Modelling and Control of the Modular Multilevel Matrix

Converter and its application to Wind Energy Conversion

Systems
Matias Diaz Roberto Cárdenas Mauricio Espinoza Andres Mora Pat Wheeler
U. of Santiago of Chile U. of Chile U. of Costa Rica UTFSM The U. of Nottingham
Santiago, Chile Santiago, Chile. San José, Costa Rica Valparaíso, Chile Nottingham, UK

Abstract—In the last past years, some countries are enforcing

stringent grid codes to regulate the connection of Wind Energy
Conversion Systems (WECSs) to the electrical network, mainly
because of the high penetration of electric power from this
renewable source. Additionally, the trend of wind turbines has
shown an ongoing power rating growth, reaching sizes up to 10
MW. Multilevel converters have appeared as a solution for large
WECSs, due to its high reliability, controllability and the
capability to reach high power ratings. This paper presents a
control strategy for the application of the Modular Multilevel
Matrix Converter in Multimegawatts Wind Turbines. Extensive
computer simulations and a downscaled laboratory prototype,
with twenty-seven power cells, are presented to validate the
effectiveness of the proposed control system.
. Fig. 1: Proposed topology to drive a High-power Wind Turbines
Keywords— Wind Turbines, Wind Energy Conversion Systems, become the focus of modern wind energy application and
Fault Ride-Through Capability, Modular Multilevel Matrix research [2], and why wind turbines manufacturers have been
Converter.
upscaling turbine dimensions.
I. INTRODUCTION However, most of the present WECSs are based on low-
Wind Energy has become in the industry with greatest and voltage two-level voltage source power converters (usually
fastest growth among all the renewable energy sources. The based on 1700-V IGBT devices for a 690-V voltage level),
wind energy production capacity for the whole world increased which is not the best technology for high power applications due
exponentially from 17.4 GW in 2000 to 432.4 GW in 2015, to the high currents required. Consequently, medium or high
positioning wind power as a significant and crucial energy voltage power electronic converters (e.g. multilevel converters)
source in areas as China, USA and Europe. Powered 30,5 GW are well suited for high-power wind turbines. Therefore,
of new installations in China, the global wind power installed in Modular Multilevel Converters appears as a suitable technology
2015 was 63 GW, representing annual market growth of 22% to reach high power ratings.
[1]. The Modular Multilevel Matrix Converter (M3C) is a
A constant increment of wind power capacity is expected in the modular AC/AC converter able to reach medium-voltage levels
near future. Truthfully, the European Wind Energy Association by the series connection of full-bridge modules. This converter
(EWEA) plan for the next years is to become wind industry in has some advantages compared to traditional two-level
the most competitive energy source, by 2020 onshore and converters for high-power applications, fundamentally focused
offshore by 2030. The EWEA has stated that “wind power on the full modularity and easy extendibility to reach high
would be capable of contributing up to 20% of EU electricity by voltage and high power levels, redundancy, control flexibility
2020, 30% by 2030 and 50% by 2050”. and power quality improvements [3], [4].

An important part of the required future installed wind power In this context, this paper introduces the application of the
will be offshore based, because of the higher wind-energy M3C to drive high-power WECSs, as is illustrated in Fig. 1.
potential and the lower environmental impacts. Going offshore Comprehensive theoretical discussion on the control and
implies several technology challenges, mainly regarding modelling of the M3C is considered. Finally, the effectiveness of
reliability, efficiency and upscaling. Considering that hugely the proposed topology is confirmed by experiments using
expensive platforms must support the total weight of the WECS, simulation models and a downscaled 5kW laboratory prototype.
the size and weight of components are critical, and possible GRID INTEGRATION REQUIREMENTS FOR WECS
weight reductions become crucial. What is more, it has been
demonstrated that high power wind turbines could reduce the In countries with a high penetration of renewable energies,
cost structure of offshore WECS, reason why upscaling has dedicated grid code regulations have been enforced to ensure the
Chilean National Fund of Scientific and Technological
Development (FONDECYT) grant Nr. 1140337 and the
Advanced Centre for Electrical and Electronic Engineering,
AC3E, Basal Project FB0008

978-1-5090-3474-1/16/$31.00 ©2016 IEEE 5052

 b)

c)
a)

Fig. 3: Modular Multilevel Matrix Converter Topology. (a) Whole converter.

