You are on page 1of 10

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO.

1, JANUARY 2007 139

Comparison of Simple Control Strategies of


Space-Vector Modulated Indirect Matrix
Converter Under Distorted Supply Voltage
Matti Jussila, Student Member, IEEE, and Heikki Tuusa

AbstractThis paper concerns the effect of symmetric supply


voltage harmonics in a space-vector modulated three-phase in-
direct matrix converter (IMC). The IMC is modulated using
indirect space-vector modulation. The effect of symmetric supply
harmonics on load voltages and currents is analyzed using
space-vector presentation. The measurement and simulation
results with distorted supply voltages are compared to the results
with pure sinusoidal supply voltage. Four simple control methods
of IMC are tested in simulations and measurements: 1) open-loop
control without measurements (Method I); 2) calculation of
modulation index using measured supply voltages (Method II);
3) closed-loop control of output currents (Method III); and 4)
combination of Methods II and III, i.e., closed-loop output current
Fig. 1. Indirect matrix converter.
control with supply voltage measurement (Method IV). All the
methods are easy to implement and do not require complicated
computing. It is noted that Methods II and IV, where modulation
index is calculated using measured supply voltage magnitudes, MCs with abnormal supply conditions have been studied,
can produce more sinusoidal load currents than Methods I and e.g., in [8][15], of which [8][10], [12], and [15] focus on
III, which suppose ideal supply conditions. unbalanced and sinusoidal situations. However, for example,
Index TermsIndirect matrix converter (IMC). in industry low-voltage mains can be distorted by three-phase,
full-bridge diode rectifiers with capacitive load. In that case,
all three phases are equally distorted and voltages remain sym-
I. INTRODUCTION metric, i.e., they are not unbalanced. These kinds of supply volt-
ages are quite likely in the future with MCs commercial ap-
ATRIX converters have been studied widely since their plications and thus their effects are studied in this paper. When
M principle was introduced in 1970 [1]. In recent years, re-
search on direct frequency conversion using matrix converters
there is harmonic distortion in supply voltages a choice between
input and output current quality has to be made [14].
(MCs) has become popular [2]. The main reason for the interest In many papers, covering abnormal supply conditions the
in MCs is that they provides a compact solution for a four-quad- main concern has been the quality of the supply current [9][12],
rant frequency converter, which produces sinusoidal input and and [15]. In this paper, the load current quality is a primary
output currents without passive components in dc link. The ab- value and the supply current quality is a secondary value. Thus,
sence of dc link also has its disadvantages: unfiltered input and the quality of output power is considered more important than
output disturbances are carried through MC and additional com- the quality of the current drawn from a supply whose voltages
mutation strategies are needed to avoid a short-circuiting of the are already distorted. Naturally, the standards and environment
supply or a cutting of the load current path. The outputinput set different limits in individual cases. Thus, it is important to
voltage ratio is also limited to 0.866 in linear modulation. have knowledge about the quality of output power produced
A conventional space-vector modulated direct matrix con- by simple open- and closed-loop control methods of the IMC,
verter (DMC) may be replaced by an indirect matrix converter which can be implemented easily and cost-effectively.
(IMC), which consists of separated line and load bridges as pre- Four simple control methods of the IMC are tested. The first
sented in Fig. 1 [3][7]. The IMC offers the same benefits and control method (Method I) is an open-loop control method,
disadvantages as the DMC, but it also provides an option to re- which contains only the space-vector modulator without the
duce the number of active switches of the line bridge to three if supply voltage or load current magnitude measurements [16].
no bidirectional power flow is needed [5], [6]. Anyhow, Method I, like every other method considered here,
requires a zero crossing detection of one supply phase voltage
Manuscript received August 24, 2005; revised January 20, 2006. This paper for synchronization of the input current reference. This does
was presented in part at PESC05, Recife, Brazil, June 1216, 2005. This work not require A/D converters but can be implemented with sim-
was supported by the Fortum Foundation (Fortumin sti). Recomended for pler electronics. The second control method (Method II) is
publication by Associate Editor J. Rodriguez.
The authors are with the Institute of Power Electronics, Tampere University
based on the measurements of the supply voltage magnitudes
of Technology,Tampere FI-33101, Finland (e-mail: matti.jussila@tut.fi). [8][15] and likewise Method I does not contain closed-loop
Digital Object Identifier 10.1109/TPEL.2006.886654 load current control either. The third control method (Method
0885-8993/$20.00 2006 IEEE
140 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO. 1, JANUARY 2007

