See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350744558

Circulating Current Control for the Modular Multilevel Matrix Converter Based
on Model Predictive Control

Article in IEEE Journal of Emerging and Selected Topics in Power Electronics · April 2021
DOI: 10.1109/JESTPE.2021.3071964

CITATIONS READS

22 1,195

4 authors:

Matías Urrutia Roberto Cardenas

Sprint Electric ltd University of Chile
21 PUBLICATIONS 254 CITATIONS 215 PUBLICATIONS 9,312 CITATIONS

SEE PROFILE SEE PROFILE

Jon C Clare Alan J Watson

University of Nottingham University of Nottingham
461 PUBLICATIONS 19,226 CITATIONS 143 PUBLICATIONS 3,613 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Matías Urrutia on 10 April 2021.

The user has requested enhancement of the downloaded file.

JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 1

Circulating Current Control for the Modular

Multilevel Matrix Converter Based on Model
Predictive Control
Matı́as Urrutia, Student Member, IEEE, Roberto Cárdenas, Senior Member, IEEE, Jon Clare, Senior
Member, IEEE, and Alan Watson, Senior Member, IEEE

Abstract—In this work, a Continuous-Control-Set Model Pre-

dictive Control (CCS-MPC) strategy, with a saturation scheme
for protection, is presented for regulating the circulating currents
of a Modular Multilevel Matrix Converter (M3C). The proposed
scheme is based on a state space model of the M3C and
allows protection and better utilisation of the devices through
a saturation scheme, which directly limits the arm currents and
cluster output voltages by integrating the corresponding bounds
as constraints of the CCS-MPC scheme. In order to solve the
inherent optimisation problem associated to the CCS-MPC, an
active-set algorithm is implemented. Experimental and simulation
results from a 27-cell M3C prototype validate the proposed
strategy and illustrate the good performance achieved with the
methodology presented in this work.
Index Terms—Modular Multilevel Matrix Converter, Model
Predictive Control, Active-set Algorithm, Circulating Currents.
Fig. 1. Modular Multilevel Matrix Converter (M3C).

I. I NTRODUCTION that the M3C will replace the thyristor-based cyclo-converters.

An overview of M3C applications is presented in [13].
The Modular Multilevel Matrix Converter (M3C) is an The control of the M3C is a challenging task due to
important power electronic topology that belongs to the Mod- the large number of floating capacitors and the complex
ular Multilevel Converter (MMCs) family [1], [2], [3], [4], arm interconnections [4]. The control schemes reported are
[5]. The M3C topology is composed of several full-bridge typically implemented using three cascaded loops [9], [3], [2],
power cells (shown at the right-side of Fig. 1), which are [14] employing different control bandwidths. The three loops
grouped into nine power clusters. The serial connection of are respectively designed to regulate the Cluster Capacitor
a cluster with an inductance, is referred to as an arm (shown Voltages (CCVs), to control the arm-currents, and to locally
at the left-side of Fig. 1). The M3C has been proposed as balance the capacitor voltages in each cluster.
a prominent topology for low-speed high-torque drives since To achieve energy balance in the M3C, i.e. to regulate the
lower amplitude circulating currents are required compared to floating capacitor voltages of the nine clusters, four circulating
those utilised by other MMCs, particularly when the electrical currents are utilised as degrees of freedom [3], [4], [15].
machine is operating at low speed [6], [7], [8]. It has also Usually these currents have two or more frequency compo-
been proposed for Wind Energy Conversion Systems (WECS) nents [15], [2], [16], and typically Proportional (P) controllers
based on Permanent Magnet Synchronous Machines [9], for have been proposed for regulating them [4], [7], [17], [15].
improved fault ride-through control of doubly-fed induction Although P-controllers have shown adequate performance in
generators [10], for low frequency transmission systems [11] some applications, they cannot provide good tracking for
and , as discussed in [12], in the future it is highly possible sinusoidal references unless the gain is increased, which
This manuscript was submitted on the 5th of September 2020 and accepted introduces a trade-off between improvement in tracking error
on the 29th of March 2021. The work of Matı́as Urrutia was supported and an undesirable amplification of the electrical noise in the
by CONICYT- PCHA/Doctorado Nacional/2018-21181386, and the work of measurements. Moreover, as typically happens in most of the
Roberto Cárdenas was supported by Fondect Nr. 1180879 and Basal Project
FB0008. practical control systems, a large increase in the control gain
Matı́a Urrutia is with the Department of Electrical Engineering, Universidad may move the closed loop poles to poorly damped positions
de Chile, Santiago 8370451, Chile and the PEMC Research Group, Faculty of which certainly affect the dynamic response of the system. To
Engineering, University of Nottingham, Nottingham NG7 2RD, UK.(e-mail:
matias.urrutiaortiz@nottingham.ac.uk). avoid the inherent tracking problem of P- and PI-controllers
Roberto Cárdenas is with the Department of Electrical Engineering, Uni- in the presence of sinusoidal references, in [18] it is proposed
versidad de Chile, Santiago 8370451, Chile (e-mail: rcd@ieee.org). the utilisation of multi-resonant controllers. However, multi-
Jon Clare and Alan Watson are with the PEMC Research Group, Fac-
ulty of Engineering, University of Nottingham, Nottingham NG7 2RD, UK resonant controllers have to be tuned online when the input
(jon.clare@nottingham.ac.uk and alan.watson@nottingham.ac.uk). and/or output frequencies are varying, which implies that some
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 2

sort of self-tuning algorithm is required (see [18], [19]). successfully integrated into the control system. However, the
In addition to the requirement for satisfactory reference- implementation of the saturation scheme was not comprehen-
tracking with high-bandwidth for the circulating current con- sively analysed , and validation of the proposed methodology
trol, a saturation scheme should be included to ensure effective was realised only by simulation.
protection of the converter. The total current in each arm of As reported in previous publications, the control of modular
the M3C has several components, i.e a contribution from the multilevel converters for drive applications is typically realised
input and output currents, and a contribution from each of the utilising different control algorithms for different operating
four circulating currents. Additionally, the circulating currents points of the electrical machine. For instance, in [28], the
usually have (at least) two frequency components and they control of the Modular Multilevel Converter (M2C) is divided
could have both positive and negative sequences. Therefore, into the Low-Frequency Mode (LFM) and the High-Frequency
as demonstrated in Section V.A.3, it is very challenging to Mode (HFM) [28]. In the LFM, the addition of common
implement protection algorithms for the total arm currents mode voltage is usually required as well as the utilisation of
and cluster voltages, which are required, for instance, to mitigation currents. A similar separation of control algorithms
ensure that the thermal limit of the devices in each arm is is utilised for the M3C, where the control system is usually
not exceeded. The arm current control schemes reported in divided into the Different Frequency Mode (DFM) and Equal
the literature saturate each voltage and current component Frequency Mode (EFM) (see [5], [15], [17]). Again in the
separately, e.g. saturating the input current vector, output EFM, common mode voltages and mitigation currents are
current vector, positive and negative sequence components of required. Nevertheless, is should emphasised that in a drive
the circulating currents etc. using heuristic criterion [15], [2], applications, e.g. when the M3C drives a high power induction
[16]. This is certainly a sub-optimal solution because limiting machine, the DFM represents most of the machine operating
each component independently does not ensure that the arm range (see [5], [15], [13]).
current or voltage is adequately limited or, conversely, fully In this paper a new CCS-MPC algorithm is presented, where
exploited. the scheme is designed to regulate the circulating current in the
Recently, the application of Model Predictive Control M3C during DFM Operation [15], using a State Space Model
(MPC) schemes for the control of power electronic converters (SSM) approach [16], [29]. This approach allows a linear
and drives has been reported in the literature. Most of the representation of the M3C using state variables, simplifying
proposed algorithms are based on Finite-control set MPC the incorporation of the constraints (for instance the current
(FCS-MPC) and, more recently, Continuous-Control Set MPC limits) into the CCS-MPC strategy. To solve the CCS-MPC
(CCS-MPC) methodologies have been reported [20], [21], problem, an active-set algorithm [30], [31] is integrated into
[22], [23], [24]. A well-known advantage of MPC algorithms the proposed scheme. The benefits of the proposed MPC
is the straightforward integration of system non-linearities and are its straightforward implementation, due to the simple and
strong constraints into the scheme [4], [25]. However, when linear nature of the circulating current dynamic relationships,
applied to MMCs, FCS-MPC strategies have an unfeasible and its low computational complexity. Furthermore, although
computational burden, due to the very large number of switch- input/output port current regulation is not included in the
ing states [20] available in a typical M3C. For this reason, proposed MPC scheme, saturation limiting of the arm currents
modified FCS-MPC algorithms for MMC applications have by adjusting the circulating currents alone is successfully
been proposed, with a restricted number of switching states addressed and it affects neither the input nor output port
at each sampling time to limit the searching space and to currents. To the best of the authors’ knowledge, this control
reduce the computational burden. This approach complicates strategy approach has not been previously reported in the
the analysis and implementation of the algorithm and does literature.
not ensure an optimal solution. Moreover, another drawback Summarising, the main contributions of this work are:
of FCS-MPC is the variable switching frequency. Consider- • A new CCS-MPC implementation for regulating the
ing these disadvantages, there has been little discussion of circulating currents of the M3C including constraints is
the applications of FCS-MPC to M3C in the literature (see presented, fully analysed and experimentally validated.
[26]) and no experimental results have been reported for this • A saturation (protection) scheme is integrated with the
methodology. proposed CCS-MPC algorithm to avoid operating with
Regarding CCS-MPC strategies, in [27] it is reported an currents and voltages outside the predefined limits. A
optimising approach for regulating the capacitor voltage os- suitable Active Set Algorithm is programmed to achieve
cillations of the M3C. This strategy is based on a cascaded these goals. Experimental validation is provided.
(two step) optimisation methodology, where the manipulated • The protection scheme, which solely adjusts the cir-
variables are the common-mode voltage and the circulating culating currents, is hidden from the input and output
current references. The performance of the proposed scheme is port voltages and currents. Therefore, the port control
good and it gives an adequate dynamic performance. However, performance is unaffected on both sides.
the two-step optimisation methodology does not necessarily Other applications are also feasible considering that regulation
produce a globally optimal operating point. Therefore it has of the arm currents and circulating currents are also required
the disadvantage of reaching sub-optimal solutions. Another in other converter topologies such as, for instance, the M2C
reported implementation of CCS-MPC for M3C applications (see [13], [29]). Therefore the MPC strategy proposed in this
is discussed in [16]. In this publication a saturation scheme is work could be modified to address other challenges.
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 3