(b) M3C Branch composition. (c) M3C Cells.
Figure 2: FRT requirements. Voltage profile for simulation of faults in
Germany, Denmark, UK and Spain
the desired voltage in each cell, the point , is the number of
proper operation of the electrical networks. These grid codes cells, represents the capacitance of each capacitor and
have focused on power quality, power controllability, and Fault symbolises the total energy in the capacitors.
Ride Through (FRT) capability. Comprehensive reviews of
III. MODELLING OF THE M3C CIRCUIT
international grid code requirements for connection of wind
turbines to generation and transmission systems are presented in Modellings of the M3C have been recently reported in [6],
[5]. FRT requirements set the behaviour under Low-Voltage [7], where the fundamental approach is to use a so-called double
Ride-Through (LVRT) and High-Voltage Ride Through 0 Transformation. The approach exposed in this paper, is
(HVRT) grid-voltages, and represent, likely, the primary based on [7], but attempts to be an alternative more intuitive.
concern for wind turbine and power converter manufacturers. As In a first stage, the input system ( − − ) is transformed
depicted in Fig. 2, FTR from different national grid codes are into the 0 reference frame, as is depicted in Fig. 4(a).
slightly different in their requirements for the type, magnitude Afterward, the output systems is transformed into the 0
and duration of grid-voltage disturbances: reference frame. As can be observed in Fig. 4(b), the zero
II. M3C CIRCUIT CONFIGURATION sequence component of , , and correspond to the
equivalent circuit seen by the output. Analogously, the zero
Fig. 3(a) shows the circuit configuration of the M3C, which
sequence component of , , and can be obtained using
consists of three Subconverters (SC), each of them forms a
the equivalent circuit that the input sees from the converter. It
single phase converter. The nine branches links the phases of the
is important to note that the current between points and is
input system ( − − ), to the phases of the output system
zero because no path is considered.
( − − ). These branches are the series connection of Full
H-Bridge based cells and one inductor (Fig. 3(b)). The voltage Analysing Fig. 4(b), the following equations represent a
levels and switching frequency of a branch voltage depend on decoupled model for the input, converter and output of the
the modulation technique and the number of cells, leading to low M3C:
low harmonic distortion and small voltage drops when a high
number of cells is considered. In this topology, DC capacitor = + (2)
voltages are floating and, therefore, can charge-discharge
during the operation of the converter, which means that the = + (3)
average value of all the DC capacitor voltage has to be regulated
with low ripple and zero power mean value in steady state
=− (4)
operation [16].
To analyse the regulation of energy in the capacitors of the where:
3
M C, a single branch as the presented in Fig. 3(b) is used. =( + /3) (5)
Neglecting internal losses, the energy stored in a cells branch =( + /3) (6)
is defined as the integral of the power at its terminals.
Analogously, the energy of a branch is proportional to the The systems described in equations (2)-(4) represents the
capacitor voltages. Therefore, it is possible to determinate the dynamics of the M3C. Equations (2) and (3) represent the input
capacitor voltage using the branch power: and output systems, respectively. A precise definition of the
circulating currents among the converter is described in (4). By
= ≈ ̅ → ≈ ̅ ( ̅) (1)
this means that it is possible to regulate the circulating currents
where: ∈ − − , ∈ − − , represents the without perturbing the input and output currents.
branch power. ̅ is the cluster average capacitor voltage, ̅ is

5053
 a) converter ‘‘ ’’, that can be expressed as the difference of the
input and output converter power and approximated by (9):
= − = − ≈ ̅ (11)

