Sunteți pe pagina 1din 7

I

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 7, NO. 4, OCTOBER 1992

101

Optimum Control of N-Input K-Output Matrix Converters


Pa010 Tenti, Senior Member, IEEE, Luigi Malesani, Member, IEEE, and Leopoldo Rossetto
Abstract-Significant developments of the general optimum control theory presented in a previous paper [3] are discussed for the specific case of multiphase matrix converters. Results hold, regardless of system configuration, input and output voltage waveforms, and loads. Applications to the most practical converter structures are illustrated, and implementation criteria of the optimum control method derived. Simulated results confirm the flexibility and effectiveness of the approach.

I. INTRODUCTION HREE-PHASE direct matrix converters, whose basic configuration and control were originally introduced in [ 13, [2], have recently received increasing interest [4][8]. In fact, PWM direct conversion techniques do not require reactive elements (except for small filters needed to remove the modulation harmonics) and also allow precise control of the phase and waveform of converter input currents 161. Small-size, high-efficiency , fast-response converters can be devised accordingly, but a penalty is paid in terms of circuit complexity since a high number of bidirectional fully controlled switches is required. Even the latter problem is being overcome, since high-power, easy-to-drive, low-cost power semiconductors are now available. The major open problem regards control, which should enforce the desired output waveforms (while taking into account the inherent voltage/current limitations of the converter) and also optimize behavior on the supply side, by keeping input currents sinusoidal and in phase with the line voltages. Solutions to this problem are available in the literature [l], [4], [6]-[8] for three-phase to three-phase matrix converters and under some simplifying assumptions (sinusoidal voltages, no energy loss). However, topologies other than three-phase to three-phase converters are of practical interest: for example, three-phase to single-phase converters for multifrequency supplies for test/lab equipment, six-phase to three-phase or six-phase to single-phase converters for aircraft applications, multiphase rectifiers, etc. Moreover, for multiconverter systems, the control problem should be approached as a whole to optimize the global input/output behavior while avoiding useless energy exchanges between converters, which affect system

efficiency [3]. Finally, for converters working, as active power filters [5], achieving input current waveforms following nonsinusoidal references is mandatory. For the above reasons, developing a general control theory capable of covering all converter topologies, connections, and operating modes is of prime interest. This problem was first approached in [3], in which a general optimum control theory was developed, based on the transformation of actual converter topology into a suitable equivalent structure. Optimum modulation laws referring to the latter structure were then derived in analytical form, but the antitransformation problem remained open. In this paper the application of the optimum control theory to N-input K-output matrix converters is discussed and general antitransformation criteria are developed. The most important converter configurations are then discussed in detail.
11. GENERAL REPRESENTATION OF N-INPUT K-OUTPUT MATRIX CONVERTERS Since the results presented here are based on the optimum control theory discussed in [3], we first summarize the main hypotheses and outcome of this theory. The general representation of a N-input K-output voltage-fed matrix converter is shown in Fig. 1, in which u;-uX are supply voltages and u;I-uk are load voltages. Two fundamental hypotheses are made on the cofiverter system. First, outputs are assumed to be decoupled, so that each of them can be controlled independently. This assumption is well verified by all voltage-fed mattix converters, if the internal impedances of the supply are small enough. Second, as shown in Fig. 1, the system is assumed to be made up only of ideal, lossless switches and transformers. This implies that possible output voltages (corresponding to all possible statuses of the converter switches) are independent of load currents and result from linear combinations of supply voltages. A. Single-Output Subsystems According to the above hypotheses, the N- to K-phase system of Fig. 1 can be split into a set of K N-phase to single-phase subsystems, as shown in Fig. 2. In turn, each single-output subsystem can be represented either by the scheme of Fig. 3(a) (star connection) or that of Fig. 3(b)

Manuscript received June 8, 1990; revised May 6, 1992. The authors are with the Department of Electrical Engineering, University of Padova, 35131 Padova, Italy. IEEE Log Number 9202486.