The rest of this work is organised as follows: modelling with

2 -1 √ -1 0 0 0
 
of the M3C, including an overall discussion of the proposed √
0 3 - 3 0 0 0 
control system is presented in Sections II and III. In Section o 12 2 2 0 0 0  uvw

vαβ0 =  v
IV the proposed control strategy for regulating the circulating 6 0 0 0 2 √
-1 √  rst
-1 
0 0 0 0 3 - 3 
currents of the M3C is analysed. Simulation and experimental
0 0 0 2 2 2
results are discussed in Section V. Finally an appraisal of the
To
proposed methodology is presented in Section VI.
as the transformed input/output port voltage vector.
II. M ODELLING OF THE M3C
The M3C modelling approach presented in this section is B. Output-Power/Capacitor-Voltage Model of the M3C
based on a state space representation of the converter that is
The energy of the j-th cluster of the M3C is defined as:
derived from [16], [29]. The resulting linear model allows a
n
simple representation of the dynamics of the input/output port 1X j 2
and circulating currents of the M3C. Additionally, the state WCj = C(vCr ) (4)
2 r=1
space model enables implementation of MIMO-based control
techniques, for instance, Model Predictive Control, Kalman where it is assumed that the cluster composed of n cells having
filters and sensor-less approaches. the same capacitance. Conveniently, the cluster stored energy
Modelling of the M3C is based on two decoupled dynamic can be rewritten as:
sub-models: The Current-Voltage Dynamic Model and The 1
Output-Power/Capacitor-Voltage Model. These are discussed WCj = CψCbj , (5)
2
below. Pn j 2
with ψCbj = r=1 (vCr ) as the sum of the square of the
A. Current-Voltage Dynamic Model capacitor voltages from the j-th cluster. With this definition,
Using the references shown in Fig. 1, the dynamics of the the following relation is stated for the nine clusters:
arm currents are defined by:
ψ 
pb1
1  
Cb1
ib1  1 0 0 -1 0 0 v 
b1 1 ψ
 Cb2  pb2 
i b2 1 0 0 0 -1 0 v  v
i 
 b3  1 0 0 0 0  vuv
-1
b2 
vb3 
1
  ψCb3  pb3 
i 1 ψ  p 
d  b4 
  1  1 0 -1
 0 0 0   1
v 
vb4  1
d  Cb4  2  b4 
  vwr  −
v   
ib5  = 0 1 0 0 -1 0  b5  − 1 vcom , ψCb5  = pb5  (6)
dt ib6  Lb 0 1 0 0 0 -1   Lb vb6  Lb  
1 dt ψCb6  C 
0  vvs pb6 
v 
  
ib7  0 0 1 -1 0
   
t
 b7   
1 ψ  pb7 
ib8 0 0 1 0 -1 0 vb8  Cb7 
1
 
0 0 1 0 0 vb9 pb8
 
ib9 -1 vrst
uvw
ψCb8
1 pb9
ib Λv vb ψCb9
1
(1) ψC pb
where the input and output ports are defined by “rst” and
where a SSM in a vector form is utilised, and pbj = vbj ibj is
“uvw” respectively, Lb is the arm inductance, and vcom is the
the instantaneous output power of the j-th cluster. Moreover,
voltage between the input and output port neutral points.
In addition, the linear transformation presented in [3], [1] as depicted in (6), the dynamic relationship between each ψC
is rearranged in the SSM formulation, and its respective output-power is linear, and the small-signal
i  2 models proposed in [2], [9], [15] are avoided if the controlled
α1 2 2 √-1 -1
√ √-1 -1
√ -1
√ -1 
√
i
 β1  0 0 0 3 3 3 - 3 - 3 - 3 variable is ψC instead of the sum of the capacitor voltages.
i  2 -1 -1 2 -1 -1 2 -1 -1 
 α2 
iβ2   √ √ √ √ √ √  If the expression (6) is pre multiplied by Tαβ0ε , the follow-
0 3 - 3 0 3 - 3 0 3 - 3
 i0  = 1  ing relation is obtained:
 
2 2 2 2 2 2 2 2 2 
 ib (2)
 
 iε1  6 2 -1
√ √-1 -1
√ -1
√ 2 √-1 2 -1 
√ ψ 
 
 
 iε2  0 - 3 3 - 3 3 0 3 0 - 3 Cα1 pα1 
ψCβ1 
2 -1 -1 -1 2 -1 -1 -1 2 
iε3
 
√ √ √ √ √ √ pβ1
iε4 0 - 3 3 3 0 - 3 - 3 3 0 ψCα2  pα2 
ψ 
p
 
iαβ0ε
Tαβ0ε d 
 C β2  2  β2 
ψC0  =  p0  (7)
dt  ψC  C  pε1 
  
Thereby, by applying (2) to (1), the following decoupled ψ  ε1
 pε2 

model is obtained:  Cε2  pε3
ψCε3 pε4
i  3 0 0 0 0 ψCε4
α1 0 v 
α1
0
iβ1  0 3 0 0 0 0  v o 
α1
vβ1  0 Regulation of the transformed variable ψC is convenient in
i  0 0 0 -3 0 0 v  0
 α2    v o   α2    terms of control purposes, which establishes the same inter-
iβ2  0 0 0 0 -3 0 β1 vβ2  0
  vo 
d  i0  = 1  0 0 3 0 0 1  v0  − 1

 
-3 01 
o −
   
3 vcom pretation shown in [9], [1]. The component ψC0 is related to
dt  iε1 
  Lb   vα2 L  vε1  Lb
   
0 0 0 0 0 0 
 v o 
 b 0 the overall energy stored in the M3C, and the remaining eight
0  β2
      
 iε2  0 0 0 0 0  vε2  0
 v02 o transformed terms represent the energy imbalance in the nine
iε3 0 0 0 0 0 0 vε3 0
      
iε4 0 0 0 0 0 0 o
vαβ0 vε4 0 clusters. If the eight variables related to the balancing are equal
iαβ0ε vαβ0ε to zero, it means that the stored energy of the M3C is evenly
(3) distributed among the clusters.
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 4

4
4

Fig. 2. Overall control scheme for the Modular Multilevel Matrix Converter
(M3C).

Regarding the regulation of the stored energy of the M3C,

the ideal condition is that all the capacitor voltages have the
same value vCref . Therefore, the set-point values to regulate the
ψc values are defined as:
Nt
ψCref0 = (vCref )2 (8)
3
ref
ψCαβε =08x1 , (9)
Fig. 3. Stored Energy Control of the M3C. (a) Total Energy Control and (b)
where Nt is the total number of modules, Balancing Energy Control.
ψCαβε =[ψCα1 ψCβ1 ψCα2 ψCβ2 ψCε1 ψCε2 ψCε3 ψCε4 ]T ,
and 08x1 is a 8x1 zero vector.
regulate the d-axis current reference of the rst port. Notice that
III. P ROPOSED C ONTROL S CHEME FOR THE M3C grid voltage orientation is assumed for this control system [see
As explained in the introduction, in this work the control Fig.3. (a)][2], [32].
system design is focused only on DFM operation [15], where 2) Balancing Energy Control: The goal for cluster-
mitigation of very large low-frequency oscillations of the balancing control is to balance the DC components of the
capacitor voltages is not required (see [5]). Discussion of the cluster stored energies, which is required during DFM Oper-
control system design and implementation is divided into three ation [15].
stages: control of the stored energy, control of the arm currents To define the balancing scheme, the dynamic equations of
and modulation integrated with capacitor balancing control. the transformed variables are established as:
Fig. 2 shows the proposed control scheme. dψCαβ1 2
˙ dt = (v c ic + v2 iε34 ) + pdαβ1
A. Stored Energy Control of the M3C 3C ˙ 2 ˙ ε12 ˙ ˙ ˙
dψCαβ2 2
Regulation of the stored energy in the M3C has two goals: ˙ dt = (v1 icε34 + v1c icε12 ) + pdαβ2
(a) to control the total stored energy and (b) to balance 3C ˙ ˙ ˙ ˙ ˙
dψCε12 2 c
the stored energy evenly amongst the nine clusters. Fig. 3 ˙dt = (v2 iε34 + v1 iε34 ) + pdε12
illustrates the implemented control schemes. 3C ˙ ˙ ˙ ˙ ˙
dψCε34 2
1) Total Energy Control: The aim of the Total Energy ˙dt = c c
(v2 i + v iε12 ) + pdε34 (11)
Control system is to regulate the variable ψC0 to the value 3C ˙ ˙ ε12 ˙ 1 ˙ ˙
depicted in (8). In terms of the transformed input/output where ψCαβ1 =ψCα1 +jψCβ1 , ψCαβ2 =ψCα2 +jψCβ2 ,
currents and voltages, the dynamic of ψC0 is given by: ˙ Cε2 , ψCε34 =ψCε3 +jψCε4
ψCε12 =ψCε1 +jψ ˙ are the complex
  ˙
form of the controlled ˙ variables. Additionally, v0 =0 is
dψC0 2  assumed, and iε12 =iε1 +jiε2 with iε34 =iε3 +jiε4 are
= 2< {v1 ic1 } + 2< {v2 ic2 } (10) the two circulating ˙ current pairs in their ˙ complex form.
dt 3C ˙ ˙ ˙ ˙
p1 p2 Furthermore, the following complex power terms are defined:
2 c c 2 c c 2
where p1 and p2 are the output and input port instanta- pdαβ1 = 3C v1 i1 , pdαβ2 = 3C v2 i2 , pdε12 = 3C (ic2 v1c + ic1 v2c ) and
neous power, respectively, and a complex notation is adopted pdε34 = 3C (i2 v1 + i1 v2 ), which depend on˙ the
2 ˙ c˙ c ˙ ˙ ˙ input/output
˙ ˙
with v1 = vα1 + jvβ1 , v2 = vα2 + jvβ2 , i1 = iα1 + jiβ1 and ˙ ˙ ˙ ˙
port variables and, therefore, they could be considered
i2 = ˙iα2 + jiβ2 as the ˙transformed output ˙ voltages and cur- as disturbances from a control perspective. Hence, the
˙rents of the M3C. Notice that the superscript c is used to manipulated variables for regulating the complex energy
indicate the application of the complex conjugate operator. vectors are the circulating currents iε12 and iε34 . In
Thus, for controlling ψC0 , the grid-side power can be used. this work, vector control systems orientated ˙ with ˙ rotating
In this regard, a simple PI controller implemented with an anti- reference frames are utilised for the regulation of the variables
windup saturation scheme with maximum value Isat d is used to [ψCαβ1 ,ψCαβ2 ,ψCαβ2 , ψCε12 ,ψCε34 ] and this can be achieved
˙ ˙ ˙ ˙ ˙
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 5

