You are on page 1of 8

Adaptive High-Frequency Injection and

Control Loops Design for


Flying-Capacitor Passive Cross-Connected
Modular Multilevel Converter Based Drive Systems
2021 IEEE Energy Conversion Congress and Exposition (ECCE) | 978-1-7281-5135-9/21/$31.00 ©2021 IEEE | DOI: 10.1109/ECCE47101.2021.9595706

Massimiliano BIASON Riccardo BREDA Mattia IURICH


Polytechnic Department of Polytechnic Department of Polytechnic Department of
Engineering and Architecture Engineering and Architecture Engineering and Architecture
University of Udine University of Udine University of Udine
Udine, ITALY Udine, ITALY Udine, ITALY
biason.massimiliano@spes.uniud.it breda.riccardo@spes.uniud.it iurich.mattia@spes.uniud.it

Simone MAZZER Roberto PETRELLA


Polytechnic Department of Polytechnic Department of
Engineering and Architecture Engineering and Architecture
University of Udine University of Udine
Udine, ITALY Udine, ITALY
mazzer.simone.1@spes.uniud.it roberto.petrella@uniud.it

Abstract—One of the main issues associated with the torque, [4]-[6]. The ripple arises from large energy
operation of Modular Multilevel Converters (MMC) as fluctuations, whose amplitude depends on the output current
variable-speed drive is the voltage ripple developed across the amplitude and frequency, and that will eventually diverge
sub-module (SM) capacitors at low speed and high torque when the converter is supplying the machine with dc current,
conditions. A hybrid MMC topology, namely the i.e. at standstill in the case of synchronous machines.
Flying-Capacitor Passive Cross-Connected arms (FC-PCC)
MMC has been recently introduced to cope with this issue. A Usually, the control methods developed for the zero/low-
comprehensive theoretical analysis, as well as a validation of speed operation of MMC-based drives make use of
both the power conversion topology and a new and optimized high-frequency injection strategies that rely on the introduction
distributed digital control strategy are reported in this paper. of large common-mode voltage component for re-balancing the
An adaptive algorithm for the management of the
high-frequency injection adopted to control the voltage ripple
across the SM capacitors is proposed, allowing reduction of
additional losses and guaranteeing enough voltage margin for
proper current regulation in the whole speed/torque range in an
adaptive fashion. The effectiveness of the proposal is
demonstrated by a 22kW permanent magnet synchronous
machine drive, even at rated torque from zero to rated speed.

Keywords—MMC converters, multilevel converters, power


electronics, digital control, electric drives

I. INTRODUCTION
Nowadays, performance and efficiency improvement in
medium voltage industrial motor drives are generally achieved
by the adoption of digitally-controlled multilevel converter
topologies. This choice is generally justified by the possibility to
employ low-voltage power semiconductor devices to obtain a
high-quality output voltage waveform, allowing at the same time
a reduction of the overall power losses in most cases. Moreover,
this solution is effective to mitigate the electromechanical
stresses on the electric machine bearings and windings.
MMC is a quite recent multilevel inverter topology based
on the cascaded connection of elementary cells named sub-
modules (SMs), as depicted in Fig. 1, [1]- [4]. Each cell is
equipped with a local dc energy source, that is implemented
by a capacitor. One of the main issues associated with the
operation of MMC as a variable-speed drive is the voltage
ripple that is developed across the SMs capacitors when the
electrical machine is operated at low speed and high Fig. 1. Phase-leg of a standard modular multi-level converter.