III) is based on the load current control and only needs load
current measurements in addition to the measurements required
in Method I. The fourth control method (Method IV) studied is
a combination of Methods II and III. Hence, it needs both the
supply voltage magnitude and the load current measurements.
Methods I and II are suitable for drives where current sensors
are not otherwise used. Methods I and III are suitable for
applications where the magnitude measurements of supply
voltages are not applied.
All the methods are easy to implement and do not require
complicated computing. In addition, the simulation and mea- Fig. 2. Input phase current space-vectors of the IMC line bridge: (a) i >0
surement results will show that the methods produce line cur- and (b) i < 0. (c) Synthesis of the input current reference vector.
rents with remarkably lower distortion than full-bridge diode
rectifiers. The results will also show that Methods I and III
cannot compensate the effects of the supply voltage harmonics
as effectively as Methods II and IV because in those methods
ideal supply voltages are assumed.

II. EFFECT OF SUPPLY VOLTAGE DISTORTION

A. Space-Vector Modulation
The indirect space-vector modulation used for MC has been
presented in [4] and is directly suitable for IMC as presented
in [7]. It is based on the space-vectors of input voltages and
currents and output voltages, , , , respectively Fig. 3. (a) Output phase voltage space-vectors of the IMC load bridge. (b) Syn-
thesis of the output voltage reference vector.
(1)

a b c When the input voltages are purely symmetric and sinusoidal


and is constant. The displacement angle be-
tween the input voltage and current vectors is , which can
be assumed constant during one modulation period. Moreover,
(2)
when the input fundamental angular frequency is one can
write an equation for average dc link voltage during one modu-
lation period

(3)

where . (5)
In (2) , , and are the switching functions of the
IMC line bridge and is their space-vector. The allowed The line bridge is always modulated to produce maximum dc
switching states producing the input current space-vectors are link voltage so that [4], [7]
presented in Fig. 2. Reference vector is synthesized as a
time average of adjacent vectors and , whose duty cycles (6)
are and , respectively, as presented in Fig. 2(c).
In (3) , , and are the switching functions of Thus, the average link voltage on one modulation period is
the IMC load bridge and is their space-vector. The allowed
switching states producing output voltage space-vectors are pre- (7)
sented in Fig. 3(a). Reference vector is synthesized as a
time average of adjacent vectors and , whose duty cycles The calculation of the duty cycles is based on the modulation
are and , respectively, as presented in Fig. 3(b). index , which is equal to the modulation index of conventional
When the line bridge is assumed to be ideal the instant active voltage source inverter:
powers of line and dc link are equal. By substituting (2) into the
power equation it is possible to define the instant dc link voltage
(8)
[7]
(4) where in linear modulation [4], [7]. Substituting (7)
JUSSILA AND TUUSA: COMPARISON OF SIMPLE CONTROL STRATEGIES 141

into (8) we get

(9)

When the separate duty cycles of the line and load bridges, ,
and , , respectively, are combined the result is four ac-
tive duty cycles [4], [7], so that the output voltages desired are
produced directly from the three-phase input voltages

(10a)

(10b)

(10c)
(10d) Fig. 4. Block diagram of the IMC control methods.

For zero vector: .


bottom of the page, where is the fundamental angular fre-
B. Reflection of Supply Voltage Harmonic Distortion quency of the output voltage desired.
The harmonic distortion in balanced three phase supply volt- As can be seen in (13) the additional components in the
ages can be presented in space-vector form as output voltage caused by input voltage distortion are in fre-
where the order of the harmonic component 6 1, quencies 6 , where is the fundamental input frequency
and 1. When the phase voltages are symmetrical and is and is the fundamental output frequency. The magnitudes of
the fundamental angular frequency of the input, the space-vec- those voltage components do not depend only on the magnitude
tors of the harmonic voltage components are of each input harmonic but with suitable values of
voltage components and can compen-
when and sate each other.
(11a)
III. CONTROL METHODS OF THE IMC
when (11b)
A. Method I
In (11), is the phase-difference angle between the
nearest zero points of 6 1 th harmonic and the fundamental The first control method of the IMC (Method I) contains only
component. With the help of (5) one gets for the average dc the vector modulator described in Section II-A and a possibility
link voltage to directly set voltage reference values and , which are
then used to produce two-axis components of voltage reference
and presented in Fig. 4. The modulation index
is calculated assuming both sinusoidal and balanced input volt-
ages, i.e., using fixed value for space-vector length
[16]. Thus, when is set at unity, which provides the
maximum voltage transfer ratio, the produced output voltages
(12) include all the harmonics presented in (13)
Thus, the averaged is not constant but it contains harmonic
ripple components in addition to the constant value presented (14)
in (7). Thus, the result is quite the same as with the unbalanced
supply voltages presented in [8], when the dc link voltage con- The input current is modulated along the switching vector
tains only the frequency component of 2 in addition to dc , whose direction is . Thus the angle is also the
value. angle of input current reference space-vector and used to de-
When the modulation frequency components of are ig- fine the angle (Fig. 2) needed in (10). It is defined with the
nored, the substitution of (12) into (3) gives (13), shown at the help of the zero crossing detection of filtered supply voltage