with relatively low bandwidth PI controllers, implemented the TMS320C6713 control board utilised to obtain the exper-
with anti-windup schemes and saturation values of ±Isat ε imental results discussed in this work. The integration of the
(see [9]). The reference frames for balancing ψCαβ1 , ψCαβ2 , bounds as inequality constraints in the MPC scheme enables a
ψCε12 and ψCε34 , are orientated along v2c , v1˙, v1c and ˙ v2 , saturation (protection) scheme for the power converter, where
˙ ˙ the
˙respectively;˙ additionally, for balanced operation, ˙ set-point
˙ the circulating currents are modified when a limit is reached.
values of all complex energy terms have to be set to zero. The target of the saturation scheme is to maintain the arm
Fig. 3(b) illustrates the Balancing Control Scheme, where currents and cluster output voltages within predefined limits
iε =[iε1 iε2 iε3 iε4 ] is the vector form of the four circulating even when the converter is operating under rapid transient
currents, and θ1 with θ2 are the angles of the output voltage conditions.
rotating vectors v1 and v2 respectively. In the approach presented here, the MPC scheme only
˙ ˙ considers the circulating currents. Whilst it may be possible
B. Arm Current Control
to include input/output port current regulation as well, a
Arm current control has also two goals: to regulate the in- scheme using only the circulating currents has a number
put/output port currents and to regulate the circulating currents. of attractive features which make it equally worthwhile to
The dynamics of the input/output current control loops are consider. These include the natural high bandwidth response,
dependent on the external systems connected to the converter, successful arm current saturation (with a straightforward im-
whereas the circulating current controllers are decoupled and plementation) during transients, and a reduced computational
are independent of whatever is connected to the input/output burden in comparison to an MPC scheme using the nine
ports. arm currents. Additionally, arm current/voltage saturation is
1) Input/Output Port Current Control: In this work, two optimally achieved without affecting the input/output port cur-
balanced grids are connected to the uvw and rst input/output rents, which can be adequately controlled using conventional
ports respectively. Without loosing generality, the output uvw approaches. Conversely, as discussed in the Introduction, and
port can be considered as equivalent to an electrical machine, demonstrated in Section V.A.3, saturation of the arm currents
where its back emf is the balanced grid. Therefore, the vector using conventional linear controllers may produce sub-optimal
uvw
vrst (see Fig. 1) can be decomposed as follows: performance.
-L
uvw 0 0 0 0 0  A. Modelling and MPC Definition
 0 -Luvw 0 0 0 0  As depicted in (3), the circulating current dynamic relations
uvw  0 0 -Luvw 0 0 0   d iuvw grid
 o
vrst =  dt rst +v12 (12) are not dependent on the input/output port voltages vαβ12 .
0 0 0 Lrst 0 0

0 0 0 0 Lrst 0
 By using a ZOH discretisation method with sampling time
0 0 0 0 0 Lrst Ts in the last four rows of (3), the following discrete-time
model for the circulating currents is obtained:
where Luvw and Lrst are the Thevenin equiva-
lent inductances of the input/output ports, and iε,k+1 = Aε iε,k + Bε vε,k (13)
grid
v12 =[vugrid vvgrid vwgrid vrgrid vsgrid vtgrid ]T are the grid voltages where Aε =I4 , Bε =- TLsb I4 and I4 is an 4x4 identity matrix.
of these ports. Notice that the dynamics of each circulating current are
For the implementation presented, regulation of the currents
dependent on its corresponding circulating voltage vε,k , and
is achieved using a vector control strategy with an anti-windup
there are neither external disturbances nor coupled interactions
saturation scheme implemented along a synchronous reference
with other circulating currents or voltages.
frame orientated along the corresponding grid voltage vector To regulate the circulating currents, a MPC strategy with
[33], [9]. finite horizon of order 1 is proposed [36], [37], [21], [38]:
2) Circulating Current Control: Regulation of the circulat-
ing currents requires knowledge of the output voltages defined Minimize J = x̂Tk+1 Qx̂k+1 + ûTk Rûk
x̂k+1 ,ûk
by the input/output port current controllers. The details of the
implemented MPC strategy is fully discussed in Section IV. subject to x̂k+1 = Aε x̂k + Bε ûk (14)
C. Modulation and Local Balancing Control -vdc,k ≤ vb,k ≤ vdc,k
The modulation scheme is based on a Phase Disposition -Imax ≤ ib,k+1 ≤ Imax
PWM with a sawtooth carrier waveform (single-edge PWM) with
[34], [35], where the Capacitor Voltage Balancing (CVB) of x̂k =iref
ε −iε,k (15)
the cells in a given cluster is accomplished by using a sorting
as the circulating current error vector, with reference iref
ε
scheme based on a priority list which defines the modules to
defined by the stored energy control (see the outer control
be utilised. This scheme is extensively discussed in [14].
loop in Fig.3(b)), and
IV. C IRCULATING C URRENT C ONTROL S CHEME
ûk =vεref -vε,k (16)
For regulating the circulating currents, an MPC strategy
is utilised, where arm currents and cluster output voltages as the circulating voltage error, with vεref is defined by a dead-
bounds are incorporated. The resulting optimisation problem beat control law [21], [38]:
is solved using an active-set algorithm [30], [31], which is Lb
suitable for implementation in a low cost DSP platform as vεref =- x̂k . (17)
Ts
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 6

ref ref
Matrices R = rε I4 and Q = qε I4 are, respectively, the with Lvdc =−vdc,k +vb,k and Uvdc =vdc,k +vb,k as inferior and
hv
costs of ûk and x̂k , with qε ≥0 and rε >0 [37]. For regulation
i
ref -1 αβ0,k
upper limits in vector form, and vb,k =Tαβ0ε v ref as the
purposes, the regulation of all current errors are considered ε
output voltage reference considering (17).
equally important.
The resultant inequality constraint depends on the value
Furthermore, the inequality constraints are specified by
of the measured currents and, therefore, it changes at every
the current limits and Cluster Capacitor Voltages (CCVs) of
sampling period.
the nine clusters, where it is important to highlight that the
measured CCVs are considered as constant values during the 2) Current limit constraints: On the other hand, the arm
sampling period. current limits can be stated by the transformed current terms:
By replacing the equality constraint of (14) in the cost
-Imax ≤ T-1 αβ0ε iαβ0ε,k+1 ≤ Imax (23)
function, the following optimisation problem is obtained [21],
[38]: h
i
i
where iαβ0ε,k+1 = αβ0,k+1
iε,k+1 . Moreover, taking into account
1
Minimize J = ûTk hε ûk − ûTunc,k hε ûk (18) (13) and (16) yields:
ûk 2
subject to -vdc,k ≤ vb,k ≤ vdc,k i  05x1

iαβ0,k+1 iαβ0,k+1
h i h
iε,k+1 = iε,k + - TLsb (vεref -ûk )
-Imax ≤ ib,k+1 ≤ Imax
iαβ0,k+1 Ts 05x1 Ts 05x4
h i h i h i
2
with hε =2(Ts /Lb ) qε + 2rε as the resulting cost of the = - ref + ûk (24)
iε,k Lb vε Lb I 4
new optimisation problem, and ûunc,k =-2(h-1 ε qε rε )x̂k as the
optimal solution of (18) without considering the inequality If the models of input/output ports are not clearly defined it
constraints. As described in (18), the inequality constraints could be stated that iαβ0,k+1 ≈iαβ0,k , which can be a good
sums a total of 36 bound limits. approximation when the sampling frequency is reasonably
B. Constraints of the MPC problem high, and the sampling time is relatively small when compared
In this section, the procedure of redefine the constraints to the time constants associated with the load dynamics. We
in terms of ûk , defined in (16), is detailed. To accomplish refer to this as the ”MPC Strategy with Saturation Scheme A”.
this goal, the dynamic relation (13), the input/output port If the models of input/output ports are known, the computation
voltages defined by the output current controllers, and the of an estimation for iαβ0,k+1 can be calculated online [for
linear transform defined in (2) are considered. instance using a discretised version of the input/output plant
1) CCVs constraints: The inequality constraints related to models described in Section III-B1] and it helps to obtain a
the available CCVs can be rewritten in terms of the trans- better definition of the bounds for each sampling period. We
formed voltages by using (2): refer to this as the ”MPC Strategy with Saturation Scheme B”.
Regardless of which saturation scheme is used, if an esti-
−vdc,k ≤ T-1
αβ0ε vαβ0ε,k ≤ vdc,k (19)
mation of the output currents is known at instant “k+1”, it can
where the vector vαβ0ε,k can be decomposed into be used in (23) to define the following bounds:
vαβ0,k
h i
vαβ0ε,k = v (20)  L
- Tsb (Imax +ib1,k )+∆vb1ε