978-1-7281-5135-9/21/$31.00 ©2021 IEEE 1924

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
energy in the converter arms, keeping the capacitor ripple under A. Converter Topology and Structure
control, [6]. The injected high-frequency common-mode voltage The basic structure of a generic FC-MMC phase-leg
can represent a potential reliability problem for the electric ( = , , ) is shown in the left part of Fig. 2: it is split into
machine, since it can drive parasitic bearing currents and stator two arms (the upper and the lower ) with respect to the
windings deterioration, leading to the premature failures. output leg terminal, the leg output current being represented by
Various hybrid MMC topologies have been recently the label . Each arm is split into two sub-arms (the light blue
introduced, [7]-[10], in order to allow the low-frequency boxes) named upper, marked with the subscript 1, and lower,
operation of the MMC converter without the deliberate injection with the subscript 2, respectively. A generic implementation
of large common-mode voltage, at the expense of increased consists of a series connection of half-bridge (HB)
architecture complexity. Among the proposals, the sub-modules per arm, each one is equipped with a capacitor
Flying-Capacitor Passive Cross-Connected arms MMC with a nominal voltage = .
(FC-PCC MMC), [7], appears to be the one that achieves the
objective with the fewest alterations to the conventional The total leg inductance (twice the arm inductance )
architecture, and with the lowest cost. However, for the FC-PCC is split into four elements, each one in series with each sub-arm
MMC design some challenges have to be addressed, mainly and named . The equivalent resistance of each sub-arm,
regarding the control algorithm optimization and implementation. given by the inductance winding resistance and the power
devices on-state resistances is also considered, i.e. .
A comprehensive theoretical analysis, as well as the
validation of a novel adaptive and distributed digital control The flying-capacitor is connected across the two
strategy for FC-PCC MMC are reported in this paper. An midpoints of the upper and lower arms, in order to create an
adaptive algorithm for the management of the high-frequency additional path for the energy flow and introduce an auxiliary
injection (till complete shutdown at high speed) adopted to degree of freedom for the converter control to flatten their
control the voltage ripple across the SM capacitors is proposed, natural imbalance in the low-speed operation of the MMC.
whilst reducing the additional losses and guaranteeing enough The subdivision of the arms into sub-arms is purely formal,
voltage margin for proper current regulation in the whole since disconnecting the flying-capacitor terminals from the
speed/torque range. The effectiveness of the proposal is created mid-nodes would restore the conventional MMC
demonstrated on a 22 kW permanent magnet synchronous converter topology structure.
machine drive, even at rated torque from zero to rated speed. A
prototype of the converter has been designed, in order to allow B. Sub-module Topology
the experimental validation of the proposed control algorithms. A classical half-bridge topology has been adopted for the
implementation of each SM, as shown in Fig. 3. The sub-module
II. FC-PCC MMC TOPOLOGY can be inserted (left) or bypassed (right), thus forming the two
The converter topology and its structure will be briefly levels for the output port voltage, i.e. = , or =0
discussed in this section, in order to highlight the specific respectively. The power switches voltage drops have been
features of the considered MMC architecture. neglected for simplicity.
The choice of HB sub-module topology is twofold: it is an
industry-standard for actual MMC converters, because of the
minimum requirements in terms of power semiconductor
devices, and allowed to test the proposed control algorithms
with the minimum efforts in terms of computational
complexity of the simulation scheme. However, once the
control is completely validated with HB sub-modules, any
other topology can be adopted, as long as an effective
modulation scheme can be implemented and any additional
degree-of-freedom coming from this choice is considered and
modeled. Each SM of the FC-PCC MMC prototype under
development is based on a sandwich structure, comprising a
power board and a universal control board specifically
designed to handle several SM topologies, e.g. half-bridge,
full-bridge (FB) and the neutral-point-clamped (NPC).
(additional details on the concept will be reported later).

Fig. 2. Phase-leg of a FC-PCC MMC (left) and


Fig. 3. Half-bridge submodule and its main operating states:
averaged equivalent circuit model (right).
inserted (left) and bypassed (right).

1925

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
III. FC-PCC MMC MODEL and they result to be in phase with respect to the
high-frequency impressed voltage , ( ), as it will explained
A. FC-PCC Mathematical Model in section IV.B.
As already discussed, the flying-capacitor is added in
order to create a shunt power flow path between the upper and B. FC-PCC Current Dynamics
lower arms, allowing balancing of the sub-modules capacitor A model of the dynamics of the converter currents is needed
voltages without the introduction of any common-mode to properly design a suitable control strategy. By considering the
component, [7][11][12]. Every sub-arm is supposed to inject averaged model shown in the right schematic of Fig. 3 (where
an additional high-frequency voltage term: the equivalent series resistance (ESR) of the flying-capacitor is
neglected), it is possible to apply the Kirchhoff voltage law to
, ( )= , ( )sin(
(1) ) both the outer and inner loops, leading to:
as shown in the equivalent average phase-leg circuit model ,
within one switching period reported in right schematic of − − , − , −
Fig. 2. The general voltage expressions for the upper and ,
(7)
lower sub-arms are reported hereinafter: − − , − , =0