0885-8993/92$03.00 0 1992 IEEE

708

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 7, NO. 4, OCTOBER 1992

Input Tranformatlon Section

Smtching Section

I
Fig. 1 . General representation of N-input K-output matrix converters.

Fig. 4. Equivalent multilevel converter.

duce, depending on the supply voltages and status of its switches. In this representation, constant matrix T k is the L X N input transformation matrix, whose coefficients determine vector uk of the possible output voltages according to the equation:
Uk =

Tk

U'

(la>

where U' is the vector of the supply voltages. In Fig. 4 , the multiplexer represents the switching action of the actual converter which, at any time, applies one among voltages u k to the output. Let sk be the vector of the status variables (1 = on, 0 = off) of the multiplexer switches. Output voltage U; is given by
=U : .

sk = urT- T:

sk

(W

.-

where index T means transposition. Note from Fig. 4 that freewheeling action, if any, is considered separately. Discrete variable equations (lb) can be turned into continuous form by applying a time-averaging approach, in which status vector s k is substituted by vector mk of the switch modulation laws [2]. Modulation laws mk are constrained by the relations:

(a)

FJ
0 0

Lk

mkj

+ mfk
j

=
=

(2a)
1, L k

ImkJ I1 Imfk I 1

(2b) (2c)

which express the conditions that only one switch is closed at any time, and that the on-time of each switch must be greater than zero and lower than the commutation period. Now applying linear transformation T l to variables m k , we obtain the transformed modulation laws:
xk =

(b) Fig. 3. N-phase to single-phase configurations. (a) Star converter. (b) Bridge converter.

Ti

'

mk

(3)
(4a)

and, from (lb), we can write


= U'T

(bridge connection) in which, for the sake of generality, an N- to J-phase transformer is also included. Note that the output voltage produced by the star converter must coincide, at any time, with one among voltages ukl to uw, or can be zero (if free-wheeling switch sfk is actually present). Ipstead, in the bridge converter the output voltage can be zero or coincide with one among the line-to-line voltages. In any case, the representation of Fig. 4 (equivalent multilevel converter) can be adopted, in which voltages ukl to ukL are all output voltages the converter can pro-

. xk

Given output current $', input current behavior can be evaluated by balancing the input and output power. The result is i; = xk i;

(4b)

Equation (4b) shows that variables xk, like currents i;, must have zero sum at any time. Thus,

TENT1 er al.: OPTIMUM CONTROL OF N-INPUT K-OUTPUT MATRIX CONVERTERS

709

Additional limitations on variables straints (2).

xk

arise from con-

B. Multioutput Converters Since the number of variables xk equals number N of the supply phases, regardless of system configuration, (4a) and (4b) give a general model of single-output converters, which can easily be extended to cover the case of multioutput systems like that shown in Fig. 2. Let X be the matrix of the transformed modulation laws of all single-output subsystems:

B. Aclual Modulation Laws Given optimum transformed modulation laws xk, we must then determine status cr of the actual converter switches. This can be done in three steps:
1) First, we must reverse (3) to derive modulation laws mk of each equivalent multilevel converter as a function of variables xk. 2) Second, status vectors sk of the multiplexers must be derived, according to some PWM technique, from modulation laws mk. 3) Finally, status b k of the actual converter switches is derived from status vectors sk of the equivalent multilevel converters.

x = Ix, * - * xk * XKI For the K-output converter, we easily obtain


= XT
K

(53) (5b)

.
=

Ckii
I

(5c)

In practice, steps 2) and 3 ) are rapidly solved once the circuit topology and PWM technique have been selected. Instead, solution of the antitransformation problem (step 1)) is more complex. A possible approach is discussed below.
I v . ANTITRANSFORMATION OF THE MODULATION LAWS Before analysing antitransfonnation criteria, some basic definitions regarding converter topology must be introduced. The corresponding properties of the transformation matrix will then be discussed. A. Topologic Properties of Converters The topologic properties that are of interest for the antitransformation of the modulation laws are symmetry and completeness. They are defined as follows: 1) A converter is symmetric if the sum of its possible output voltages is zero, irrespective of actual input voltage behavior. 2) A converter is complete if the possible output voltage can be grouped in pairs of opposite voltages. Note that, by definition, completeness involves symmetry. Symmetry is normally met by polyphase converters with symmetric topology, but without neutral connection on the supply side (star converters with deltaconnected input transformers, etc .). Completeness is normally met by symmetric bridge converters.