ε,k
- Lb (I +i )+∆v 
Taking into consideration the definition shown in (16), here-  Ts max b2,k b2ε 
 Lb
inafter the following decomposition is used for the transformed - Ts (Imax +ib3,k )+∆vb3ε 

 Lb
voltage vector:

- T (Imax +ib4,k )+∆vb4ε 
hv  Ls 
h
vαβ0,k
i
αβ0,k
i - b (Imax +ib5,k )+∆vb5ε  ≤ Cu ûk ≤
=  Ts
vε,k ref 
v -ûk - Lb (Imax +ib6,k )+∆vb6ε 
hvε  Ts 
05x4
i h i - Lb (I +i )+∆v 
αβ0,k
= ref
vε − I4 ûk (21)  Ts max b7,k b7ε 
 Lb
- Ts (Imax +ib8,k )+∆vb8ε 

where the first term of the right hand of the relation is already
known from the input/output port current control scheme and - TLsb (Imax +ib9,k )+∆vb9ε
the circulating current control law used in the MPC scheme, LImax
and 05x4 is a 5x4 matrix composed of zeros. Replacing this L 
T (Imax −ib1,k )+∆vb1ε
b
decomposition into (19), yields:
 Lsb (I −i )+∆v 
 Ts max b2,k b2ε 
ref 2 √ 0 2 0
 ref  Lb
 Ts (Imax −ib3,k )+∆vb3ε 
    
vb1,k − vdc1,k vb1,k + vdc1,k

√
ref
vb2,k − vdc2,k  -1 - 3 -1 -√ 3 ref
vb2,k + vdc2,k   Lb
 T (Imax −ib4,k )+∆vb4ε 

 ref   √   ref   Ls
vb3,k − vdc3,k  -1 3 -1 3 vb3,k + vdc3,k  
√ √   b (Imax −ib5,k )+∆vb5ε 

 ref    ref 
 Ts (25)
 1 -1 -√ 3 3
vb4,k − vdc4,k   -1 vb4,k + vdc4,k  
 ref   ref   Lb (Imax −ib6,k )+∆vb6ε 
vb5,k − vdc5,k  ≤ -1 3 2 0 
√  ûk ≤ vb5,k + vdc5,k  (22)  Ts
 
 ref  3  Lb (I −i )+∆v 
v − vdc6,k  2 0 -1 -√3 v ref + vdc6,k   Ts max b7,k b7ε 
 b6,k   √  b6,k
 ref   Lb
v ref − vdc7,k  -1 3 -√ 3 v  Ts (Imax −ib8,k )+∆vb8ε 
-1  b7,k + vdc7,k 
 
 b7,k   
v ref − vdc8,k  2 0 -1 3 v ref
+ v Lb
Ts (Imax −ib9,k )+∆vb9ε
  dc8,k

b8,k
ref
√ b8,k
ref
vb9,k − vdc9,k -1 - 3 2 0 vb9,k + vdc9,k
UImax
Lvdc Cu Uvdc
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 7

where LImax =- TLsb (Imax +ib,k )+∆vbε is the lower bound, and the solution of the following Equality Constrained QP (EC-
UImax = TLsb (Imax −ib,k )+∆vbε is the upper bound in vector QP),
05x1 1 T
û hε ûk -ûTunc,k hε ûk
h i
form, with ∆vbε =T-1 . Minimize
αβ0ε v ref
ε
ûk 2 k
3) Incorporation of the Constraints into the Cost Function: subject to Ku∗ ûk =Mu∗ (32)
The constraints defined in (22) and (25), expressed in terms
(Ku∗ , Mu∗ ) ∈ W ∗ ,
of the argument ûk , are computed at every sampling period.
In the following, it is shown that the original 36 bounds can where W ∗ is the set of active constraints defined from (30),
be reduced to 18 to significantly reduce the complexity of the with Ku∗ and Mu∗ as the matrices composed of the active row-
problem to solve. vectors and elements from Ku and Mu , respectively. Because
For the j-th cluster, the following bounds are defined, the set W ∗ is not known initially, an iterative procedure
  for finding W ∗ and solving (32) is implemented. For every
LImax,j
≤ Cu,j ûk (26) iteration, a rearranged EC-QP obtained from (32) is solved,
Lvdc,j
which has the following structure for the j-th step,
where LImax,j and Lvdc,j are the j-th element of vectors LImax
and Lvdc , respectively. Additionally, Cu,j is the j-th row of 1 T T
Minimize p hε pk,j + gk,j pk,j
matrix Cu . pk,j 2 k,j (33)
As depicted in (26), if the constraint related to the maximum subject to Ku,j pk,j = 0
between LImax,j and Lvdc,j value is met, the other constraint is
where pk,j is the argument, gk,j =-hε (ûk,j-in +ûunc,k ) is a
automatically met too. Therefore, (26) can be replaced by just
vector defined at every iteration according to the input ûk,j-in ,
one constraint as,
and Ku,j is assembled from the selected row-vectors defined
Max {Lvdc ,j ; LImax ,j } ≤ Cu,j ûk (27) in W j according to the active constraints in the j-th iteration.
The solution of (33) is obtained using:
With a similar procedure, the upper bounds for the same  ∗  
T
 
j-th cluster can be stated as, hε I4 Ku,j −pk,j gk,j
= (34)
  Ku,j 0 λ∗k,j 0
−UImax,j
≤ (−Cu,j )ûk . (28)
−Uvdc,j where p∗k,j is the optimal solution, and λ∗k,j is the vector of
Lagrange multipliers that has a length equal to the number of
where UImax,j and Uvdc,j are the j-th element of vectors UImax
active constraints [31].
and Uvdc , respectively. With the same analysis done for the
If p∗k,j = 0, it means that ûk,j-in is the optimal value that
lower bounds, the two upper bounds can be replaced by the
minimizes (33). Moreover, if all the elements of λ∗k,j are zero
following constraint:
or positive, it implies that ûk,j-in is the global optimal solution
−Min {Lvdc ,j ; UImax ,j } ≤ (−Cu,j )ûk (29) that satisfies (18) , and W j =W ∗ .
If p∗k,j = 0 and at least one element of λ∗k,j is negative,
With the aforementioned considerations, the inequality con- it implies that the computed solution does not satisfy all the
straints of (18) rewritten in terms of ûk are incorporated as required conditions for (18) [31]; therefore, according to the
follows:     active-set methodology, the constraint ηj ∈ W j related to
Lu Cu
≤ û (30) the most negative element of λ∗k,j is removed, and a new
−Uu −Cu k
iteration starts with a reduced set of active constraints (i.e,
Mu Ku
W j+1 =W j −ηj and ûk,j = ûk,j-in ).
where On the other hand, if p∗k,j 6= 0 it implies that ûk,j-in is not
the optimal solution of (33) and it has to be updated according
Lu = Max {Lvdc ; LImax }
to
Uu = Min {Uvdc ; UImax } (31) ûk,j = ûk,j-in + αk,j p̂k,j , (35)
are the constraints utilised for the proposed MPC strategy. where
The operators “Max{}” and “Min{}” compute the maximum !
and minimum values for each element of the respective input mγ − kγT ûk,in
αk,j = min 1; min (36)
vectors. The outputs of these operators have the same size as (mγ ,kγ )∈V j kγT pk,j
the inputs. Therefore, the 36 constraints of the space defined
from (22) and (25), are significantly reduced to 18 as shown is the ”step-length” parameter, kγ is the γ-th row-vector of
in (31). These 18 constraints define the size of the matrices Ku , mγ is the γ-th element of Mu , and V j is defined as the
Mu and Ku as 18x1 and 18x4, respectively. set of all non-active constraints that, for every γ-th element,
satisfies kγT pk,j < 0.
C. Implementation of the Strategy If αk,j <1 due to some constraint ρj ∈ V j , this element
To solve the optimisation problem stated in (18), an Active- ρj is added to the active-set W j for the next iteration
set algorithm is used with the methodology discussed in [31]. (W j+1 =W j + ρj ).
The Active-set method states that the solution of the main A flowchart of the Active-set algorithm is illustrated in
Quadratic Programming (QP) problem of (18) is equivalent to Fig. 4. To avoid the potential problem of the algorithm failing
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 8

TABLE I
S IMULATION PARAMETERS BASED ON THE DIMENSIONING PROPOSED IN
[39], [40].
Rated power S 12kVA
Input/output port VL-L rms grid voltage vrst /vuvw 173V/173V
Capacitor Voltage vC 107V
Total number of cells Nt 27
Input/output port inductance Lrst /Luvw 1mH/1mH
Arm inductance Lb 1mH
Cell capacitance C 4.7mF
Maximum arm current Imax 40A