,
, = − − , , = − + , (2) − , − − , − ,
4 2 4 2
,
(8)
= + − = + + − − , =0
, , , , (3)
4 2 4 2
Substituting the expressions for the sub-arms currents (4)
Fundamental frequency and dc voltage components are
and (5), and rearranging the terms to have the forcing
found as in a conventional MMC. The high-frequency voltage
quantities on the left side, the following relations are obtained:
, ( ), is introduced alternatively in-phase and out-of-phase by
the sub-arms and adds a degree of freedom for the low-frequency −( + ) ,
power balancing. It is worth noticing that those components = + , (9)
2
cancel out in the total arm voltage and hence will not produce
any effect on the output voltage, nor on the dc-bus voltage. −2 −( −,
)
=2 +2 ,
(10)
The integration of the flying-capacitor into the converter 2
architecture allows the realization of two additional paths, where where and are the total voltages inserted in the outer
two high-frequency currents , (inner) and , (outer) can flow, loop and inner loop sub-arms, respectively.
as shown in the right circuit of Fig. 2. Assuming balanced Two relevant results can be drawn from (9) and (10):
impedances of the arms and sub-arms and the same high-frequency
voltage perturbations, the overall flying-capacitor current will be  the dynamic behavior of , is exactly the same as found
= , + , = 2 , , where , = , = , . for the conventional MMC topology, therefore allowing all
the control strategies already known to be applied, [5][6];
The sub-arm current expression can be computed starting
from the conventional MMC leg model, where the  the control of the low-frequency leg circulating current
fundamental frequency and dc current components are and high-frequency ones , is naturally fully-decoupled,
preserved, and the high-frequency terms are added. Therefore, since these two are controlled by different combinations of
the currents that flow in the upper and lower sub-arms are: the voltages inserted by the sub-modules.
C. Compensation of the low-frequency energy fluctuations
, = , + + , , = , + − , (4) Voltage ripple across SM capacitors are mainly due to the
2 2
low-frequency energy fluctuations, related to the output
, = , − − , + , , (5) = , − voltage and current of the leg (feeding the electric machine).
2 2 The interaction of the high-frequency voltage and current
where the term , has been defined to include, in a components can be adopted to compensate for those
steady-state balanced operation, both the leg common-mode fluctuations, as reported hereinafter. In order to find out the
dc current and the low-frequency leg circulating current, appropriate values of the amplitude of the high-frequency
components , (for the voltage) and , (for the current),
The high-frequency injected currents flow through the it is necessary to evaluate the overall power exchanged by
flying-capacitor and interact with the high-frequency voltages each (upper and lower) arm:
introduced by the sub-arms, thus causing a power flow
between the upper and lower arms. In particular, the power is
exchanged with the top sub-arm of the upper arm and the = − − , , + + , +
4 2 2
bottom sub-arm of the lower arm, which are part of the outer
loop. Similarly for the bottom sub-arm of the upper arm and + − + , , + − ,
4 2 2
the top sub-arm of the lower arm, being part of the inner loop.
The combination of these interactions allows to control the = , − + (11)
power exchange from the upper and lower arm and vice versa, 2 2
hence to compensate for any energy unbalance between them. ,

The high-frequency currents are controlled to be sinusoidal, i.e.


+ − , − , , + , , sin(2 )
( )= ( ) sin(ω ) 4
, , (6) ,

1926

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
= + − , , − − , +
4 2 2
+ + + , , − + ,
4 2 2
= , − + (12)
2 2
,

