Modern Solutions for Industrial Matrix-Converter Applications

Transcription

1 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL Modern Solutions for Industrial Matrix-Converter Applications Olaf Simon, Jochen Mahlein, Mark Nils Muenzer, and Manfred Bruckmann Abstract Thus far, matrix converter topology has not reached industrial use. Reasons for this have been complex control schemes, inefficient solutions for overvoltage protection, questionable robustness of the bi-directional switch commutation at disturbed supply voltages, and missing power semiconductor modules. Recent research results concerning these problems have led to innovative solutions and, from a technical point of view, industrial use seems to be reasonable now. Index Terms AC AC power conversion, matrix converter, matrix module, overmodulation, overvoltage protection, robust commutation, space-vector control. I. INTRODUCTION THE concept of matrix-converter (MC) topology, a forced commutated direct ac-to-ac converter, has been known for more than 25 years [1], [2]. The advantages are the capability of power regeneration, a sinusoidal input current, and no need for energy storage. These features are very advantageous for the modern trends in drive technology such as decentralized drive systems. If the converter is located near the motor or even integrated with the motor, compact converter solutions are necessary. Voluminous braking resistors are no longer needed with the capability of regenerating energy. Furthermore, several MC drives are able to exchange energy between each other. This is state of the art for centralized dc-link converters but not for ac-coupled decentralized drives. Without electrolytic capacitors, the MC converter has a long lifetime even with high-temperature surroundings which is usually the case with motor integrated drives. Despite all these advantages, MC technology has not yet found applications. Reasons for this have been the difficulty in understanding the control scheme, economic aspects due to the large amount of switches, expensive solutions for overvoltage protection, and questionable robustness concerning the commutation process. During the last years novel solutions for the mentioned problems have been developed and will be summarized in this paper. II. MATRIX CONVERTER The MC directly connects three input phases with three output phases (Fig. 1). Additionally, a three-phase input filter Fig. 1. Matrix converter. consisting of a capacitor and an inductor is necessary to eliminate current components with pulse frequency from the line current. Discussing the basic features of the MC from a technical point of view, it is a more advanced solution compared to a diode-fed voltage-source converter but provides less features as an active front end (forced commutated line-side converter). Often, the limited output voltage range of the MC, which is for sinusoidal operation, is mentioned as a disadvantage. But this isn t quite true. Motor integrated converters usually operate with a reduced dc-link capacitor. Therefore, the dc-link voltage follows of Fig. 2 and not the smoothed line above. For generating sinusoidal output voltages, only can be taken into account which is almost equal to the fictitious dc-link voltage of the MC. Consequently, the MC can generate the same output voltage as a diode-fed voltage source converter with a reduced dc-link capacitor. Certainly, it is true that the MC does not have the capability of stepping up the input voltage like an active front end. On the other hand, concerning the shape of the line current and the power regeneration, the MC equals active front-end technology. III. IMPLEMENTATION OF RIV-MODULATION To reach acceptance for a new technology like the MC, it is quite helpful to give an easy explanation. Using rectifying and inverting vector modulation (RIVM), the MC control can be explained by the same methods known from dc-link converters. In [3], the validity of the following space-vector transfer functions has been shown: Manuscript received April 21, 2001; revised August 15, Abstract published on the Internet January 9, O. Simon and M. Bruckmann are with Siemens AG, D Erlangen, Germany ( olaf.simon@erlf.siemens.de). J. Mahlein is with the Elektrotechnisches Institut, University of Karlsruhe, D Karlsruhe, Germany ( mahlein@eti.etec.uni-karlsruhe.de). M. N. Muenzer is with Eupec GmbH + Company KG, D Warstein, Germany ( marc.muenzer@eupec.com). Publisher Item Identifier S (02) The complex modulation indices and are free of choice independently within the region of the corresponding hexagon given in Fig. 3. (1) /02$ IEEE