(13)
142 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO. 1, JANUARY 2007

resulting fundamental and the use of a software imple-


mented phase-locked loop (PLL). This is the simplest system
to synchronize input current reference vector to the supply volt-
ages [9]. With 1 the input current reference vector
has the same direction as the fundamental supply voltage
vector. The relation between the zero crossing of based fun-
damental phase angle and the direction of the switching
vector is 2. The phase shift caused
to by low-pass filtering of the for fundamental fre-
quency component is known and is compensated in the soft-
ware. The small fundamental phase shift between supply and
capacitor voltages is not compensated in this case because it was
Fig. 5. Supposed low voltage equivalent circuit of an industrial power supply
found negligible. The other possibility to synchronize the input network.
current reference vector would be to detect the zero crossing
points of the filter capacitor voltage as suggested, e.g., in [12],
[17], and [18]. However, capacitor voltage contains more high stability of the MC system with supply voltage feedforward
frequency modulation ripple and additional transients, caused, depends on several parameters like grid impedance and supply
e.g., by load transients, than the supply voltage. The result is filter parameters. However, it has been presented in [17] that
that supply voltage can be considered simpler and more secure the PLL based synchronization of supply switching vector
voltage for the zero crossing detection than capacitor voltage. leads up to the stable operation even though measured supply
voltages are used in the calculation of modulation index by (15).
B. Method II Thus, the choice made in Section I because of the simplicity
The second IMC control method (Method II) is based on the should also provide stable operation in whole output power
calculation of the modulation index using measured line volt- range, as will be presented in Sections IV and V. In addition,
ages. The use of the input voltage measurement with space- unlike the method presented in [14], the input displacement
vector modulation to compensate the distortion has been pre- factor correction could also be used.
sented in [10][13]. In [14], it is used with the Venturini method With Method II the effects of supply voltage harmonics on
and it is also applied in [15]. the load voltages should be mainly prevented. The drawback of
When the modulation index is calculated by (9) using actual the method is that if the supply voltages are distorted the line
input voltages, it includes the harmonic content and the voltage current waveforms also become distorted due to the power bal-
produced does not contain input harmonics ance between the input and output. But as presented in [11], the
RMS of the supply current distortion should be smaller com-
pared to the case when the current was modulated along input
voltage vector .

C. Method III
(15) In the third control method of the IMC (Method III) a load
current feedback loop is added to the open-loop control system
where again it is assumed that 1. of Method I as presented in Fig. 4. The supply voltages are not
Also in this case the supply voltages , , and are measured, thus the calculation of the modulation index is based
measured instead of capacitor voltages as presented in Fig. 4. on the assumption of ideal sinusoidal mains. The synchroniza-
To filter modulation frequency components the first order tion of the line current reference vector is also defined with PLL
digital low-pass filter with the cut-off frequency of 3.18 kHz as in Method I.
was used in the software implementation. The reason for the The current references are given in two-axis form in
arrangement is that when the zero crossings are detected from stationary coordinates and . The measured load
the supply voltage it is easier to measure the voltage magnitudes currents are and , which are used to define the actual
from the same point than arrange another voltage measurement two-axis currents and . Both current controllers are iden-
setup. Naturally, the phase differences between the higher order tical discrete time antiwindup PI-controllers [7]. The PI-con-
voltage harmonics of the supply and capacitor voltages are trollers are used because they can be considered to be the basic
more remarkable and better results for load magnitudes could current control strategy when current control is an inner loop of
be obtained if the capacitor voltages were used [13]. cascade closed-loop control system in electric drives.
The angle of the input current reference vector is defined ex- The tuning of the controllers is based on the time constant
actly same as in Method I, i.e., with zero crossing detection of of the load on such a bandwidth that the frequency components
and PLL with 1. The measured voltage magni- reflected from the supply can be damped. The time constant of
tudes are used only in the calculation of the modulation index as the simulated load presented in Section IV is 0.33 ms, which is
presented in (15). Thus, the system is similar to those presented also the value of the integration time of the controller. To attain
and analyzed in [9], [11], [17], and [18]. the requirement of high bandwidth the gain of the controller was
With supply voltage feedforward the stability of the matrix set to 50. Thus, the theoretical bandwidth of the closed-loop cur-
converter system is an essential question [13], [17][19]. The rent control is 800 Hz, which in many practical drives can be too
JUSSILA AND TUUSA: COMPARISON OF SIMPLE CONTROL STRATEGIES 143