TABLE II
C ONTROLLER PARAMETERS U SED FOR THE S IMULATION T ESTS
Arm Currents Control
Output rst uvw Circulating
ξ 0.99 0.99 ξ 0.6
BW [Hz] 143 143 BW [Hz] 360
rε ; qε 5;1
Stored Energy Control
Total Energy Balancing
ξ 0.99 ξ 0.7
Fig. 4. Flowchart of the implemented Active-set algorithm described in BW [Hz] 33.2 BW [Hz] 14
Section IV-C.The algorithm is performed at every sampling period.

to converge, an iteration limit value jmax is set. If this value is possible value set by the speed control output saturation level.
reached, the iteration is stopped and the unconstrained solution Depending on the mechanical inertia connected to the machine
is used. shaft, the maximum torque reference could last the order of
In this work a value of jmax =9 has been used because, after seconds, which implies that all the internal circulating currents
extensive simulation and experimental work, the algorithm and the input/output currents have to be adequately regulated
never required more than 9 iterations to obtain the solution. to maintain the correct operation of the converter avoiding
Therefore, in accordance with (16), the applied circulating overcurrent in the arms.
voltage is determined by The test starts at 0 seconds (see Fig. 5), where the d-axis
current reference of the uvw port is stepped from 5A to 34A
vε,k = vεref − ûk (37) for 200ms; afterwards, it is changed from 34A to 22A. The
From this formulation, it is worth to mention that if null q-axis currents of the input/output ports are set to a constant
constraints are active, then value of -1A and 1A, respectively. The simulation results are
shown in Fig. 5, where the arm currents, CCVs, circulating
vε,k = kp x̂k . (38) currents and input/output port currents are illustrated. Two
scenarios are considered, the first one (left hand side of Fig. 5)
 
with kp =- TLsb +2h-1
ε qε rε as a constant value, which is ob-
shows the simulation results when the arm currents are not
tained by using expressions (17) and (18). Therefore, regula- saturated. At the right hand side the saturation scheme-B is
tion of the circulating currents is accomplished by a propor- enabled.
tional control scheme when no constraints are active.
As shown, to maintain the energy balance, in both scenarios
the sudden change of the uvw d-axis power current generates a
V. S IMULATION AND E XPERIMENTAL R ESULTS
fast increase of the d-axis power current current at the rst port,
A. Simulation Results which is driven by the Total Energy Control system. As shown
Using the Plexim PLECS software, a simulation model of a in the arm-currents, when the no-saturation scheme is utilised,
27-cell M3C with two balanced grids, connected at the input the arm currents have initial peaks near 52A (30% higher
and output ports, was implemented. The parameters of the than the 40A limit), and there are frequent and relatively high
M3C-grid system are depicted in Table I with the sampling current excursions outside the 40A during the 200ms transient.
period for the controllers set to 320µs. The parameters of the For the scheme-B, the arm current peaks are decreased in
controllers are shown in Table II. amplitude to short-duration over-currents at around 42A, 5%
1) Performance of the Proposed Control Strategy for Step higher than the limit.
Changes in the Output Load: In this section, the active power It should be highlighted that the 5% excursions outside
is stepped to a relatively high value on the load-side by a the predefined limit of 40A do not necessarily indicate a
step-change in the d-axis current reference of the output port. problem with the modelling of the system and/or a problem
Without losing generality, this test could emulate (for instance) with the algorithm. In this work a relatively low switching
a load impact on a motor drive. For a typical motor drive based frequency (for a single-edge PD-PWM) is being used to obtain
on a PI speed controller, sudden mechanical load impacts the experimental and simulation results. Therefore, the current
cause a speed reduction. In order to restore the speed, the ripple in the simulation results has a peak to peak value of
torque current component is usually increased to the maximum up to ≈ 4A which is produced by the switching. Hence, it
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 9

the non-saturated and saturated cases show good performance

in the regulation of the external port currents. Additionally,
in both scenarios, the CCVs exhibit relatively high magnitude
short-duration peaks at the very beginning, which are related
to the natural response due to the high load impact. Although
the circulating currents in the saturated case have higher values
than the non-saturated ones, the arm-currents have lower
peaks. This interesting feature is inherent to the proposed
control strategy to maintain the cluster currents inside the
predefined limits.
In Fig. 6, harmonic spectra of the input/output port currents
are shown when no saturation is required, and the uvw port
dq-axis currents are set to 22A and 1A, respectively. As
depicted in the results, high frequency bands are centred
around the carrier frequency as expected. Notice that for the
simulation and experimental results presented in this work, the
modulation scheme is based on a sawtooth carrier waveform
(single-edge PWM scheme). This modulation method has been
selected considering experimental simplicity.
2) Performance of the Strategy when CCVs constraints are
active: Although the proposed saturation scheme is intended
to be mainly used during transient instances, the performance
of the strategy is evaluated when the output cluster voltages
periodically reach their CCVs limits. This emulates a system
where the addition of the peak voltages at the rst and uvw
ports are comparable to the CCV values.
To enter this operating condition, the input/output port grid
voltages were increased an 18%. Fig. 7 illustrates simulation
results of this condition for the MPC saturation scheme-B with
the output uvw d-axis and q-axis currents set to 22A and to
Fig. 5. Simulation results of sudden load injection in the uvw port. (a) Without 1A, respectively; additionally, the input rst q-axis current was
saturation; (b) Saturation scheme-B.
set to -1A, and the d-axis current was specified according to
the Total Energy Control. As part of the test, at 0 seconds
a sudden change of the q-axis currents to 5A and -5A was
made for the uvw and rst port, respectively. As depicted in
the upper-left side graph, the output cluster voltages (vb ) have
several saturation periods. The number of active constraints
plot depicts each moment when one of the constrains is active.
By inspecting the graphic located at the left-top of Fig. 7
it is concluded that most of the times it is a CCV-related
constrain which is being active (see the saturated tops of some
Fig. 6. Simulation results of input and output port currents when iref
d =22A, of the waveforms). During all the tests it is shown that the
iref ref
q =1A in the uvw port, and iq =−1A in the rst port. (a) Harmonic spectrum
of input port current ir ; (b) Harmonic spectrum of output port current iu . input and output port currents are unaffected when the arm is
saturated, as depicted in their respective dq-axis current plots.
The circulating current could have some relatively high current
is concluded that most of the current excursions outside the peaks when a constraint is active, but the arm currents do not
predefined boundaries are produced by the switching ripple. surpass the maximum limit of 40A when these peaks occur.
Due to the sampling process, the peak ripple current is not 3) Comparison of MPC scheme-B with a Saturation Method
necessarily measured and the algorithm cannot take any action based on Linear Controllers.: To compare the arm current
to limit this value. To obtain a better regulation, the peak value saturation performance of the proposed strategy with previous
of the ripple could be considered when defining the value works, such as those reported in [2], [15], a comparison of
Imax [see (18)], for instance by setting the current limit to the MPC scheme-B with a current limitation scheme based on
36A instead of 40A. However, to compensate the effect of the adjusting the value Iεsat is presented [see Fig. 3.(b)]. For both
ripple produced by PWM-related issues is considered outside cases, the arm current saturation limit is set to Imax =40A,
the scope of this work. while the saturation values of the stored energy PI-controllers
For the simulation results presented in Fig. 5, using the are set to Iεsat =5A for the scheme based purely on linear
saturation scheme-B in the control algorithm does not affect controllers and to Iεsat =30A for the MPC based scheme.
the regulation of the input/output currents of the M3C. Both It is necessary to mention that defining Iεsat =5A for the
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 10

/ CCVs/ -CCVs linear approach is not an unfair comparison. In previous work

400 40A
50 considering linear controllers (see [9], [15]), each PI controller
200
related to the stored energy balancing control is limited to an
0
independent saturation value Iεsat , and there are no additional
-200 current saturation schemes implemented in the downstream
-50 -40A
-400 control stages. In the MPC case, each PI saturation limit is
- n° active constraints circulating currents
20 set to an apparently large value of Iεsat =30A; however, in
2 10 the circulating current scheme, the MPC strategy additionally
limits the magnitude of the circulating currents to fulfil the
1 -10 restriction of having a total current in the cluster below
-20 Imax =40A when it is required [see (18)]. Therefore, some of
the components of the circulating currents shown in Fig. 8.(b)
20 reach a higher peak than those depicted in Fig. 8.(a) between
20
10 0< t <40ms and this allows faster balancing of the capacitor
10
voltages. This faster response is produced because the MPC
0 0 algorithm manages to utilise more circulating current, without
-10 surpassing the maximum value allowed in the total arm
0 50 100 0 50 100
[ms] [ms] current. Conversely, as discussed in Section I, linear controllers
Fig. 7. Simulation results considering saturation of output cluster voltages.
saturate each component of the circulating current separately,
Description from upper-left to the right and down: (1) output cluster voltages using an heuristic criterion (see [15], [16]), rendering sub-
vb [coloured lines], CCVs and -CCVs [grey lines]; (2) arm currents; (3) optimal performance.
number of active constraints; (4) circulating currents; (5) dq-axis rst-port
currents; (6) dq-axis uvw-port currents.
In summary, in the MPC strategy, the algorithm optimally
decides the maximum circulating current component which
can be applied at a particular operating point, aiming to
50 40A 50 40A
achieve the lowest circulating current error with the active
constraints. This is a very important feature during transient
operation, where saturation of the arm currents is more likely,
such as the case demonstrated in section V-A1.
-50 -40A -50 -40A In the linear controller-based case, the circulating current
control is regulated by P-controllers, as shown in (38); ad-
cap. voltages cap. voltages ditionally, the parameters of the input/output port current
150 150 controller and the stored energy controllers are identical to
those used in the MPC strategy.
To compare both saturation methods, the following Tran-
100 100 sient Balancing Test (TBT) is performed: the energy stored
in the capacitors of the M3C is intentionally unbalanced
40 40
circulating currents circulating currents at t<0 by using non-zero references for the energy bal-
20 20
ancing controllers depicted in Fig. 3(b). At 0 seconds,
the references for the ψ values representing imbalances i.e
-20 -20 [ψCα1 ψCβ1 ψCα2 ψCβ2 ψCε1 ψCε2 ψCε3 ψCε4 ]T , are step
-40 -40 changed to zero to induce high circulating currents for balanc-
30 30 ing the converter energy. Furthermore, to induce an extreme
20 20 unfavourable scenario, the q-axis references are stepped from
10 10 1A to 10A at the uvw side, and from -1A to -10A at the
0 0 rst side at 10ms. The d-axis current of the uvw-side is fixed
-10 -10 during the test at 24A.
30 30 As illustrated in Fig. 8, at 0 seconds, the circulating current
20 20 injection increases for both schemes, with their respective arm
current values remaining within the margin of ±40A, with
10 10
minimal peaks above the limits. Although the peak values
0 0
of the circulating currents for the linear case are lower, the
-10 -10
arm current peaks have higher values than the MPC strategy.
0 40 80 120[ms] 0 40 80 120[ms]
(a) (b) Moreover, the arm currents in the MPC strategy remain within
Fig. 8. Simulation results for TBT for Imax =40A. (a) Scheme based on linear the predefined limits and the balance of the capacitor voltages
controllers with Iεsat =5A;(b) MPC scheme-B with Iεsat =30A. for the linear scheme takes nearly 20ms longer than the MPC-
based strategy. This feature is achieved due to the MPC
saturation scheme, which is required during specific periods
of the test and allows a higher value of Iεsat . The saturation
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 11