where ‘‘ ’’ and ‘‘ ’’ are the -axis input voltage and current,

respectively. Moreover, ‘‘ ’’ could be considered as a
perturbation and can be neglected ---or feed-forwarded--- for
control purposes, which lead to the following relationship:
̅
= (12)
b) In consequence, the control system regulates the average
voltage of all capacitors by imposing the direct current
command ∗ :
∗ ( ) ∗ ( ∗00 − 3 0 )
= (13)
Where ( ) is the transfer function of the PI controller that
can be designed using the plant depicted in (12).
2) Balancing Capacitor Voltage Control
The eight remaining components of (10) should be
Fig.4: Electric Circuit Modelling of the M3C.
controlled to zero to mitigate the ripple in the capacitor voltages.
IV. PROPOSED CONTROL STRATEGY FOR THE M3C The unbalance voltage components ̅ , ̅ , ̅ , ̅
The proposed control strategy is based on the model presented represent the mismatches between capacitor voltages in the
in the previous section and provides decoupled control of the same SC (Inner-SC). On the other hand, components ̅ , ̅ ,
input (2), the output (3), and the M3C (4). An overview of the ̅ , ̅ represent the unbalance different SC (Inter-SC). In
proposed control system is presented in Fig. 5, which considers a both cases, has been probe that this components can be mitigated
hierarchical structure of the following sub-control systems: either regulating the circulating currents or the common mode
A. Control of the M3C voltage . In this paper, the circulating currents commands
have been calculated in order to mitigate the unbalance and
1) Average Capacitor Voltage Control
oscillating components produced by the variable speed at the
Using (1), the relationship power/capacitor voltage could be input side and to perform inter-subconverter balancing:
expressed as: ∗
= + + + − (14)
̅ = + ̅ (7) ∗
= − + + (15)
Using matrix notation, (7) yields: ∗
= + + − + − (16)
̅ ̅ ̅ 1 1 1 ∗
= − + − + (17)
̅ ̅ ̅ = ̅ + ̅ 1 1 1 Where the constants 1, and are:
̅ ̅ ̅ 1 1 1
√ √ √
(8) =− , =− , =− (18)
Applying the two-stage 0 transformation to (8): These constants are calculated using the double-stage 0
̅ ̅ ̅ transformation to the power terms of (9).
̅ ̅ ̅ = ̅ + 3) Circulating Current Control
̅ ̅ ̅ ̅ Because of the topology of the converter, the branch
(9) currents, and consequently circulating currents, contain different
When all the elements left side of (8) are equal to ̅ , frequency components (rotating at and , which represent
capacitor voltages in the two-stage 0 frame are: the input and output frequencies, respectively). Considering that
̅ ̅ ̅ the plant is integrative, proportional controllers could be well
0 0 0 suited for this application. According to (4), the voltage
̅ ̅ ̅ = 0 0 0 (10) commands to achieve decoupled current control of the four
̅ ̅ ̅ 0 0 3 ̅ circulating currents should be:
∗ ∗ ∗ ∗
Then, when all capacitor voltages are equal to the desired
set-point, just the component ‘‘ ̅ ’’ has to be regulated. In fact, ∗ ∗ =− ( ∗ ∗ − ) (19)
this component is related to the active power flowing into the
where represent the proportional gain transfer function.

5054
 Fig. 5: Overview of the proposed control strategy
∗
B. Input Current Control: ∗

In this proposal, the αβ reference frame is used to regulate ∗

− − ∗
the input system. Consequently, the current reference ‘‘ ∗ ’’ is = (22)
∗
− − 0
transformed into the αβ reference frame the using the input ∗ 0
angle . Besides, using superposition this control scheme can
be easily linked with maximum power point tracking control to Because of Resonant Controllers (RC) can be used to
extract the maximum wind power. The use of a resonant regulate positive and negative sequence current at the same time,
controller is suitable for this application [14]. Considering (2), a only two RC are needed instead of four for d-q axis based control
simple Resonant Regulator could be utilised to control the input systems [10], [11]. Therefore, this type of regulators is applied
currents as follow: to output current control. At this point and considering (3),
∗ ∗ voltage references to achieve decoupled output current control
∗ = − + ( )( ∗ − ) (20) using RC can be calculated as:
∗ ∗
Where ( ) represents the input current Resonant ∗ +
∗ = − − ( )( ∗ ∗
− ) (23)
Controller transfer function. +
C. Output current control
Where ( ) represents the Output Current Resonant
For the implementation of FRT control systems, usually, the
Controller transfer function.
measured currents and voltages have to be separated into
positive and negative sequence components. To achieve D. Cell balancing Method and Modulation
sequence separation, Delayed-signal-cancellation (DSC) is The voltage references obtained in control loops presented
probably the best-suited method [8], but has an intrinsic delay of abode ( ∗ , ∗ , ∗ , ∗ , ∗ , ∗ , ∗ , ∗ ) are transformed to the
5ms (for ωg=50Hz) before achieving the separation of the natural reference frame using the two stage inverse 0-
sequence components. To reduce the inherent delay present in Transformation.Then, an AC voltage reference is obtained for
traditional DSC, the fast convergence DSC proposed in [9] is each branch: ∗ , ∗ , ∗ , ∗ , ∗ , ∗ , ∗ , ∗ , ∗ .
utilised in this work to calculate the sequence components.
An additional control loop is utilised to ensure equal dc
As is reported in [10], the currents, the powers and the capacitor voltages within a branch [12]. The voltage reference
voltages are related by: “ ∗ ” is distributed among the number of cells in the −
branch. The capacitor voltage for the cell“i”, with ∈ (1, ), is
− compared with the desired value “ ̅ ”. Then, the error is
− multiplied for the sing of the branch current, resulting in the
= (21)
− − addition/subtraction of a ∆ to the cell voltage reference.
Finally, the phase-shifted PWM is used to synthesise the
where the superscripts , and are utilised to symbolise the multilevel waveforms. It has been validated that the phase-
positive and negative sequence components, respectively. The shifted PWM is easy to enlarge the count of bridge cells per
terms “ ” and “ ” represents double frequency oscillations cluster, the power losses are equally distributed in the cell of the
in active power that can be mitigated considering the next same branch. Moreover, using phase-shifted unipolar
current reference calculation: modulation, the output switching frequency is 2*3n times the
triangular carrier frequency.