Fig. 6. Supply phase voltages in simulations: (a) waveforms, (b) spectrum, and (c) locus.

TABLE I
DISTORTED SUPPLY VOLTAGE
included is the 11th order. The relative magnitudes of harmonics
are the maximum allowed voltage harmonics allowed in the
standard [20] and their phase angles are similar to the phase
angles of the harmonics in industrial mains distorted by three-
phase diode rectifiers with capacitive loads. A total harmonic
distortion (THD) of the supply voltage is about 8.5%, which is
0.5% unit over the limit of 8% defined in [20].

high but now needed to damp the frequency components caused B. Simulation Model
by the distorted supply voltage. Thus, the capability of Method Simulations were performed with a combination of Simplorer
III depends on the controller tuning, which cannot necessarily and Matlab Simulink softwares. In the simulations -connected
always be done with the adequate bandwidth. As Method II, RL load with resistors of 30 and inductors of 10 mH was used
Method III should also produce more distorted supply currents that gives the load power of 3.9 kVA when the voltage ratio is the
than Method I. maximum 0.866. The modulation frequency was 5 kHz and the
modulator model was updated twice in one modulation period
D. Method IV [16].
The fourth IMC control method (Method IV) is a combination The prototype and type filter used was originally de-
of Methods II and III, as presented in Fig. 4. Thus it contains signed for 5-kVA rated power [16]. Thus, the supply filter design
both the supply voltage and load current measurements. Method was made supposing ideal sinusoidal supply conditions and its
IV should prevent the reflection of supply harmonics to the load main point was to fulfill the EMC requirements [21]. Desired
most effectively among the methods presented here. It should THDs of the supply current and filter capacitor voltage were
also produce the most distorted line currents. 4% and 5% when the frequency components up to 500 kHz
were taken into account. The resonance frequency was set to
IV. SIMULATIONS 1049 Hz. In the 5-kVA converter it means 2.3-mH (2.3%) in-
ductors and 10- F (10%) capacitors in -connection [21]. In
A. Distorted Supply Voltage the simulations, the damping and dc resistances of the real in-
In the typical low voltage industrial network presented in ductors were modelled connecting resistances of 88 in par-
Fig. 5 the impedance consists mostly of impedance caused allel and 55 m in series with each ideal inductance of 2.3 mH.
by distribution transformer and medium and high voltage net- The impedance of the power supply used in the measurements
works. Compared to the impedances , , of consisted of 0.03- resistor and 0.1-mH inductance in series
low voltage network are usually negligible, i.e., and these values were also used in the simulations to model the
, , , and . When power supply impedance.
the high frequency EMC in low voltage network is considered
the filtering caused by is not necessarily adequate C. Simulation Results
to provide safe operation of load in common case. Thus, As an ideal starting point to the other cases the simulation
or type filters of an IMC drive cannot be reduced to pure results with ideal sinusoidal supply voltages are presented in
type filter even though the would sometimes offer ade- Fig. 7 when Method I is applied. The voltage transfer ratio is 0.8
quate impedance for filtering, i.e., the point of common coupling 260 V 325 V and the load frequency is 40 Hz. As can be
(PCC) of the IMC is not generally in the supply side of the seen, the method, containing only the discrete-time space-vector
but inside the industrial network. Thus, in this study type modulator, produces slight distortion to currents and load volt-
line filter is applied. ages even with sinusoidal supply voltage because of inaccu-
The components of distorted supply voltage used are pre- racies produced by sampling, averaging of space-vector refer-
sented in Table I. Its relative phase voltage waveforms, spec- ences and discrete calculation. The spectra are presented with
trum, and locus are presented in Fig. 6. The highest harmonic the frequency resolution of 10 Hz up to 1 kHz.
144 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO. 1, JANUARY 2007

TABLE II
SIMULATION RESULTS

Fig. 7. Simulated waveforms and spectra (f = 10 Hz) of Method I with


sinusoidal supply voltage when f = 40 Hz: (a) supply voltage u and current
i , (b) spectrum of i , (c) waveform of load current i , and (d) spectrum of i .