where U and i f are vectors of output voltages and total input currents, respectively. Equation (5) gives a synthetic but complete input/output representation of whatever N- to K-phase matrix converter.
111. OPTIMUM CONTROL THEORY According to general representation (3,in [3] an optimum control theory was analytically developed, aimed at 1) Constraining system output voltages Uto track ref-

erences U* 2) Minimizing the discrepancy between total input currents if and references i* 3) Minimizing energy interchange between subsystems, causing minimum energy loss in the whole system. A. General Solution In [3] the optimum modulation laws were derived as a function of output voltage references U* and input current references i*. Here, instead of considering the general solution, we refer to the particular but practical case in which references i* are set at zero. According to the second property mentioned above, the corresponding modulation laws produce minimum total rms current absorption from the supply. These modulation laws are called inherent modulation laws and are given by

B. Properties of Transformation Matrix and Modulation Laws According to the above relations and definitions, the following properties of the transformation matrix can be demonstrated :
The sum of the elements of each row of matrix T is zero. This property, which is always valid, is simply derived from the condition that the sum of the supply currents is zero. I f the sum of all possible output voltages is zero, we can add the same quantity to all modulation laws without afecting the converter output voltage. This

Note that the inherent modulation laws only depend on the output voltage reference of the corresponding subsystem. Therefore, control of the various subsystems is completely decoupled.

710

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 7, NO. 4, OCTOBER 1992

is normally verified even in nonsymmetric converters, if the supply voltages are symmetric. 3) For symmetric converters the sum of the elements of each column of matrix T is zero. This property can be demonstrated by applying only one nonzero supply voltage in sequence to the various input terminals and then assuming the symmetry condition. It has the consequence that we can add the same quantity to all modulation laws o f symmetric converters without affecting output voltage or input currents. In other words, converter performance remains the same. f complete converters, the transfor4) In the case o mation matrix is made up of pairs of opposite rows. This property is easily demonstrated from the definition of completeness. It has the consequence that we can halve the number of possible output voltages by considering only one voltage of each pair of opposite voltages and accepting negative modulation laws. This both reduces system complexity and removes the constraint to accept positive modulation laws only. C. General Antitransformation Equations In the following, according to the assumption of independent output voltages, the antitransformation problem is discussed with reference to single-output subsystems. Note first that, due to constraint (4c), (3) gives only N 1 independent equations on variables m (subscript k will be omitted for brevity). Thus, (3) can be reversed only if number L of the possible output voltages is greater than or equal to N - 1. This condition is verified for all practical converters and does not affect the validity of the approach. In general L is greater than N - 1, and transformation matrix T cannot be directly reversed. However, let Lp be rank of matrix T: we can split T into a full-rank principal submatrix Tp and a secondary submatrix T,, which collects the remaining rows of T. Of course, choice of Tp and T, is arbitrary. Correspondingly, since each row of T is associated to one of the output voltages (i.e., to one modulation law), we can also split vector m into subvectors mp and m,. Note also that full-rank square matrix Tp T i (of size L,) can be reversed. Let

Establishing which are the acceptable solutions, if any, is not easy in general, since constraints (2) are difficult to handle in analytical terms. However, the general existence criteria of the solutions can be obtained in the cases of symmetric and complete converters.

D. Case of Symmetric Converters An important consequence of the above properties is that antitransformation is always possible for symmetric converters. In fact, assume all modulation laws m, to be zero. Accordingly, Eq. 8 gives:

m,=Q.x

(9)

in which variables mp do not necessarily satisfy constraints (2). However, we can add a term to all modulation laws (including m,), which makes the lowest among them equal to zero. Accordingly, we obtain the minimum positive modulation laws that satisfy (8): this gives the maximum range of regulation.