2 402 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL 2002 Fig. 2. Comparison of dc-link voltages. Fig. 3. Value ranges of the complex modulation indices. The real part of the first equation in (1) can be considered as a fictitious dc-link voltage and kept constant with an appropriate choice of for any input voltage. Afterwards, is chosen to create the desired output voltage. The method of obtaining the switching states and corresponding switching times per modulation period from the two complex modulation indices is similar to the space-vector modulation of voltage source converters. Considering Fig. 3, the geometric position of within the hexagon defines how long the two edges (2) and (3) have to be used. Therefore, the times and are defined. All other times are set to zero because the corresponding positions are not used. In this way, a vector of times for each edge called the modulation index vector is defined. The same procedure has to be applied for the complex modulation index and is similar to a current source converter. Next, the vector product has to be calculated, leading to a matrix called the modulation index matrix. Only four adjacent elements are unequal to zero. This matrix corresponds to the switching-state matrix given in Fig. 4, thus defining which switching states have to be chosen and how long. During the remaining time of the modulation period, a zero switching state 111, 222, or 333 has to be chosen. Considering the example, the switching state 333 is useful because in this way the phase 2 has no switching instant over the modulation period. As a last step, a useful sequence within the modulation period has to be defined. The procedure is illustrated in Fig. 5. To minimize switching instances, it is necessary to follow an -shape Fig. 4. Derivation of switching states from the complex modulation indices of the MC. order from to and to place the zero switching state in the center. Afterwards, the whole switching sequence is followed vice versa from to. For other cases, if the sum of sector numbers of and is unequal, a -shape, as indicated in Fig. 4 in gray, is necessary. The results of the proposed algorithm can be fed to a standard pulsewidth modulator (PWM) using a triangular carrier to control a single output phase. Fig. 6 gives the example for output phase 1. Different compared to conventional modulators, two reference values are used and information about the order of

3 SIMON et al.: MODERN SOLUTIONS FOR INDUSTRIAL MC APPLICATIONS 403 Fig. 7. Switching sequence for voltage-controlled commutation from A to B (+ = turning on, 0 = turning off). Fig. 5. Sequence of switching states. Fig. 8. Output voltage of one phase with respect to the line neutral by operations with RIVM and zero input phase shift. Fig. 6. Modulator for output phase 1. input voltages that have to be taken during the modulation period (sequence) is necessary. Using the RIVM with an input phase shift of zero, which is desirable for maximum output voltage range, a very interesting side effect has been observed and will be explained in the following section. IV. ROBUST COMMUTATION Due to not avoidable time lags between two switching commands, the commutation process in MCs is quite complicated. To avoid short circuits at the input and open circuits at the output, only bi-directional switches with two separate gates for each current direction are applicable for snubberless operation. Such a switch can be realized by two anti-serial insulated gate biplar transistors (IGBTs). Nevertheless, the knowledge of the voltage or current sign is necessary to choose the right switching sequence. In this case, the voltage-controlled commutation (Fig. 7) is chosen for two reasons. First, the risk of short circuits is tolerable for IGBTs, whereas the risk of open circuits is dangerous. Second, an input voltage measurement is necessary for the voltage control in any case whereas a current measurement is optional. So far, realizations required a very precise sign measurement usually implemented in addition to the analog measurement used for the control. This is expensive and there is a remaining risk of false measurements still, especially if the input voltage is strongly disturbed. Additional stress to the semiconductors or even destruction might follow. Ziegler has been the first researcher to consider this problem and proposed an interesting solution [4]. The idea was the replacement method. If the voltage sign measurement between phase A and B is unsure, a commutation to the third available phase C is executed, directly followed by a second commutation from phase C to A. Disadvantages are a small output voltage error and increasing switching losses. A new method has been found by using the RIVM with zero input phase shift [5]. It is called the avoiding method [6] and is explained with Fig. 8 demonstrating the resulting output voltage under the stated conditions. It can be found that simply no commutation occurs between the two voltages which are close together. The modulation automatically prevents critical commutations. It turns out that this is a general feature over the whole input voltage period. Using this method, it is no longer necessary to have a precise sign measurement and the analog input voltage measurement used for the control anyway is sufficient to provide the voltage sign information for the commutation control. The presented experimental results are the first ones measured for an MC without an explicit voltage or current sign measurement. To verify the achieved robustness of the commutation process using the proposed technique, the MC has been operated at a heavily disturbed input voltage (Fig. 9). The voltage has been measured at the line filter capacitor of the MC. Whereas rare short-circuit currents have been observed with conventional explicit voltage sign measurement for commutation, no short-circuit currents have been found during operation with the new method. Because critical voltage slopes at the converter input are limited by the line side filter, it can be concluded that no line voltage disturbance can harm the commutation process as long as a line filter resonance is damped by additional resistors or varistors.