Fig. 10. Experimental set-up of a Modular Multilevel Matrix Converter

(M3C).

TABLE III
C ONTROLLER PARAMETERS OF THE E XPERIMENTAL S ET- UP
Arm Currents Control
Output rst uvw Circulating
ξ 0.99 0.99 ξ 0.99
BW [Hz] 253 165 BW [Hz] 268
Fig. 9. Simulation results of saturation scheme-A with TBT at different rε ; qε 10 ; 1
sampling periods. (a) Scheme-A at 320µs (b) Scheme-A at 106.67µs. Stored Energy Control
Total Energy Balancing
ξ 0.99 ξ 0.99
BW [Hz] 6.5 BW [Hz] 5.58

is active during short periods, especially when the circulating

currents reach the initial peaks. B. Experimental Results
To experimentally validate the proposed MPC strategy, an
4) Performance of the Saturation Scheme-A Considering experimental prototype of a M3C employing 27 H-bridge
Different Sampling Frequencies: As discussed before, to modules was implemented (see Fig. 10). The control plat-
implement the saturation scheme-B, estimation of the in- form has a Texas Instrument 6713C DSP board, three Actel
put/output port currents iαβ0,k+1 is needed. However, if the A3P1000 FPGA boards and additional external boards for ana-
input/output port models are not known, the approximation of logue to digital conversion and for computer communication
iαβ0,k+1 =iαβ0,k can be an attractive alternative. with the DSP. Each arm of the M3C is based on three H-bridge
modules with one Lb =2.5mH inductor, and each module
In this section, simulation results of a TBT test was made
has a 4.7mF/200V capacitor. At input port rst, an Ametek
for scheme-A at two different sampling periods:(a) 320µs
Programmable Power Source Model CSW5550 is connected
and (b) 106.67µs. For both cases, the system conditions and
in series with inductances of Lrst =5mH, whereas at output
controller bandwidth of the stored energy and the output
port uvw another Ametek Programmable Power Source Model
current controllers were the same. The d-axis current reference
MX45 is utilized with Luvw =2.5mH filters. These inductances
of the uvw-side was set to 24A, and the q-axis references are
(Luvw and Lrst ), represent the Thevenin impedances of the
stepped from 1A to 10A at the uvw side, and from -1A to -10A
power sources connected to the M3C ports [see Section III-B1]
at the rst side at 10ms. The CCVs reference level was 400V.
which have been selected based on laboratory availability.
The results are illustrated in Fig. 9, where the arm currents,
The DSP sample-time and PWM-carrier period are
number of active constraints and CCVs are included.
Ts =320µs, the DC-voltage reference for all modules is
As depicted in the arm currents for the two cases, the peaks 127V, the arm currents limit is Imax =12A, and the line-
where the current exceed the predefined limit are significantly voltage/frequency of the grid connected at rst-input port is
reduced from nearly 46A (15% above the limit) to 41A (2.5% 182V/50Hz. The parameters of the controllers of the M3C are
above), when the sampling period is reduced from Ts =320µs shown in Table III.
to Ts =106.67µs. As shown in the corresponding CCVs plot, 1) Performance of the MPC strategy Considering Different
the balancing controllers exhibit the same behaviour and Voltage and Frequency in the External Ports.: To show the
duration for balancing the clusters. From these results it can performance of the MPC scheme at different output frequen-
be established that scheme-A gives reasonable results when cies, the voltage and frequency of the programmable power
the sampling frequency is increased. source at the uvw port was programmed for a ramp variation
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 12

Fig. 14. Experimental results for the TBT without a saturation scheme.

The experimental results for this test are shown from Fig. 11
to Fig. 13.
The two upper plots of Fig. 11 illustrate the instantaneous
grid voltages at the uvw port, where the amplitude variation
Fig. 11. Experimental results of the proposed MPC strategy with cyclic
variations on voltage/frequency at the uvw-port power source. can be identified; the right upper plot shows the grid-frequency
of both the uvw and rst ports. The two middle graphics
show the peak grid voltages for both ports and the Cluster
Capacitor Voltages (CCVs). The lower graphics depict the
dq-axis currents of the uvw and rst ports, respectively. From
these experimental results it is concluded that the dq-axis
uvw output currents follow their respective references with
negligible tracking error, while the d-axis reference at the rst-
input port changes according to the active power demanded
by the total energy control system. Notice that in this test the
power at the port uvw is not constant because the uvw voltage
varies between 182V to 90V with constant d-axis current
reference. As shown in the middle right-side graphic, this
dynamic variation produces small low frequency oscillations
in the capacitor voltages.
Additional results corresponding to this experimental test
are shown in the scope waveforms illustrated in Fig. 12. The
steady state performance at 20Hz is shown in Fig. 12(a) and
Fig. 12. Experimental results of output and arm currents of the M3C at
at 45Hz in Fig. 12(b). Notice that the system is stable with
different voltage/frequency values at the uvw-grid. (a)Operation at 20Hz, and low distortion in the input and output current waveforms. The
(b) Operation at 45Hz. input/output currents ir , iu , and arm currents ib1 , ib4 are also
shown in Fig. 12. Additionally, harmonic spectra of the port
(a) (b) currents, when the output side frequency is 20Hz, are shown
[%] [%]
in Fig. 13. From the spectrum depicted in this figure it is
4 4 concluded that there are relatively high frequency components
which are centred at 3.125kHZ, which is the carrier frequency.
2 2 Additionally, just as observed in Fig. 6, low-order harmonic
content is present for both port currents. The magnitude of
0 0 the low frequency components could be improved by using
0 1 2 3 4 0 1 2 3 4
[kHz] [kHz] a triangular waveform as a carrier (instead of a saw-tooth
Fig. 13. Harmonic spectrum of input and output port currents. (a) Input port waveforms) and optimising some features in the experimental
current ir at 50Hz;(b) Output port current iu at 20Hz. system. However, the current in the grid side THD (input
port) is below 5% which is lower than the maximum value of
grid-side THD recommended in the EEE Std-519-2014 (see
from 182V/45Hz to 90V/20Hz and vice versa in ≈ 1.25s. For [41]) and it is considered appropriate for this application. The
the whole of this test the d-q axis current references uvw are output port side (load-side), working at 20Hz, has higher low-
set to 1A and 5A respectively. At the rst-input port, the q-axis frequency harmonic content in comparison with the input side.
reference is also constant at -5A whereas the d-axis reference This difference is produced because the output port inductance
is defined by the Total Energy Control system [see Fig.3.(a)]. is lower than the input port inductance.
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 13

Fig. 15. Experimental results for Saturation Scheme-A, and -B during TBT.

2) Comparison of Performance Achieved by the Proposed

Saturation Schemes.: The experimental test used in this sub-
section is designed to demand high magnitudes of circulating
currents that exceed a predetermined arm-current limit. This Fig. 16. Experimental results of output and arm currents of the M3C during
allows experimental verification of the saturation schemes TBT. (a) MPC strategy without saturation scheme, (b) MPC strategy with
discussed in Section IV. saturation scheme-A and (c) MPC strategy with saturation scheme-B.
In order to compare the performance of the saturation
scheme-A and -B, a TBT (described in section V.A.2) is per-
formed with the power source at the rst side set to 182V/50Hz, inside the predefined limits, with few undesirable spikes that
and the uvw power source set to 182V/25Hz. Balancing of the surpass the ± 12A band, reaching a maximum value of ≈ 14A.
CCVs starts at tON =0s, followed by a transient at tQ =20ms Notice that the circulating currents have some high-frequency
when the q-axis reference is stepped from 1A to 4A (445Var to oscillating components which are produced by a combination
1782Var) at the uvw side, and from -1A to -4A (-445Var to - of switching ripple and the fast dynamic response typically
1782Var) at the rst side. The d-axis uvw current reference is set achieved by the predictive control algorithms. However, these
at 6A (2670W). The results of this test for the strategy without oscillations in the circulating currents are not reflected in
saturation, and for Scheme-A and Scheme-B (considering either the uvw or in the rst output port currents because the
Imax = 12A) are shown in Figs. 14-17. proposed CCS-MPC maintains even the fast oscillations of the
As shown in Fig. 14, when no saturation scheme is used, circulating currents within the defined constraints.
the arm currents reach peak values near 20A and exceed As depicted in Fig. 15, scheme-B exhibits better arm
the defined ±12A limits several times. The input/output port current saturation performance than that achieved by scheme-
currents are unaffected during the balancing of the CCVs, A, maintaining the arm currents mostly inside the predefined
which lasts around 95ms. On the other hand, as shown in band. In comparison with the unconstrained results shown
Fig. 15, the arm peak currents are limited with saturation in Fig. 14, CCV imbalance is corrected more rapidly when
scheme-A, but there still are a couple of short-duration current the non-saturation scheme is applied, but this is achieved
peaks close to 17A. In scheme-B the arm currents are mostly by increasing the arm currents above the predefined limits.
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 14