Fig. 9. Simulated waveforms and spectra (f = 10 Hz) of Method II with


distorted supply voltages when f = 40 Hz: (a) line current i , (b) spectrum of
i , (c) load current i , and (d) spectrum of i .

Fig. 8. Simulated waveforms and spectra (f = 10 Hz) of Method I with


distorted supply voltages, when f = 40 Hz: (a) supply voltage u and current
i , (b) spectrum of i , (c) load current i , and (d) spectrum of i .

Method I was also simulated using the distorted supply


voltage of Fig. 6(a) instead of an ideal supply. The waveforms
and spectra of the currents produced are presented in Fig. 8.
When the load current spectra of Figs. 7(d) and 8(d) are
compared to each other, it can be seen that the supply voltage
harmonics reflect to output frequencies 260 Hz, 340 Hz, 560 Hz,
and 640 Hz, i.e., they are the frequencies 6 when
has values 1 ( 5, 7) and 2 11 as predicted in
Section II-B. The THDs of the simulated currents are presented
in Table II, where all THDs are calculated up to 2 kHz.
Fig. 10. Simulated waveforms and spectra (f = 10 Hz) of Method III with
Methods IIIV were simulated only with distorted supply distorted supply voltages when f = 40 Hz: (a) line current i , (b) spectrum of
voltage. The waveforms and spectra of the currents produced i , (c) load current i , and (d) spectrum of i .
by these methods are presented in Figs. 911. The THDs of cur-
rents and load voltage up to 2 kHz are presented in Table II.
In the simulations the current controllers of Methods III and filter inductance of 1 mH instead of 2.3 mH the resonance fre-
IV were tuned as presented in Section III-C. The feedback loops quency is 1591 Hz and the THDs with Method IV ( 40 Hz)
of the load currents included the first order digital low-pass fil- for , and were 17.0%, 4.3%, and 2.7%, respectively.
ters with 3.18-kHz cut-off frequency, thus the simulation models In the simulations, the approximated dc and parallel resistance
are equal to the system in the measurements presented in Sec- values of the inductor were also scaled with the smaller induc-
tion V. tance value. When the inductance value was reduced to 1 mH
As presented in Section IV-B the filter design was based remaining the original resonance frequency of 1049 Hz, i.e., ca-
on high frequency EMC requirements. However, the parameters pacitance of 23 F was used instead of 10 F, the THDs with
are not necessarily the optimum parameters when the low fre- Method IV ( 40 Hz) for , , and were 32.1%, 4.8%,
quency distortion of the supply is considered, e.g., by using the and 3.1%, respectively.
JUSSILA AND TUUSA: COMPARISON OF SIMPLE CONTROL STRATEGIES 145

Fig. 13. Photograph of the IMC prototype.


Fig. 11. Simulated waveforms and spectra ((f = 10 Hz) of Method IV with
distorted supply voltages when f = 40 Hz: (a) line current i , (b) spectrum of
i , (c) load current i , and (d) spectrum of i .

Fig. 12. Block diagram of the IMC system.