E. Case of Complete Converters In this case, solution (9) is always valid. In fact, as mentioned before, if some modulation laws turn out to be negative, we can simply make them positive and reverse the corresponding voltage (volpges with same amplitude but opposite polarity are available).
V. EXAMPLES OF ANTITRANSFORMATION In the following, antitransformation equations are derived for the most practical converter topologies.
A . Three-phase to Single-phase Star Converter The scheme is shown in Fig. 5 . Assuming, for simplicity, a transformer with a turn-ratio equal to 1, possible output voltages are u l , u2, u3, given by
U1

= U; = U;

U;
U;

(loa) (lob) (10c)

U2 U3

= U ; - U;.

The corresponding transformation matrix is

Q
and

(T, T,

Ti)-'
QT.

Tp

(74

R
From (3) we obtain

(7b)
=

(8) which is the fundamental antitransformation equation. System (8) gives Lp conditions on L variables m. Thus, if L > Lp there are infinite solutions, among which only those satisfying constraints (2) can be accepted.

mp + R T *m,

Q* x

and has rank equal to 2. Assuming, for instance, that Tp includes the first and second rows of T, and T, the third row, from (7) we obtain

Q=?l
3

-2 -1

'1

2 -1

TENT1 et al.: OPTIMUM CONTROL OF N-INPUT K-OUTPUT MATRIX CONVERTERS

71 I

1 1 1 \
Fig. 5. Three-phase to single-phase star converter. Fig. 6. Three-phase to single-phase bridge converter.

ml = -x2 System (8) gives two equations on variables m l , m2, and m3. They are ml - m3 = (--2.X1 m 2 - m3 = (-x1 m1

m2 = x3

m3 = 0.

(16c)

+ x2 + x3)/3
+ 2x2 - x3)/3

(12b) The third equation is obtained from constraint (2a):

+ m2 + m3 = 1

(1W

which holds, since the converter does not include the freewheeling switch. Finally, we have mi = (X3 - x1)/3 m 2 = (x2 - x3)/3 m3 = (xI - x2)/3.
(1 3 4 ( 13b)

These solutions are equivalent, in theory, as regards input and output converter performance. In practice, only the solutions that do not cause saturation of the modulation laws can be accepted. A good criterion is to modulate the output voltage, at any time, between the maximum and minimum possible output voltages: this avoids saturation, when possible, while giving the widest range of output voltage regulation.
C. Three-phase to Three-phase Matrix Converter The scheme is shown in Fig. 7. It includes three singleoutput subsystems of the kind shown in Fig. 8, with the difference that the three-phase configuration does not actually require the neutral connection. Possible output voltages are
U1 = U ;
U2

(1 3 4

B. Three-phase to Single-phase Bridge Converter The scheme is shown in Fig. 6. Possible output voltages are the six line-to-line voltages or zero voltage. In practice, instead of considering pairs of opposite output voltages, each associated to a positive modulation law, we refer only to three nonopposite voltages, while allowing their modulation laws to become negative. Assuming as possible output voltages:
U1 U2

U; U; U;

(1 7 4 ( 17b)

= U;

U3 = U;

(17c)

and, correspondingly, matrix T becomes 1 0 0 - 1

= U ; - U; = U; - U ;

(144 (14b) (14c)

T = 0 1 0 -1.
0 0 1 - 1 Note that this converter is not symmetric (the sum of all possible output voltages is zero only if supply voltages are symmetric). The rank of T is 3, and (8) gives ml = (3xl - x2 - x3 - x4)/4 m2 = ( - x i m3 = (-x1 (19d (19b) (194

U3 = U;

U;

the transformation matrix becomes 1 -1

0
1 (15)

T = -1
0

1 -1

+ 3x2 - x3 - x4)/4 - x2 + 3x3 - x4)/4.

Moreover, since neutral voltage U; is zero, variable x4 is also zero (6). Recalling now, from (4(c)), that the sum of the modulation laws x is zero, (19) gives: ml = x1 m2 = x2 m3 = x3. (20) Unfortunately, this solution does not comply with constraint (12c), which holds even in this case. Nevertheless,

712

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 1, NO. 4, OCTOBER 1992

U*-I I / I
Fig. 7. Three-phase to three-phase matrix converter.