5055
a) b) c)

d) e) f)

Fig 6: Simulations Results (a) Output Voltages and Currents. (b) Amplified view of (a). (c) Input Voltages and Currents. (d) Active and Reactive Power injected into
the grid. (e) Average Branch Capacitor Voltages for SubConver A, B and C, respectively. (f) Capacitor Voltages, branch “a-s”.

V. SIMULATION AND EXPERIMENTAL RESULTS reducing the current to zero (Fig. 6(c)). The output currents are
controlled using the calculation presented in (22), which
The effectiveness of the proposed control strategy has been generates unbalanced references to mitigating the effects of the
tested through simulation and experimental work. The grid voltage dip and provide full reactive power injection, as is
simulations have been developed using PLECS software, shown in Fig.6(d). There are not double-frequency oscillations in
considering a five cell per branch M3C connected to the grid. The active power, but reactive power presents double frequency
general parameters of the simulated model are depicted in Table oscillations that cannot be controlled because there are not enough
I. This system has been tested under variable speed operation and degrees of freedom [11]. As presented in Fig. 6(e) and Fig. 6(f),
grid voltage dips to validate the novel application of M3C driving proper energy balancing is achieved through the grid voltage dip.
high-power grid-connected WECS.
Experimental results have been obtained with downscaled
Table I: Simulation Parameters Table II: Experimental Setup laboratory prototype, composed of nine branches, each of them
Nominal Power 10 MW Nominal Power 5 kW
considering the series connection of 3 H-Bridge modules and 1
Cells per branch 5 Cells per branch 3
cluster inductor. The system is controlled using one Digital Signal
Input Voltage/Freq, 5.4kV/10-40Hz Input Voltage/Freq, 200V/10-40Hz Processor (DSP) Texas Instrument TMS320C6713 and three
Branch Inductor 2.5 mH Branch Inductor 2.5 mH Actel ProAsic3 field programmable gate array (FPGA) boards
Cap. in each cell 2.8 mF Cap. in each cell 4.7 mF equipped with 50 14-bit analogue-digital channels. The 108
Capacitor Voltage 2.4 kV Capacitor Voltage 155V switching signals are obtained in the FPGA boards using unipolar
Output Voltage/Freq. 5.4kV/60 Hz Output Voltage/Freq. 185V/50 Hz phase-shifted PWM and are transmitted using fibre optic
Switching frequency 0.8kHz Switching frequency 2.5kHz communications. More details about the experimental prototype
are depicted in Table II, and a picture of the system is shown in
The performance of the proposed control strategy is tested for Fig. 7(a).
a Dip Type C (two phases decrease their nominal voltage to 0V) The synchronised operation of the experimental prototype is
based on the regulations present in the German grid code [5], as presented in Fig. 7(b)-(g). The input and output currents are
is shown in Fig. 6(a) and Fig 6(b). When the fault appears, the controlled to 13.5A (peak value) and are not affected by the
input current control stops the active power generation by

5056
 b) 0.1s c) 0.1s
a)

Currents (A)
10
0
-10
d)

Inner-SC Unbal. Inter-SC Unbal.