Comparing the results of the control methods presented it can Fig. 14. Measured waveforms and spectra (f = 10 Hz) of Method I with
sinusoidal supply voltage when f = 40 Hz: (a) supply voltage u and current
be observed that the supply voltage feedforward of Method II i , (b) spectrum of i , (c) waveform of load current i , and (d) spectrum of i .
can reduce the load current distortion considerably but not to-
tally. The first reason is that the magnitude of is used instead
of capacitor voltages. As an example Method IV ( 40 Hz) To attain the same current with the same voltage transfer ratio
was simulated using the capacitor voltage measurements when as in the simulations the -connected load consisted of
the THDs for , , and were 21.5%, 5.3%, and 3.2%, 27- resistors and 10-mH inductors that gives the load power
respectively, so that the effect of voltage measurement point is of 3.9 kVA with the voltage ratio of 0.866. filter parameters
not very remarkable in this case. The second reason is that cur- were 2.3 mH and 10 F as in the simulations.
rent is not modulated along the instantaneous supply voltage The control system was implemented by microcontroller
vector, which would be required to attain exactly constant output (Motorola MPC555), which defines the switching states and
power as presented in Section III-B. However, that would prob- their durations after every 100 s, i.e., it updates the modulator
ably cause stability problems as presented in [13], [17], and [19]. twice in the modulation period [16]. All control methods were
The current control of Method III cannot alone reduce the dis- implemented with the microcontroller. The required currents
tortion as well as Method II. To achieve sufficient compensation and voltages were measured using the ten bits A/D converters
with current control it has to be combined to the supply voltage of the microcontroller. The current filters and controllers of
feedforward as in Method IV. Open-loop control of Method I Methods III and IV had the same parameters as in the simula-
produces the most sinusoidal supply current but on the other tions.
hand its load current is highly distorted. As presented in Fig. 13, the size of the prototype is not
minimized but the aim has been to build a prototype easy to
V. MEASUREMENTS modify and test different control systems. The microcontroller
and driver boards are the general boards of the institute. In
A. IMC System addition, supply filter inductor and both line and load bridges
The IMC prototype is described in [16]. Its nominal load have higher current ratings than required by the system.
power is 5 kVA and modulation frequency is 5 kHz. Its block
diagram is presented in Fig. 12 and the prototype is shown in B. Measurement Results
Fig. 13. In the measurements, the distorted supply voltage with Measurement results of Method I with ideal sinusoidal
the amplitude of the fundamental component of 325 V was not supply voltage are presented in Fig. 14 and in Table III, where
available thus 305 V was used instead. Its locus and the relative all THDs are calculated up to 2 kHz. The amplitude of the
magnitudes of the harmonic components are the same as in the sinusoidal supply voltage was 305 V. Method I was also
simulations presented in Fig. 6 and in Table I. tested with distorted supply voltage. Measurement results are
146 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO. 1, JANUARY 2007

TABLE III
MEASUREMENT RESULTS

Fig. 16. Measured waveforms and spectra (f = 10 Hz) of Method II with


distorted supply voltages when f = 40 Hz: (a) line current i , (b) spectrum
of i , (c) load current i , and (d) spectrum of i .

Fig. 15. Measured waveforms and spectra (f = 10 Hz) of Method I with


distorted supply voltages, when f = 40 Hz: (a) supply voltage u and current
i , (b) spectrum of i , (c) load current i , and (d) spectrum of i .
Fig. 17. Measured waveforms and spectra (f = 10 Hz) 10 Hz) of Method
III with distorted supply voltages when f = 40 Hz: (a) line current i , (b)
spectrum of i , (c) load current i , and (d) spectrum of i .
presented in Fig. 15 and in Table III. The differences between
the results with ideal and distorted supply voltages are similar
as in the simulations. In each figure the waveforms are from
the same instant, e.g., the waveforms of the Fig. 14(a) and (b)
are simultaneous, etc. In addition, each spectrum calculation
has been started from the beginning point of the respective
waveform.
The measured waveforms and spectra of currents produced
by Methods II, III and IV are presented in Figs. 1618, respec-
tively. The THDs of currents and load voltages up to 2 kHz are
presented in Table III.

C. Comparison of Simulation and Measurement Results


Comparing Tables II and III it can be seen that the measure-
ment and the simulation results are essentially similar. The con- Fig. 18. Measured waveforms and spectra (f = 10 Hz) of Method IV with
trol methods containing compensation, i.e., Methods IIIV, can distorted supply voltages when f = 40 Hz: (a) line current i , (b) spectrum of
compensate the reflection of supply voltage harmonics to load i , (c) load current i , and (d) spectrum of i .
voltage and current. In all cases the compensation distorts the
line currents more than in Method I, which produces the most
sinusoidal line currents and the most distorted load voltages and of Method IV are presented in Fig. 21. The locus of the supply
currents. This can also be seen when the loci of currents pre- voltage in the simulations is presented in Fig. 6(c) and it is iden-
sented in Figs. 1921 are compared. tical to the locus of the voltage used in the measurements.
The relative loci of currents produced by Method I in simu- In the measurements Method IV produces the least distorted
lations and measurements are presented in Fig. 19. The corre- load currents in all cases, but the least distorted load voltages
sponding loci of Method II are presented in Fig. 20 and the loci are produced with Method II as also in the simulations because
JUSSILA AND TUUSA: COMPARISON OF SIMPLE CONTROL STRATEGIES 147