saturation scheme-A and saturation scheme-B respectively. For

all the schemes it is concluded that the output currents are
not affected by the large changes in the circulating currents
(after tON ) confirming that the operation of the proposed M3C
control system achieves decoupled operation.
In Fig. 17 the four circulating currents are shown for
the three saturation schemes studied in this work. These
values have been obtained using the data acquisition system
embedded in the control platforms. Fig. 17 (a) depicts the
performance of the non-saturation scheme and Fig. 17 (b)
shows the performance of the scheme-A and scheme-B re-
spectively. Of course the tracking performance is better for
the non-saturated case because the control algorithm does not
consider the constrains to limit the circulating currents.
As shown in Fig. 17 (b), the effects of the proposed satura-
tion algorithms are clearly shown in the circulating current
waveforms. Although relatively high-frequency oscillations
are produced when the saturation scheme is activated, these
values are not necessarily reflected in the arm currents in
the natural coordinate domain. To reinforce this idea, notice
that the tracking of the circulating currents is worst when
the saturation scheme-B is used, which has better saturation
performance than scheme-A. This behaviour is inherent in
the MPC algorithm, because there is a trade-off between
good tracking of the circulating current references and proper
limitation when the maximum set values are reached by the
arm currents.

VI. C ONCLUSIONS
This paper has presented a continuous-control-set model
predictive control strategy for regulating the circulating cur-
rents of an M3C. The proposed methodology considers the
Fig. 17. Experimental results of circulating currents with their respective
references during Dynamic test. (a) MPC without Saturation, (b) Scheme-A maximum currents allowed in each cluster as constraints as
and Scheme-B. well as the maximum voltage which can be synthesised by
a given cluster without producing overmodulation. The most
important feature of the proposed strategy is that the MPC
The step changes in the active and reactive dq-currents at employed includes a saturation scheme that regulates the
the rst-input and uvw-output ports are also shown in Fig.15. thermal limit of the arm currents of the M3C and the saturation
According to the experimental results, the proposed saturation of the arm output voltages. These features are attractive during
scheme included in the control strategy does not affect the high demand of circulating currents in transient conditions.
dynamic or steady state performance of the input/output port The proposed saturation scheme is conveniently integrated
currents, confirming that decoupled operation is achieved. into the MPC formulation for the regulation of the circulating
As shown at the lower part of Fig.15, the number of currents. To solve the intrinsic optimisation problem formu-
activated constraints during the test for scheme-A and -B lated from the MPC, an Active-Set algorithm, suitable for
differ considerably. For scheme-A, there is mainly one ac- experimental implementation, is integrated into the proposed
tive constraint for most of the transient, but there are some control methodology.
instances during the test where two constraints are activated si- Two control schemes are derived (A and B) in this work.
multaneously. Analysing the experimental data it is concluded In scheme-A the model of the load/grid is not known and the
that this occurs when both the upper and the lower current prediction of the output currents is replaced by the sampled
limits are reached in some of the arms. On the other hand, measurement. On the other hand, in scheme-B the models of
for scheme-B, there are more instances when two constraints the load/grid are known and are integrated in the algorithm.
are active in comparison with scheme-A. Even more, up to The simulation and experimental results are promising and
three arm-current constraints are activated simultaneously at validate the usefulness of the proposed strategy and algorithm.
a specific instant, achieving a better limitation of the arms The attractive features of the scheme are shown during tran-
currents than that obtained by scheme-A. sient operation, where the saturation scheme performance is
In Fig. 16 the currents in the ports uvw and rst are shown, adequate for a sampling and carrier period of 320µs. Scheme-
as well as two arm currents ib1 and ib4 . Figs. 16(a), (b), (c) B has better regulation of the arm currents than scheme-
show the experimental results for the no saturation scheme, A during transient operation. However, it is shown that the
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 15

inherent prediction error of scheme-A is reduced when the [9] M. Diaz, R. Cardenas, M. Espinoza, F. Rojas, A. Mora, J. C. Clare, and
sampling frequency is increased. P. Wheeler, “Control of wind energy conversion systems based on the
modular multilevel matrix converter,” IEEE Transactions on Industrial
As discussed in this paper, saturation and thermal protection Electronics, vol. 64, no. 11, pp. 8799–8810, 2017.
are important tasks which are very difficult to address using [10] F. Kammerer, M. Gommeringer, J. Kolb, and M. Braun, “Benefits
conventional linear controllers considering that, in a typical of operating Doubly Fed Induction Generators by Modular Multilevel
Matrix Converters,” in PCIM Europe Conference Proceedings.
modular multilevel matrix converter, each cluster current has Nuremberg: VDE-Verl, 2013, pp. 1149–1156. [Online]. Available:
components of several frequencies and positive and negative https://publikationen.bibliothek.kit.edu/1000036276
sequences. Therefore, using linear methods it is difficult to [11] S. Liu, X. Wang, Y. Meng, P. Sun, H. Luo, and B. Wang, “A decoupled
control strategy of modular multilevel matrix converter for fractional
implement optimal limitation of these circulating currents and, frequency transmission system,” IEEE Transactions on Power Delivery,
at the same time, to avoid affecting the input/output currents. vol. 32, no. 4, pp. 2111–2121, 8 2017.
However, this task is simple to achieve using the CCS-MPC [12] H. Akagi, “Multilevel Converters: Fundamental Circuits and Systems,”
methodology that, unlike conventional FS-MPC strategies, is Proceedings of the IEEE, vol. 105, no. 11, pp. 2048–2065, 2017.
[13] M. Diaz, R. Cardenas Dobson, E. Ibaceta, A. Mora, M. Urrutia,
suitable for experimental implementation with present control M. Espinoza, F. Rojas, and P. Wheeler, “An overview of applications of
platforms. In addition, the MPC approach discussed in this the modular multilevel matrix converter,” Energies, vol. 13, no. 21, p.
work is utilised to regulate only the circulating currents 5546, Oct 2020.
[14] A. Mora, M. Urrutia, R. Cárdenas, A. Angulo, M. Espinoza, M. Dı́az,
providing saturation when required. This strategy has less and P. Lezana, “Model-predictive-control-based capacitor voltage bal-
computational processing time in comparison with the burden ancing strategies for modular multilevel converters,” IEEE Transactions
which would be expected if an overall control of the mod- on Industrial Electronics, vol. 66, no. 3, pp. 2432–2443, 2019.
ular multilevel matrix converter were implemented using an [15] M. Diaz, R. Cardenas, M. Espinoza, C. M. Hackl, F. Rojas, J. C. Clare,
and P. Wheeler, “Vector control of a modular multilevel matrix converter
MPC-based approach. Nevertheless, even though the proposed operating over the full output-frequency range,” IEEE Transactions on
control scheme may have less controllability, the advantages of Industrial Electronics, vol. 66, no. 7, pp. 5102–5114, 2019.
the proposed strategy, demonstrated using several experimental [16] M. Urrutia, F. Donoso, A. Mora, E. Espina, M. Diaz, and R. Cárdenas,
“Enhanced circulating-current control for the modular multilevel matrix
and simulation tests, have been clearly identified, i.e. when converter based on model predictive control,” in 2019 21st European
the arm voltages and/or currents reach the specified bounds, Conference on Power Electronics and Applications (EPE ’19 ECCE
they are modified in order to operate the arm-voltage/-current Europe), 2019, pp. 1–9.
[17] W. Kawamura, Y. Chiba, M. Hagiwara, and H. Akagi, “Experimental
magnitudes within the limits, without affecting either the input verification of an electrical drive fed by a modular multilevel tsbc
or output port currents. converter when the motor frequency gets closer or equal to the supply
To study and implement a more complex global MPC frequency,” IEEE Transactions on Industry Applications, vol. 53, no. 3,
pp. 2297–2306, 2017.
control scheme for modular multilevel converters will be the [18] A. Duran, E. Ibaceta, M. Diaz, F. Rojas, R. Cardenas, and H. Chavez,
subject of a future publication. “Control of a Modular Multilevel Matrix Converter for Unified Power
Flow Controller Applications,” Energies, vol. 13, no. 4, p. 953, feb 2020.
[19] R. Cardenas, M. Diaz, F. Rojas, J. Clare, and P. Wheeler, “Resonant
R EFERENCES control system for low-voltage ride-through in wind energy conversion
systems,” IET Power Electronics, vol. 9, no. 6, pp. 1297–1305, 2016.
[1] F. Kammerer, M. Gommeringer, J. Kolb, and M. Braun, “Energy [20] J. Rodriguez and P. Cortes, Predictive Control of Power Converters and
balancing of the modular multilevel matrix converter based on a new Electrical Drives, ser. Wiley - IEEE. Wiley, 2012.
transformed arm power analysis,” in 2014 16th European Conference
[21] T. Geyer, Model Predictive Control of High Power Converters and
on Power Electronics and Applications, 2014, pp. 1–10.
Industrial Drives. Wiley, 2016.
[2] W. Kawamura, M. Hagiwara, and H. Akagi, “Control and experiment
of a modular multilevel cascade converter based on triple-star bridge [22] S. Kouro, M. A. Perez, J. Rodriguez, A. M. Llor, and H. A. Young,
cells,” IEEE Transactions on Industry Applications, vol. 50, no. 5, pp. “Model predictive control: Mpc’s role in the evolution of power elec-
3536–3548, 2014. tronics,” IEEE Industrial Electronics Magazine, vol. 9, no. 4, pp. 8–21,
[3] F. Kammerer, J. Kolb, and M. Braun, “Fully decoupled current control 2015.
and energy balancing of the modular multilevel matrix converter,” [23] S. Vazquez, J. Rodriguez, M. Rivera, L. G. Franquelo, and M. No-
in 2012 15th International Power Electronics and Motion Control rambuena, “Model predictive control for power converters and drives:
Conference (EPE/PEMC), 2012, pp. LS2a.3–1–LS2a.3–8. Advances and trends,” IEEE Transactions on Industrial Electronics,
[4] M. A. Perez, S. Bernet, J. Rodriguez, S. Kouro, and R. Lizana, “Circuit vol. 64, no. 2, pp. 935–947, 2017.
topologies, modeling, control schemes, and applications of modular [24] A. Dekka, B. Wu, V. Yaramasu, R. L. Fuentes, and N. R. Zargari,
multilevel converters,” IEEE Transactions on Power Electronics, vol. 30, “Model predictive control of high-power modular multilevel converters
no. 1, pp. 4–17, 2015. - an overview,” IEEE Journal of Emerging and Selected Topics in Power
[5] M. Diaz, R. Cardenas, E. Ibaceta, A. Mora, M. Urrutia, M. Espinoza, Electronics, vol. 7, no. 1, pp. 168–183, 2019.
F. Rojas, and P. Wheeler, “An overview of modelling techniques and [25] A. Dekka, B. Wu, R. L. Fuentes, M. Perez, and N. R. Zargari, “Evolution
control strategies for modular multilevel matrix converters,” Energies, of topologies, modeling, control schemes, and applications of modular
vol. 13, no. 18, p. 4678, Sep 2020. multilevel converters,” IEEE Journal of Emerging and Selected Topics
[6] Y. Okazaki, W. Kawamura, M. Hagiwara, H. Akagi, T. Ishida, in Power Electronics, vol. 5, no. 4, pp. 1631–1656, 2017.
M. Tsukakoshi, and R. Nakamura, “Experimental comparisons between [26] A. Mora, M. Espinoza, M. Dı́az, and R. Cárdenas, “Model predictive
modular multilevel dscc inverters and tsbc converters for medium- control of modular multilevel matrix converter,” in 2015 IEEE 24th
voltage motor drives,” IEEE Transactions on Power Electronics, vol. 32, International Symposium on Industrial Electronics (ISIE), 2015, pp.
no. 3, pp. 1805–1817, 2017. 1074–1079.
[7] W. Kawamura, K. Chen, M. Hagiwara, and H. Akagi, “A low-speed, [27] B. Fan, K. Wang, P. Wheeler, C. Gu, and Y. Li, “An optimal full
high-torque motor drive using a modular multilevel cascade converter frequency control strategy for the modular multilevel matrix converter
based on triple-star bridge cells (mmcc-tsbc),” IEEE Transactions on based on predictive control,” IEEE Transactions on Power Electronics,
Industry Applications, vol. 51, no. 5, pp. 3965–3974, 2015. vol. 33, no. 8, pp. 6608–6621, 2018.
[8] D. E. Soto-Sanchez, R. Pena, R. Cardenas, J. Clare, and P. Wheeler, [28] M. Espinoza, R. Cardenas, M. Diaz, and J. C. Clare, “An enhanced dq-
“A cascade multilevel frequency changing converter for high-power based vector control system for modular multilevel converters feeding
applications,” IEEE Transactions on Industrial Electronics, vol. 60, variable-speed drives,” IEEE Transactions on Industrial Electronics,
no. 6, pp. 2118–2130, 2013. vol. 64, no. 4, pp. 2620–2630, 2017.
JOURNAL OF EMERGING AND SELECTED TOPICS IN POWER ELECTRONICS, 2021 16