− − , − , , + , , sin(2 )
4
, Fig. 5. Adopted distributed control layers hierarchy for FC-PCC MMC.
A remarkable results emerges from (11) and (12), i.e. the A novel adaptive algorithm for the management of the
arm power ( ) can be split into two components, namely a high-frequency injection (till complete shutdown) has been
common-mode ( , ) and a differential-mode ( , ) , developed, aiming at optimizing the control of the voltage ripple
identical for both the arms. The first component is responsible across the sub-module capacitors, whilst reducing the additional
for deviating the average value of the SM capacitor energy of losses and guaranteeing enough voltage margin for machine
the entire phase leg, whereas the second one accounts for an current regulation.
imbalance of the average energy stored by the SM capacitors
of one arm (upper or lower) compared to the other. As a IV. CONTROL SYSTEM ARCHITECTURE
result, the presence of these two components deviates the SM The high-level functional block diagram of the overall drive
capacitor voltages from the rated value. In order to avoid this control system is sketched in Fig. 4, where the electric machine,
issue, it is necessary to eliminate both power contributions by the power converter and a certain number of logical blocks
exploiting the available degrees-of-freedom, i.e. related to the control algorithm are shown. The system exploits
low-frequency and high-frequency voltage and currents. a distributed control approach, with one central microcontroller
Forcing the common-mode term to zero: and a number of decentralized computational units, each one
situated within one SM. Five control tasks, namely , , ,
, = , − =0 and , can be identified and organized as shown in the layer
2 2
(13) hierarchy of Fig. 5, where a brief description of each one and the
⇒ , = corresponding control domain are reported. The overall
reference voltage at each SM is formed by the sum of each
As far as the differential-mode power component control task output, being either calculated by the central
concerns, zeroing the last term of (11) (or (12)) provides a processing unit or by each SM, and then actuated through a
mean to calculate the amplitude of the controlled phase-shifted-carrier pulse-width-modulation (PSC-PWM) at
high-frequency current in accordance with the injected voltage SM level. The corresponding data paths of the various control
term amplitude , , i.e.: voltage components are shown Fig. 6. Subscripts , , , are
referring to the output phase ( = , , ) of the converter, to the
upper or lower arm ( = , ), to upper or lower sub-arm ( =
, = − , − , , =0
4 1,2), and to the SM within a sub-arm ( = 1,2,3, …).
− (14) The adoption of a distributed control system for a modular
4 ,
⟹ , = topology such as the MMC offers various advantages:
,
 separation of concerns, enables faster processing as
The higher the amplitude of the high-frequency injected compared to either centralized or decentralized approach.
voltage, the lower would be the corresponding peak (and Indeed, it is possible to employ a high-performance
RMS) value of the high-frequency circulating current microcontroller for the centralized tasks and low-cost
required for the power compensation, thus reducing the Joule MCUs in the SMs for decentralized processing, thus
losses on the converter leg. The amplitude of the injected parallelizing a huge amount of control operations;
voltage is limited by the converter input dc voltage and
constrained by the voltage required to control the electric
machine in the whole torque/speed range.

Fig. 6. Distributed control architecture: the sources of each reference


Fig. 4. Control system high-level functional block diagram. voltage component fed to the SM PSC modulator are highlighted.