1.1

*
v\/
-

\I

0-

f [HJ

Fig. 9. Simulated waveforms of three-phase to single-phase bridge converter.

Fig. 8. Three-phase to single-phase star converter with neutral connection.

we can find a solution in the common assumption that the supply voltages are symmetric. In this case, the sum of the possible output voltages is zero and, as mentioned before, the output performance of the converter is not affected if an arbitrary quantity is added to the modulation laws. We can therefore add, to (20), constant term $ to all modulation laws, thus obtaining m, = xI

+f m2 = x2 + f m3 = x3 + f

(21a)
(21b) (2 1c)

which is the desired solution, since it satisfies constraint (12d) while ensuring the required output voltage. In order to take full advantage of the inherent output voltage capability of the matrix converter, output voltage references should then be chosen according to the control method described in [4], by which the load midpoint voltage is controlled so as to achieve maximum load phase voltages. For this purpose, the quantity
&U* =

A. Three-phase to Single-phase Bridge Converter The scheme is that of Fig. 6. The output voltage reference was assumed to be sinusoidal at 130 Hz with the maximum amplitude, which does not cause saturation of the modulation laws, i.e., 1.5 pu. The output current was also assumed to be sinusoidal, of amplitude 1 pu, and delayed by 30. The optimum transformed modulation laws follow (6) the antitransformation being performed by (16). The criterion to modulate, at any time, between the maximum and minimum available output voltages was followed. The resulting converter performance is described by the waveforms of Fig. 9, which were obtained according to a time-averaging technique. In particular, from top to bottom, Fig. 9 shows supply voltage U;, input current i ; , output voltage U , load current i, and the harmonic spectrum of the input current. The latter shows only two harmonic components in addition to the fundamental. These harmonics, respectively, at 210 Hz and 310 Hz, are needed to compensate for output power fluctuation.

i[max {U}

+ min {U}]

must be added to all voltage references. VI. EXAMPLES OF APPLICATION As significant applications of the above theory, the cases of three-phase to single-phase bridge converter and three-phase to three-phase star converter were simulated. In both cases, supply voltages were assumed to be sinusoidal at 50 Hz and symmetric, their peak value being 1 pu.

B. Three-phase Matrix Converter The scheme of this converter, whose performance has been deeply analyzed in the literature in the hypothesis of sinusoidal operation, is shown in Fig. 7. Here, a different case is considered, which refers to a six-step 80-Hz output voltage with the maximum possible fundamental component (0.95 pu). This case is of practical interest for motor drives in which distorted output voltages can be allowed for high rotational speed if this gives better exploitation of converter voltage capability (maximum output voltage under sinusoidal operation is 0.86 pu). Optimum transformed modulation laws are obtained from (6), while (21) is used for the antitransformation. Output voltage references were modified according to (22). The corresponding simulated waveforms are shown in Fig. 10. The order is the same as in Fig. 9. From top to bottom:

TENT1 et al.: OPTIMUM CONTROL OF N-INPUT K-OUTPUT MATRIX CONVERTERS

713

[6] D. G. Holmes and T. A. Lipo, Implementation of a controlled rectifier using ac-ac matrix converter theory, in Proc. IEEE-PESC Conf., Milwaukee, WI, 1989, pp. 353-359. [7] J. Oyama, T. Higuchi er a l . , Novel control strategy for matrix converter, in Proc. IEEE-PESC Conf.,Milwaukee, WI, 1989, pp. 360367. [8] G. Roy and G. E. April, Cycloconverter operation under a new scalar control algorithm, in Proc. IEEE-PESC Conf. , Milwaukee, WI, 1989, pp. 368-375.