Fig. 19. Loci of currents with Method I, when f = 40 Hz: (a) simulated and
Fig. 21. Loci of currents with Method IV, when f = 40 Hz: (a) simulated and
(b) measured line currents; (c) simulated and (d) measured load currents.
(b) measured line currents; (c) simulated and (d) measured load currents.

semiconductors, and other non-idealities are not modelled ex-


actly in the simulations. Thus, the output current control system
of Methods III and IV also shows lower performance in the ex-
perimental tests than in the simulations so that the measured
output current distortion is higher and the input current distor-
tion is lower than in the simulations. However, the differences
mentioned are small and the simulations match generally the
measurements quite well.
The measurements confirm the characteristics already ob-
served in the simulations: voltage feedforward is needed to
attain less distorted load magnitudes with the IMC when the
supply voltage is distorted as expected.

VI. CONCLUSION
First, the conduction of harmonics through an MC was ana-
lyzed using space-vector presentation. After that, four control
methods of the IMC were presented and compared in simula-
tions and measurements. The methods required neither compli-
cated computing nor complex measurements.
It was observed that the open-loop control method of the
IMC without information of supply voltage magnitude or load
current produced most distorted load currents. The current
Fig. 20. Loci of currents with Method II, when f = 40 Hz: (a) simulated and feedback control could not alone compensate the effects of
(b) measured line currents; (c) simulated and (d) measured load currents. supply voltage harmonics on the load current sufficiently when
the supply voltage was assumed to be ideal. In addition, the cur-
rent controllers cannot necessarily be tuned in all drives so that
Method II does not contain current control. The differences be- adequate bandwidth is achieved. The sufficient compensation
tween simulation and measurement results are caused by the was achieved only with the methods where the measured supply
non-idealities of the realization. The accuracy of the voltage voltages were used in the calculation of modulation index. In
measurement in the prototype is worse than assumed in the sim- all cases, the compensation increased line current distortion.
ulations therefore Methods II and IV are not as effective in the The analysis presented is valid for both DMC and IMC. Thus,
measurements as in the simulations even though the difference the control methods presented as well as their comparison are
is quite small. In addition, stray inductances, delays, non-ideal also applicable to DMC.
148 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO. 1, JANUARY 2007

REFERENCES [15] J.-K. Kang, H. Hara, A. M. Hava, E. Yamamoto, E. Watanabe, and T.