Cap. Volt. (V) Cap. Volt. (V)
8

-8
e)
8

-8
f)

Voltages (V)
155

Capacitor
150
145

g)
4

(kW - kVAr)
Grid Power
P
2 g
Q
g
0
0 0.2 0.4 0.6 0.8 1
Time (s)
Fig. 7: (a) Experimental Prototype. (b)-(g) Experimental Results. (b) Grid Currents. (c) Input Currents. (d) Inter-SC Unb. Voltages (e) Inner-SC Unb. Voltages.
(f) 27-Capactitor Voltages. (g) Active and React. Power Injected into the grid.

balancing algorithm. As shown in Fig. 7(b) and Fig. 7(c), the [4] J. Korn, M. Winkelnkemper, P. Steimer, and J. W. Kolar, “Direct
input and output currents present low harmonic distortion (≈ modular multi-level converter for gearless low-speed drives,” in
Proceedings of the 2011 14th European Conference on Power
2%) and are balanced. The Inter-SC ripple components ( , Electronics and Applications, 2011, no. direct MMC, pp. 1–7.
, , ) are presented in Fig. 7(d), whereas the [5] F. Iov, A. D. Hansen, P. Sørensen, and N. A. Cutululis, “Mapping
Inner-SC ripple components ( , , , ) are of grid faults and grid codes,” Wind Energy, vol. 1617, no. July, pp.
shown in Fig. 7(e). The eight unbalance voltage components are 1–41, 2007.
[6] F. Kammerer, J. Kolb, and M. Braun, “Fully decoupled current
successfully regulated under a +-5V band. The average value of control and energy balancing of the Modular Multilevel Matrix
all capacitor voltages is regulated to 150V (see Fig. 7(f)). Converter,” in 15th International Power Electronics and Motion
Finally, the unitary power factor operation of the system, Control Conference and Exposition, EPE-PEMC 2012 ECCE
injecting 4kW into the grid, is presented in Fig. 7(g). Europe, 2012, p. LS2a.3-1-LS2a.3-8.
[7] W. Kawamura, M. Hagiwara, and H. Akagi, “Control and
VI. CONCLUSIONS Experiment of a Modular Multilevel Cascade Converter Based on
Triple-Star Bridge Cells,” IEEE Trans. Ind. Appl., vol. 50, no. 5, pp.
A fully decoupled control strategy for the application of the 3536–3548, Sep. 2014.
Modular Multilevel Matrix Converter in high-power WECS has [8] S. Alepuz, S. Busquets-Monge, J. Bordonau, J. A. Martinez-
been proposed in this paper. Velasco, C. A. Silva, J. Pontt, and J. Rodriguez, “Control strategies
based on symmetrical components for grid-connected converters
The proposed control strategy enables independent current under voltage dips,” IEEE Trans. Ind. Electron., vol. 56, no. 6, pp.
and energy balancing regulation, by using the electrical circuit 2162–2173, Jun. 2009.
transformation based on the two-stage 0 transformation. [9] R. Cardenas, M. Diaz, F. Rojas, and J. Clare, “Fast Convergence
Delayed Signal Cancellation Method for Sequence Component
Experimental results obtained with a 27 H-Bridges Prototype Separation,” IEEE Trans. Power Deliv., vol. 30, no. 4, pp. 2055–
and Simulation tests have been carried out to validate the 2057, Aug. 2015.
[10] M. Díaz, R. Cárdenas, P. Wheeler, J. Clare, and F. Rojas, “Resonant
effectiveness of the proposed control strategy for wind energy control system for low-voltage ride-through in wind energy
applications. In fact, the proposed control approach could be conversion systems,” IET Power Electron., vol. 9, no. 6, pp. 1297–
suitable for a wider range of electrical machine applications. 1305, May 2016.
[11] M. Diaz and R. Cardenas, “Analysis of synchronous and stationary
VII. REFERENCES reference frame control strategies to fulfill LVRT requirements in
[1] Global Wind Energy Council, “Global Wind Statistics 2015,” 2015. Wind Energy Conversion Systems,” in 2014 Ninth International
[2] N. Fichaux, J. Beurskens, and P. Jensen, “Upwind: Design limits and Conference on Ecological Vehicles and Renewable Energies
solutions for very large wind turbines,” 2011. (EVER), 2014, pp. 1–8.
[3] A. Mora, M. Espinoza, M. Diaz, and R. Cardenas, “Model Predictive [12] H. Akagi, S. Inoue, and T. Yoshii, “Control and Performance of a
Control of Modular Multilevel Matrix Converter,” in 2015 IEEE Transformerless Cascade PWM STATCOM With Star
24th International Symposium on Industrial Electronics (ISIE), Configuration,” IEEE Trans. Ind. Appl., vol. 43, no. 4, pp. 1041–
2015, pp. 1074–1079. 1049, 2007.

5057

Powered by TCPDF (www.tcpdf.org)