, multilevel converter,” IEEE Trans. Power Electron., vol. 33, no. 7,
“Reliability analysis of MMCs considering submodule designs with pp. 5621–5628, Jul. 2018.
individual or series-operated IGBTs,” IEEE Trans. Power Del., vol. 32, [157] S. Zhou, B. Li, M. Guan, X. Zhang, Z. Xu, and D. Xu, “Capacitance
no. 2, pp. 666–677, Apr. 2017. reduction of the hybrid modular multilevel converter by decreasing av-
[138] J. Xu, H. Jing, and C. Zhao, “Reliability modeling of MMCs consid- erage capacitor voltage in variable-speed drives,” IEEE Trans. Power
ering correlations of the requisite and redundant submodules,” IEEE Electron., vol. 34, no. 2, pp. 1580–1594, Feb. 2019.
Trans. Power Del., vol. 33, no. 3, pp. 1213–1222, Jun. 2018. [158] M. Espinoza-B et al., “An integrated converter and machine control
[139] P. Tu, S. Yang, and P. Wang, “Reliability- and cost-based redundancy system for MMC-based high-power drives,” IEEE Trans. Ind. Elec-
design for modular multilevel converter,” IEEE Trans. Ind. Electron., tron., vol. 66, no. 3, pp. 2343–2354, Mar. 2019.
vol. 66, no. 3, pp. 2333–2342, Mar. 2019. [159] Y. S. Kumar and G. Poddar, “Medium-voltage vector control induc-
[140] D. Ronanki and S. S. Williamson, “Device loading and reliability anal- tion motor drive at zero frequency using modular multilevel con-
ysis of modular multilevel converters with circulating current control verter,” IEEE Trans. Ind. Electron., vol. 65, no. 1, pp. 125–132,
and common-mode voltage injection,” IEEE J. Emerg. Sel. Top. Power Jan. 2018.
Electron., vol. 7, no. 3, pp. 1815–1823, Sep. 2019. [160] X. Liu, J. Lv, C. Gao, Z. Chen, and S. Chen, “A novel statcom based
[141] S. Du, B. Wu, and N. R. Zargari, “A star-channel modular multilevel on diode-clamped modular multilevel converters,” IEEE Trans. Power
converter for zero/low-fundamental-frequency operation without in- Electron., vol. 32, no. 8, pp. 5964–5977, Aug. 2017.
jecting common-mode voltage,” IEEE Trans. Power Electron., vol. 33, [161] A. F. Cupertino, J. V. M. Farias, H. A. Pereira, S. I. Seleme, and R.
no. 4, pp. 2857–2865, Apr. 2018. Teodorescu, “Comparison of DSCC and SDBC modular multilevel
[142] S. Du, B. Wu, and N. R. Zargari, “A delta-channel modular multi- converters for statcom application during negative sequence compen-
level converter for zero/low-fundamental-frequency operation,” IEEE sation,” IEEE Trans. Ind. Electron., vol. 66, no. 3, pp. 2302–2312,
Trans. Ind. Electron., vol. 66, no. 3, pp. 2227–2235, Mar. 2019. Mar. 2019.
[143] M. M. C. Merlin et al., “The extended overlap alternate arm converter: [162] N. Li, F. Gao, T. Hao, Z. Ma, and C. Zhang, “Soh balancing control
A voltage-source converter with DC fault ride-through capability and method for the MMC battery energy storage system,” IEEE Trans. Ind.
a compact design,” IEEE Trans. Power Electron., vol. 33, no. 5, Electron., vol. 65, no. 8, pp. 6581–6591, Aug. 2018.
pp. 3898–3910, May 2018. [163] M. Quraan, P. Tricoli, S. D’Arco, and L. Piegari, “Efficiency as-
[144] H. R. Wickramasinghe, G. Konstantinou, and J. Pou, “Gradient-based sessment of modular multilevel converters for battery electric vehi-
energy balancing and current control for alternate arm converters,” cles,” IEEE Trans. Power Electron., vol. 32, no. 3, pp. 2041–2051,
IEEE Trans. Power Del., vol. 33, no. 3, pp. 1459–1468, Jun. 2018. Mar. 2017.
[145] S. Liu, M. Saeedifard, and X. Wang, “Zero-current switching control [164] C. Gan, Q. Sun, J. Wu, W. Kong, C. Shi, and Y. Hu, “MMC-based
of the alternate arm HVDC converter station with an extended overlap SRM drives with decentralized battery energy storage system for hy-
period,” IEEE Trans. Ind. Electron., vol. 66,no. 3, pp. 2355–2365, brid electric vehicles,” IEEE Trans. Power Electron., vol. 34, no. 3,
Mar. 2019. pp. 2608–2621, Mar. 2019.
[146] Y. Wang, Q. Song, B. Zhao, J. Li, Q. Sun, and W. Liu, “Quasi-square- [165] J. Wang and P. Wang, “Decoupled power control for direct-
wave modulation of modular multilevel high-frequency DC converter modulation-based modular multilevel converter with improved sta-
for medium-voltage DC distribution application,” IEEE Trans. Power bility,” IEEE Trans. Ind. Electron., vol. 66, no. 7, pp. 5264–5274,
Electron., vol. 33, no. 9, pp. 7480–7495, Sep. 2018. Jul. 2019.