0-

f [Hz]

Fig. 10. Simulated waveforms of three-phase direct matrix converter.

supply voltage U;,input current i;, output voltage U;, load current i; (the latter is assumed to be sinusoidal and delayed by 30), and the harmonic spectrum of the input current. Note that, in spite of the high output voltage distortion, the harmonic content of the input currents is relatively small. In fact, in this case too, the input current harmonics are only those that are strictly needed to compensate for output power fluctuation. This was confirmed by further simulations, which showed that, in the same operating conditions but for a resistive load (with six-step output currents and constant output power), input currents were perfectly sinusoidal and in phase with the supply voltages. VII. CONCLUSIONS A general model of N-input K-output matrix converters is presented and applied to derive optimum control criteria. The topologic and control properties of the most common matrix converters are then analyzed, in view of the practical applications of the optimum control theory. Examples of application are given, showing the flexibility and power of the proposed approach, which offers a general optimum control tool, and allows the inherent capabilities of the converter to be exploited fully.

Paolo Tenti (M85-SM91) was born in Bolzano, Italy, in 1951. He received the Dr. degree with honors in electrical engineering from the State University of Padova, Italy, in 1975. Subsequently, he began work as a lecturer of electrotechnics in the Institute of Electric and Electronic Engineering at the State University of Padova and a contract researcher of the National Research Council of Italy (CNR). In 1979 he became a contract professor of power electronics at the universitv. , , in 1981 a oermanent researcher and in 1985, an associate professor. In 1990 he became full professor of power electronics at the State University of Catania, Italy, and since 1991 he is professor of Industrial Electrical Applications and Power Electronics at the University of Padova. His main interests involve industrial electronics, static power conversion, and industrial drives. His current activities include PWM rectifiers, active power filters, dc/dc power supplies, control of PWM converters, and ac drives. Dr. Tenti is a member of the Industry Applications Society, the Power Electronic Society, the Industrial Electronics Society, the IAS Industrial Power Converter Committee (IPCC), and the Italian Association of Electric and Electronic Engineers (AEI). He is also a member of the European subcommittee of the IPCC and Member-at-Large of the IAS Executive Board.

Luigi Malesani (M63) was born in Lonigo, Vicenza, Italy, on Sept. 18, 1933. He received the doctor degree in electrical engineering, with honors, from the University of Padova in 1962. From 1963 to 1964 he was employed as a researcher in the Centro Gas Ionizzati of CNR. From 1964 to 1975 he was assistant of electrical engineering and, from 1968 to 1975, Associate Professor of electronic components at the University of Padova. Since 1975 he has been a Professor of Applied .. Electronics at the University of Padova. His interests are in power electronics, circuit design, electrical machines and automatic control. He has written a number of papers on these subjects. Dr. Malesani is a member of the Italian Association of Electric and Electronic Engineers.

REFERENCES
[I] M. Venturini, A new sine wave in, sine wave out, conversion technique eliminates reactive elements, in Proc. Powercon 7, San Diego, CA, 1980, pp. E3-El5. [2] A. Alesina and M. Venturini, Solid state power conversion: A Fourier analysis approach to generalized transformer synthesis, IEEE Trans. Circuits Sysr., vol. CAS-28, no. 4, pp. 319-330, 1981. [3] P. Tenti, L. Malesani, and L. Rossetto, Optimum control of PWM multiconverter systems, in Proc. IEEE-IAS Annu. Meer., Pittsburgh, PA, 1988, pp. 888-894. 141 A. Alesina and M. Venturini, Intrinsic amplitude limits and optimum design of 9-switch direct PWM AC-AC converters, in Proc. IEEE PESC Conf.,Kyoto, Japan, 1988, pp. 1284-1291. 151 L. Rossetto and P. Tenti, Using ac-fed PWM converters as instantaneous reactive power compensators, IEEE Trans. Power Electron., vol. 7, no. 1, pp. 226-230, 1992.

Leopoldo Rossetto was born in Santa Maria di Sala (Venezia) Italy, in 1960. He received the Dr. degree (with honors) in electronic engineering from the State University of Padova, Italy, in 1985 and the Ph.D. degree in electrical engineering at the Department of Electrical Engineering of the same University where, since 1990, he works as a Permanent Researcher. His research interests are in the fields of applied electronics, power electronics, control techniques, and digital simulation.

S-ar putea să vă placă și