1927

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
 full scalability, allowing to realize a MMC with an A. Leg common-mode energy control (task )
arbitrary number of sub-modules in the phase-leg (in The aim is to control the energy stored in the phase-leg by
principle, provided that the required throughput of the regulating the average voltage across the SMs capacitors to
communication bus is actually achievable). the nominal value ∗ = ⁄ . The overall control block
On the other hand, this brings also a challenge to address diagram consists of two nested loops, as represented in Fig. 7.
for the effective implementation of the control strategy. The The outer one regulates the average value of the SM capacitor
various entities, SMs and central unit, have to exchange voltages (as calculated by the feedback branch) to the set-point
information, being measurements, partially processed data or value ∗ by selecting the proper level of leg common-mode dc
voltage references to be synthesized with accurate timing current and low-frequency leg circulating current, i.e. the
constraints. Therefore, a reliable high-speed communication reference ∗ , , being fed to the inner loop. A feed-forward term
link and tight synchronization among all the microcontrollers
are required. Moreover, being the SMs series-connected, their , is added to the output of the common-mode voltage regulator
control units are referred to potentials that depend on their according to (13), providing an instantaneous action
position in the converter leg. In general, a galvanically isolated (subscript is related to the energy control performed by the
link should be realized for exchanging information between outer loop). The fine-tuning relies to the regulator to cope with
the different control units. This would introduce additional inevitable errors due to a pure feed-forward action.
challenges to overcome, in order to guarantee the required The inner loop regulates the current , by synthesizing
performances for the communication bus.
the appropriate voltage perturbation to the sub-arms
A summary of the different operations implemented by voltage, i.e. (2) and (3). The equivalent plant model needed to
each control task, the sources of information required for the tune the common-mode current regulator can be obtained
processing and needed communication requirements are directly by (9), as the injected voltage perturbation is a forcing
reported hereinafter: term for the current , . The differential equation in time
 Control task : implements the field-oriented-control of domain and the corresponding transfer function in the Laplace
the electric machine (both current and speed loops); it is domain are reported hereinafter.
executed by the central μC only, and requires the ,
measurement of the machine phase currents and rotor = + , (15)
position; the computed reference voltage component is then
transmitted to each SM to form the overall reference ; , ( ) 2
(s) ≜ = (16)
( ) +
 Control task : implements the leg common-mode energy
control; it is executed by the central μC only; it requires The injected current , is then responsible for the voltage
the measurements of each SM capacitor voltage and each across each SM capacitor based on the duty-cycle , , of
sub-arm current, being sent by the SMs through the the module. The control design takes also the control and
communication bus; the computed reference voltage PSC-PWM actuation delays into account.
component is then transmitted back to each SM to form the
overall reference ; B. Arm differential-mode energy control (task )
The aim is to regulate the differential-mode power
 Control task : implements the arm differential-mode components reported in (11) and (12), causing the energy
energy control; a first part is computed by the central μC, stored in the converter arms to be different from one to the
that is sent to each SM to complete the processing and to other. The overall control block diagram consists, also in this
form the overall reference ; similarly to task , it case, of two nested loops, as represented in Fig. 8.
requires the measurements of each SM capacitor voltage
and each sub-arm current; The outer loop regulates the energy difference stored in the
converter arms, i.e. , to zero, by selecting the proper level of
 Control tasks and : implement sub-arm and SM the high-frequency current, i.e. the reference ∗ , being fed to the
equalization controls, respectively; they are executed at SM inner loop. The interaction of this current component with the
level, based on local measurements; computed reference
high-frequency voltage , , (2) and (3), establishes the required
voltage components are used to form the overall reference .
power flow to counteract the differential-mode capacitor voltage
The detailed analytical developments of the control tasks fluctuations (i.e. 2 ). The reference current ∗ , is a sine-wave
will be limited to tasks and , since the FOC ( ) is well at the injection frequency, whose amplitude is dynamically
known in literature, and and have the structure of a regulated by the differential-mode energy regulator . Also in
classical control loop present in almost all the PWM MMC this case, a feed-forward term , is considered to provide an
converters, [6][11][15]. instantaneous action, according to (14), and fine-tuning is left to
the regulator.

Fig. 7. Leg common-mode energy control loop (task ).