Kume, The matrix converter drive performance under abnormal input
voltage conditions, IEEE Trans. Power Electron., vol. 17, no. 5, pp.
[1] L. Gyugyi, Generalized Theory of Static Power Frequency Changers, 721730, Sep. 2002.
Ph.D. dissertation, Elect. Eng. Dept., Univ. Salford, Salford, U.K., [16] M. Jussila, M. Salo, and H. Tuusa, Realization of a three-phase indi-
1970. rect matrix converter with an indirect vector modulation method, in
[2] P. W. Wheeler, J. Rodrguez, J. C. Clare, L. Empringham, and A. Wein- Proc. IEEE PESC03 Conf., Acapulco, Mexico, Jun. 1519, 2003, vol.
stein, Matrix converters: a technology review, IEEE Trans. Ind. Elec- 2, pp. 689694.
tron., vol. 49, no. 2, pp. 276288, Apr. 2002. [17] F. Liu, C. Klumpner, and F. Blaabjerg, Stability analysis and ex-
[3] Y. Minari, K. Shinohara, and R. Ueda, PWM-rectifier/voltage-source perimental evaluation of a matrix converter drive system, in Proc.
inverter without dc link components for induction motor drive, Proc. IEEE IECON03 Conf., Roanoke, VA, Nov. 26, 2003, vol. 3, pp.
Inst. Elect. Eng. B, vol. 140, pp. 363368, Nov. 1993. 20592065.
[4] L. Huber and D. Borojevic, Space vector modulated three-phase to [18] , A robust method to improve a stability in matrix converters, in
three-phase matrix converter with input power factor correction, IEEE Proc. IEEE PESC04 Conf., Aachen, Germany, Jun. 2025, 2004, vol.
Trans. Ind. Appl., vol. 31, no. 6, pp. 12341246, Nov./Dec. 1995. 5, pp. 35603566.
[5] J. W. Kolar, M. Baumann, F. Schafmeister, and H. Ertl, Novel three- [19] D. Casadei, G. Serra, A. Tani, and L. Zarri, Stability analysis of elec-
phase AC-DC-AC sparse matrix converter, in Proc. IEEE APEC02 trical drives fed by matrix converter, in Proc. IEEE ISIE02 Conf.,
Conf., Dallas, TX, Mar. 1014, 2002, vol. 2, pp. 777791. LAquila, Italy, Jul. 811, 2002, vol. 4, pp. 11081113.
[6] L. Wei, T. A. Lipo, and H. Chan, Matrix converter topologies with [20] Voltage Characteristics of Electricity Supplied by Public Distribution
reduced number of switches, in Proc. IEEE PESC02 Conf., Cairns, Systems, European std. EN 50160, CENELEC, 1999.
Queensland, Australia, Jun. 2327, 2002, vol. 1, pp. 5763. [21] N. R. Zargari, G. Joos, and P. D. Ziogas, Input filter design for PWM
[7] M. Jussila, M. Salo, and H. Tuusa, Induction motor drive fed by a current-source rectifiers, IEEE Trans. Ind. Appl., vol. 30, no. 6, pp.
vector modulated indirect matrix converter, in Proc. IEEE PESC04 15731579, Nov./Dec. 1994.
Conf., Aachen, Germany, Jun. 2025, 2004, vol. 4, pp. 28622868.
[8] P. Nielsen, F. Blaabjerg, and J. K. Pedersen, Space vector mod- Matti Jussila (S05) was born in Tammela, Finland,
ulated matrix converter with minimized number of switchings and in 1977. He received the M.S. (with honors) and Lic.
a feedforward compensation of input voltage unbalance, in Proc. Tech. degrees in electrical engineering from Tampere
IEEE PEDES96 Conf., New Delhi, India, Jan. 811, 1996, vol. 2, pp. University of Technology, Tampere, Finland, in 2001
833839. and 2005, respectively.
[9] P. Nielsen, D. Casadei, G. Serra, and A. Tani, Evaluation of the Since 2000, he has been with the Department of
input current quality by three different modulation strategies for SVM Electrical Engineering, Tampere University of Tech-
controlled matrix converters with input voltage unbalance, in Proc. nology. He was a Research Assistant from 2000 to
IEEE PEDES96 Conf., New Delhi, India, Jan. 811, 1996, vol. 2, pp. 2001. Since 2002, he has been a Researcher. His re-
794800. search interest is matrix converters.
[10] D. Casadei, G. Serra, and A. Tani, Reduction of the input current
harmonic content in matrix converters under input/output unbalance,
IEEE Trans. Ind. Electron., vol. 45, no. 3, pp. 401411, Jun. 1998.
[11] , A general approach for the analysis of the input power quality
in matrix converters, IEEE Trans. Power. Electron., vol. 13, no. 5, pp. Heikki Tuusa was born in Hattula, Finland, in 1957.
882891, Sep. 1998. He received the M.S., Lic.Tech., and Dr.Tech. de-
[12] F. Blaabjerg, D. Casadei, C. Klumpner, and M. Matteini, Compar- grees in electrical engineering from Tampere Univer-
ison of two current modulation strategies for matrix converters under sity of Technology, Tampere, Finland, in 1983, 1987,
unbalanced input voltage conditions, IEEE Trans. Ind. Electron., vol. and 1993, respectively.
49, no. 2, pp. 289296, Apr. 2002. Since 1983, he has been with the Department of
[13] D. Casadei, G. Serra, A. Tani, and L. Zarri, Effects of input voltage Electrical Engineering, Tampere University of Tech-
measurement on stability of matrix converter drive system, Proc. Inst. nology. He was a Laboratory Manager from 1983 to
Elect. Eng., vol. 151, pp. 487497, Jul. 2004. 1986 and in 1993, a Senior Assistant from 1986 to
[14] L. Zhang, C. Watthanasarn, and W. Shepherd, Control of AC-AC ma- 1992, an Acting Professor from 1994 to 1995, and an
trix converters for unbalanced and/or distorted supply voltage, in Proc. Associate Professor from 1995 to 1998. Since 1998,
IEEE PESC01 Conf., Vancouver, BC, Canada, Jun. 1721, 2001, vol. he has been a Professor and Head of the Institute of Power Electronics. His areas
2, pp. 11081113. of interest are PWM converters and controlled motor drives.

You might also like