[29] D. Karwatzki and A. Mertens, “Generalized control approach for a Roberto Cárdenas (S'95-M'97-SM'07) was born in
class of modular multilevel converter topologies,” IEEE Transactions Punta Arenas, Chile. He received the B.Sc. degree
on Power Electronics, vol. 33, no. 4, pp. 2888–2900, 2018. in electrical engineering from the University of
[30] A. Potschka, C. Kirches, H. Bock, and J. Schlöder, “Reliable solution Magallanes, Punta Arenas, in 1988, and the M.Sc.
of convex quadratic programs with parametric active set methods,” degree in electronic engineering and the Ph.D. de-
Technical Report 2010 11 2828, Heidelberg University, Interdisciplinary gree in electrical and electronic engineering from
Center for Scientific Computing, 2010. the University of Nottingham, Nottingham, U.K., in
[31] J. Nocedal and S. J. Wright, Numerical Optimization, 2nd ed. New 1992 and 1996, respectively. From 1989 to 1991 and
York, NY, USA: Springer, 2006. 1996 to 2008, he was a Lecturer with the University
[32] F. Blaabjerg, Control of Power Electronic Converters and Systems: of Magallanes. From 1991 to 1996, he was with
Volume 1. Elsevier Science, 2018. the Power Electronics Machines and Control Group,
[33] R. Teodorescu, M. Liserre, and P. Rodriguez, Grid Converters for University of Nottingham. From 2009 to 2011, he was with the Electrical
Photovoltaic and Wind Power Systems, ser. Wiley - IEEE. Wiley, 2011. Engineering Department, University of Santiago. He is currently a Professor
[34] B. Wu and M. Narimani, High-Power Converters and AC Drives, ser. of power electronics and drives in the Electrical Engineering Department,
IEEE Press Series on Power Engineering. Wiley, 2017. University of Chile, Santiago, Chile. Prof. Cardenas is an associate editor of
[35] D. Holmes and T. Lipo, Pulse Width Modulation for Power Converters: the IEEE Trans. On Industrial Electronics. In 2005 and 2019 he received the
Principles and Practice, ser. IEEE Press Series on Power Engineering. IEEE Transactions on Industrial Electronics Best Paper Awards.
Wiley, 2003.
[36] B. Kouvaritakis and M. Cannon, Model Predictive Control: Classical, Jon C. Clare (M'90-SM'04) was born in Bristol,
Robust and Stochastic, ser. Advanced Textbooks in Control and Signal UK, in 1957. He received the BSc and PhD. de-
Processing. Springer International Publishing, 2015. grees in electrical engineering from the University
[37] J. Rawlings, D. Mayne, and M. Diehl, Model Predictive Control: Theory, of Bristol, UK, in 1979 and 1990, respectively. From
Computation, and Design. Nob Hill Publishing, 2017. 1984 to 1990, he was a Research Assistant and
[38] T. Orłowska-Kowalska, F. Blaabjerg, and J. Rodrı́guez, Advanced and Lecturer with the University of Bristol, where he
Intelligent Control in Power Electronics and Drives, ser. Studies in was involved in teaching and research on power
Computational Intelligence. Springer International Publishing, 2014. electronic systems. Since 1990, he has been with
[39] F. Kammerer, J. Kolb, and M. Braun, “Optimization of the passive the Faculty of Engineering at the University of
components of the modular multilevel matrix converter for drive ap- Nottingham, UK. He is currently Professor of Power
plications,” in PCIM Europe Conference Proceedings. Mesago PCIM Electronics and is the Head of the Electrical and
GmbH [u.a.], 2012, pp. 702–709. Electronic Engineering Department. He is a member of the Power Electronics,
[40] W. Kawamura, M. Hagiwara, H. Akagi, M. Tsukakoshi, R. Nakamura, Machines and Control Research Group at Nottingham. His research interests
and S. Kodama, “Ac-inductors design for a modular multilevel tsbc are in power-electronic converters and their applications and control. Jon Clare
converter, and performance of a low-speed high-torque motor drive using is a Fellow of the IET and the recipient of a Royal Society Wolfson Research
the converter,” IEEE Transactions on Industry Applications, vol. 53, Merit Award.
no. 5, pp. 4718–4729, 2017.
[41] “Ieee recommended practice and requirements for harmonic control in Alan Watson (S'03–M'08-SM'21) received an
electric power systems,” IEEE Std 519-2014 (Revision of IEEE Std 519- M.Eng. (Hons.) degree in Electronic Engineering
1992), pp. 1–29, 2014. from the University of Nottingham, UK in 2004,
and a PhD, also from the University of Nottingham
in 2008. In 2009, he became a Research Fellow
with the Power Electronics Machines and Control
Group, University of Nottingham. Since 2009, he
has been involved in various projects in the area
Matı́as Urrutia (S'17) was born in Santiago, Chile, of high-power electronics including resonant con-
in 1991. He received the BSc and MSc degrees in verters, high voltage power supplies, and multilevel
electrical engineering from Federico Santa Marı́a converters for grid connected applications such as
Technical University (UTFSM) in 2017, where he HVDC and Flexible AC Transmission Systems. In 2012, he was promoted
is also a part-time lecturer. Currently, Mr Urrutia is to Senior Research Fellow before becoming an Assistant Professor in High
pursuing a dual-degree PhD in electrical engineering Power Electronics at the University of Nottingham in 2013. His current
offered by the University of Chile and the University research interests include the development and control of advanced high-power
of Nottingham. Since 2021, he has been part of conversion topologies for industrial applications, grid connected converters
the PEMC group of the University of Nottingham and HVDC Transmission.
as a Research Associate. His main interests are
Modular Multilevel Converters control, Model pre-
dictive Control and modern FPGA-based digital control schemes for power
electronics applications.

View publication stats