1928

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
Fig. 8. Arm differential-mode energy control loop (task ).
The inner loop regulates the current , by synthesizing choice is to maximize the voltage in order to reduce the current,
the appropriate voltage perturbation , between the two
thus limiting the Joule losses in the circuit. Voltage amplitude
sub-arms of the same arm, with the appropriate phase, i.e. its , can be dynamically maximized by taking into account the
sign is the same as the upper-arm high-frequency voltage , , actual available voltage, i.e. the sub-arm maximum voltage
whereas for the lower arm has an opposite contribution, in minus the output phase voltage, [12][14]:
order to create a voltage imbalance that establishes a current
flow (see also the equations reported in Fig. 6). , = − (20)
4 2
The equivalent plant model needed to tune the
In practice, only a certain fraction of the actual available
differential-mode current regulator can be obtained directly
voltage (20) can be allocated for the high-frequency injection,
by substitution of the voltage terms in (10). The differential
since a certain amount should be considered as control margin
equation in time domain and the corresponding transfer function
for the other regulation loops discussed above.
in the Laplace domain are reported hereinafter (ESR neglected).
The need to change the injected high-frequency voltage
, , , amplitude as a function of the operating conditions of the machine,
= + + , (17)
results in a time-varying plant for the arm differential-mode
( ) energy control loop, (19). Therefore, an adaptive approach is
,
(s) ≜ = (18) proposed here in order to maintain the loop gain constant, thus
, ( ) +2 + ensuring the best dynamical performance and stability over the
where = and = . entire speed/torque range of the machine, as shown in Fig. 10.
C. Arm differential-mode energy control loop:
Equation (18) suggests the resonant nature of the circuit path high-frequency injection shutdown
created by the flying-capacitor between the midpoints of the
The amplitude of the capacitor voltage fluctuations is directly
phase-leg. The switching-cycle averaged phase-leg equivalent
proportional to the amplitude of the output current and inversely
circuit is shown in Fig. 9. In order to minimize the magnitude of
proportional to the output frequency, [5][6]. The value of the SM
the voltage perturbation , required to drive the current, the capacitance is designed to absorb most of the power fluctuations
injection frequency is selected in order to match the resonance when the output frequency is high, thus limiting the ripple voltage
frequency of the equivalent series RLC circuit, [5][13][14], to a tolerable value without any control action.
thus assuring that high-frequency voltage and current are in-phase.
High-frequency injection is required at low output
In order to minimize the size, and thus the cost, of passive frequency. However, its action needs to be adapted for two
components, it is necessary to select the highest possible main reasons at least, namely avoiding unnecessary Joule
injection frequency, by taking into account the limitations of the losses coming from the injected current, and increasing the
digital control, e.g. one-tenth of the switching frequency. Then, available voltage for output current control.
the arm inductance is chosen to limit the leg circulating currents,
to mitigate the effect of the capacitor voltage ripple and to A gradual turn-off of the high-frequency components
perform the required filtering action on the output voltage. As a should be ensured, in order to avoid discontinuities in the
result, the flying-capacitor capacitance can be designed. control loops and in the converter response. The solution
proposed in this paper is recalling this need, but introduces two
The red box in Fig. 8 highlights the plant of the outer loop, different smooth weighting functions for high-frequency
whose transfer function is: voltage and current injections, that progressively reduce the
( ) 2 , contribution of the arm differential-mode control until
(s) ≜ = − (19) complete shutdown as a function of the machine speed. The
, ( ) s control scheme is reported in Fig. 10, where the shutdown
It can be highlighted that loop gain is a function of the block is included in the direct branch.
amplitude of the high-frequency injected voltage , , (1), An original strategy is proposed to delay the voltage shutdown
therefore affecting the design of the outer loop regulator . profile with respect to the current one, as shown in Fig. 11. This
Also, since the differential-mode energy is controlled by means allows a sustained voltage injection, ensuring an adequate power
of the power resulting from the interaction between the transfer during the current shutdown transient. In this way, the
high-frequency injected voltage , and current , , a good differential-mode power component can be effectively
compensated, thus keeping the capacitor voltage ripple limited.

Fig. 10. Adaptive differential-mode energy control loop


Fig. 9. Averaged phase-leg equivalent circuit at high-frequency. with high-frequency injection shutdown strategy.

1929

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
high-frequency injection is turned off in both cases. Therefore,
the behavior of the capacitor voltage is equivalent, since the
differential-mode energy compensation is not active anymore.
Finally, the electric machine currents are reported in Fig. 15,
both in the synchronous and in the stationary three-phase
reference frames. The waveforms result smooth and with limited
distortion, due to the high number of voltage levels, leading to a

Fig. 11. High-frequency injection shutdown profiles.

V. SIMULATION RESULTS
An accurate simulation model of the FC-PCC MMC drive
was built to validate the effectiveness of the proposals and the
design choices. The complete implementation constraints has
been included to study and optimize the control architecture and
to gain useful recommendations for the final hardware
implementation. Among the most important aspects, the Fig. 12. Torque/speed response: electrical speed (top) (reference: green,
following ones have been considered: the distributed digital actual: red) and torque (bottom) (reference: green, actual: red, load: blue).
control architecture and topology; the control and PSC-PWM
delays (which are different for each carrier); the synchronization
and communication temporization of the SMs, which are based
on a forecast of the latency and spread of data transmission and
processing, in turn related to the type of adopted communication
bus and protocol, and to the specification on the galvanic
isolation. Most part of this aspects are not covered in this paper
due to space limitations. Main parameters of the developed
converter are reported in TABLE I. A non-salient permanent
magnet synchronous machine has been considered for the test,
whose parameters are reported in TABLE II.
A comprehensive test pattern has been designed to validate
control effectiveness in the most representative operating Fig. 13. High-frequency injected voltages (top) and current amplitudes
conditions, i.e. during speed and torque transients, including (bottom) (averaged values within a switching period).
torque disturbance grasping and releasing, even at zero-speed. A
series of figures will be reported showing the behavior of
relevant control variables within the same time range.
Torque and speed responses are reported in Fig. 12, proving a
correct control in all the considered conditions. Particularly
interesting is the possibility of correct operation of the machine at
high load torque with zero- or low-speed, thanks to the adopted
converter topology and high-frequency injection scheme.
The corresponding injected voltages and currents
averaged values within a switching period for the three
converter legs are shown in Fig. 13. It is possible to
appreciate the variation of the amplitudes in the different
conditions and the proper operation of the proposed shutdown Fig. 14. SMs capacitors voltages: adaptive (green) and non-adaptive gain (red)
strategy at increasing speeds. The effect of the turn-off
sequence based on delayed smooth weighting functions is
visible, leading to the current profile going to zero faster than
the voltage one, as explained in section IV.C.
In Fig. 14 the capacitor voltages of the SMs of leg are
reported to validate the proposed balancing strategies. The
voltage ripple is limited to less than 10% of the rated SM
voltage during all the test pattern. The green curves are obtained
with the proposed adaptation of the arm differential-mode
energy control (and shutdown strategy), as discussed in section
IV.B. A further reduction of the voltage ripple is achieved with
the proposed solution with respect to the standard fixed-gain
regulation. It is worth mentioning that at higher output Fig. 15. Electric machine phase currents: and (top and middle,
frequencies (i.e. near the end of the speed transient) the reference: green; actual: red), three-phase output currents (bottom).

1930

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.
TABLE I. MMC BASED DRIVE PARAMETERS
Rated dc-bus voltage, 800 [V]
Number of SM per arm, 8
SM capacitor rated voltage, 100 [V]
Rated peak phase voltage, 400 [V]
Rated peak phase voltage, 75 [A]
Total arm inductance, 1 [mH]
Total arm resistance, 10 [mΩ]
Rated SM capacitance, 3900 [μF]
Rated flying capacitance, 50 [μF]
PWM carrier frequency, 10 [kHz]
HF injection frequency, 1 [kHz]
Fig. 16. SM hardware prototype (3D rendering). TABLE II. SPMSM RATED PARAMETERS
fine generation of the output phase voltage and an accurate Rated torque, 62.7 [Nm]
control of the machine currents. Also, it is worth mentioning that Rated speed, 4000 [rpm]
the waveforms are not affected by the high-frequency voltage Pole pairs, 3
Rated current, 52 [A ]
and current injections, as already recalled.
Phase resistance, 54 [mΩ]
Phase inductance, = 470 [μH]
VI. CONVERTER CONCEPT
PM flux linkage, Λ 0.187 [Wb]
Development of the prototype of the proposed converter is
on-going. A concept 3D rendering of one SM assembly is shown REFERENCES
in Fig. 16, featuring a very compact and modular structure. The [1] R. Marquardt, A. Lesnicar, J. Hildinger, et al. “Modulares
sandwich structure includes: an insulated-metal-substrate (IMS) Stromrichterkonzept für Netzkupplungsanwendungen bei hohen
power board (bottom), hosting the power semiconductors and Spannungen,” in ETG-Fachtagung, Bad Nauheim, Germany, 2002.
interconnections (both power, gate driving and measurements); [2] S. Debnath, J. Qin, B. Bahrani, M. Saeedifard, and P. Barbosa, "Operation,
a control board (middle), hosting the SM μC, the gate drivers and Control, and Applications of the Modular Multilevel Converter: A Review,"
communication devices; and the capacitor board (top, rear). An IEEE Trans. on Power Electron., vol. 30, no. 1, pp. 37-53, Jan. 2015.
additional board (top, front) is used to interface the SM control [3] A. Dekka, B. Wu, R. L. Fuentes, M. Perez, and N. R. Zargari,
"Evolution of Topologies, Modeling, Control Schemes, and
electronics with the central controller via a new generation Applications of Modular Multilevel Converters," IEEE J. Emerg. Sel.
high-speed optical bi-directional transceivers. Top. Power Electron., vol. 5, no. 4, Dec. 2017.
[4] A. Lesnicar and R. Marquardt, "An innovative modular multilevel
VII. CONCLUSIONS converter topology suitable for a wide power range," in Proc. of the
This paper provides insights and original proposals on the IEEE Bologna Power Tech Conference, 2003.
optimized distributed digital control for FC-PCC MMC for [5] A. J. Korn, M. Winkelnkemper, and P. Steimer, "Low output frequency
operation of the Modular Multi-Level Converter," in Proc. of the IEEE Energy
electric drive system. An in-depth analysis has been carried
Conversion Congress and Exposition, Atlanta, GA, pp. 3993-3997, 2010.
out to identify the power flow components responsible for the
[6] J. Jung, H. Lee, and S. Sul, “Control Strategy for Improved Dynamic
capacitor voltage ripple of the converter at low output Performance of Variable-Speed Drives With Modular Multilevel
frequency and a hierarchical control structure has been Converter,” IEEE J. Emerg. Sel. Top. Power Electron., June 2015.
developed to allow operation of the drive system in the whole [7] S. Du, B. Wu, N. R. Zargari and Z. Cheng, "A Flying-Capacitor
torque/speed range. A distributed implementation of the Modular Multilevel Converter for Medium-Voltage Motor Drive,"
hierarchical architecture has been proposed, considering IEEE Trans. on Power Electron., vol. 32, no. 3, March 2017.
actual implementation constraints. A novel adaptive algorithm [8] S. Du, B. Wu, K. Tian, N. R. Zargari, and Z. Cheng, "An Active Cross-
for the management of the high-frequency injection needed to Connected Modular Multilevel Converter (AC-MMC) for a Medium-
Voltage Motor Drive," IEEE Trans. on Ind. Electron., vol. 63, no. 8,
control the voltage ripple across the SM capacitors has also pp. 4707-4717, Aug. 2016.
been proposed, providing excellent results in controlling the [9] D. D. Le, and D. -C. Lee, "A Modified Hybrid Modular Multilevel
capacitor voltage fluctuations under the maximum target Converter with Reduced Capacitor Voltage Fluctuations and Fault-
ripple, i.e. the 10% of the SM nominal voltage. Additionally, Tolerant Operation Ability," in Proc. of the 10th International
the dynamic shutdown procedure allows the reduction of Conference on Power Electronics and ECCE Asia, 2019.
supplementary Joule losses, and, at the same time, it [10] B. Li, Y. Zhang, G. Wang, W. Sun, D. Xu, and W. Wang, "A Modified
guarantees an adequate voltage margin for proper machine Modular Multilevel Converter With Reduced Capacitor Voltage
Fluctuation," in IEEE Trans. on Ind. Electron., vol. 62, no. 10,
operation in the whole speed/torque range. pp. 6108-6119, Oct. 2015.
The effectiveness of the proposal is demonstrated by an [11] M. Hagiwara, and H. Akagi, "Control and Experiment of Pulsewidth-
accurate simulation model, where all the implementation Modulated Modular Multilevel Converters," IEEE Trans. on Power
Electron., vol. 24, no. 7, pp. 1737-1746, July 2009.
constraints have been considered, aiming at studying and
[12] Q. Song, W. Liu, X. Li, H. Rao, S. Xu, and L. Li, "A Steady-State
optimizing the control algorithms and architecture, and to gain Analysis Method for a Modular Multilevel Converter," in IEEE Trans.
useful recommendations for the final hardware implementation on Power Electron., vol. 28, no. 8, pp. 3702-3713, Aug. 2013.
too. Correct operation in standard, as well as critical operating [13] D. D. Le, and D. Lee, "Integration of Coupled Inductors for Compact Design
conditions (i.e. rated torque from zero- to rated-speed) is of Flying-Capacitor Modular Multilevel Converters," in Proc. of the IEEE
demonstrated. Results highlight both the capacitor voltage Energy Conversion Congress and Exposition, Baltimore, MD, USA, 2019.
ripple reduction, as well as the absence of distortion of output [14] L. M. Cunico, G. Lambert, R. P. Dacol, S. V. G. Oliveira, and Y. R. de
waveforms, allowing unaffected machine operation. Novaes, "Parameters design for modular multilevel converter (MMC),"
in Proc. of the Brazilian Power Electronics Conference, 2013.
Development of the prototype of the proposed converter is [15] S. Debnath, and M. Saeedifard, "Optimal control of modular multilevel
on-going. A concept 3D rendering of one SM assembly has been converters for low-speed operation of motor drives," in Proc. of the IEEE
reported, featuring a very compact and modular structure. Applied Power Electronics Conference and Exposition, pp. 247-254, 2014.

1931

Authorized licensed use limited to: UDESC - Universidade do Estado de Santa Catarina. Downloaded on February 02,2022 at 00:12:19 UTC from IEEE Xplore. Restrictions apply.

You might also like