4 404 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL 2002 Fig. 9. Input voltage during robustness test. Fig. 10. Output voltage (phase to neutral) and input current during overmodulation. V. IMPROVING THE VOLTAGE GAIN Although the voltage gain of the MC, which is for sinusoidal operation, is comparable to diode-fed voltage source converters as previously shown, an academic race to higher gains has broken out. The mentioned gain for sinusoidal operation is a theoretical limit because step-up operation is not possible without changing the given topology. Nevertheless, nonsinusoidal operation in the overmodulation mode can lead to a higher magnitude of the fundamental frequency of the output voltage. The RIVM theory can give a theoretical answer. The largest output magnitude can be expected if each possible switch state of (edges of the hexagon) and is kept for 60. With a Fourier series, the complex modulation indices can be formulated as (a) Reducing this expression to the first harmonic and inserting it into (1) with the assumption of a sinusoidal symmetric source, we obtain It is interesting to find that the first harmonic of the output voltage is even slightly higher than the input voltage. Experimental validation has been published in [7]. Measurements from that publication made with a 30-kVA MC with an induction motor load are shown in Fig. 10. Fig. 11 shows the Fourier transformation of these values. As the first harmonic of the input voltage was 331 V, an output voltage factor of 99.3% has been experimentally validated. Certainly, this type of operation in overmodulation mode has several disadvantages. Neither the output voltage nor the input current is sinusoidal. Additionally, due to the nonsinusoidal current, the input filter will be periodically forced to a resonance and longterm operation without strong damping measures probably will not be tolerable. (2) (3) (b) Fig. 11. Fourier transformation of (a) output voltage and (b) input current during over modulation (magnitudes). VI. OVERVOLTAGE PROTECTION Due to a missing free-wheeling path in the MC topology, the shutdown of all gate signals, which might occur if the supply voltage fails, e.g., during a three phase short circuit, is very critical. The inductive load would force the output current to flow, and destructive overvoltage at the semiconductors would result. To overcome this problem, an additional six-pulse rectifier added by a capacitor has been used in the past as a snubber circuit at the output to take the inductive load energy after a gate shutdown [8]. In [9], it is shown that a simple solution with transil diodes at the gate drives and output varistors can solve this problem with less volume and expense. VII. NEW IGBT POWER MODULE Compact realizations of MCs are only achievable if specialized IGBT modules are available containing BDSs. A first step has been done by Eupec developing a module containing tree

5 SIMON et al.: MODERN SOLUTIONS FOR INDUSTRIAL MC APPLICATIONS 405 Fig. 12. Matrix converter module by Eupec. (a) (b) Fig. 14. Measurements of voltages and currents at the (a) input and (b) output. Voltages are measured with reference to an artificial neutral point of each three-phase system. Fig. 13. Power stage of the Siemens MC prototype. BDSs that could be realized by a single phase of an MC. A further step of integration has been reached by Eupec in cooperation with Siemens by introducing an all-in-one MC module (Fig. 12, [10]). It contains all 18 necessary IGBTs and diodes of the 3 3 switch matrix in a single housing. The common collector switch topology is used leading to only six emitter connections that need to be supplied with energy for the gate drivers. Several 35-A/1200 V IGBT Chips and EMCON HE diodes are mounted on a centrally symmetrical PCB structure based on EconoPIM3 frame technology. All terminals are sorted into six groups that are placed around the module, leading to a very compact solution but keeping all insulation necessities. VIII. EXPERIMENTAL RESULTS An MC prototype with 7.5-kW nominal power for a 400-V line has been realized at Siemens (Fig. 13). The converter is running at a pulse frequency of 10 khz. The Eupec MCmodule is mounted on the heat sink. On the right-hand side, the input capacitors in delta connection and the line inductor are located. These components are rated to 1 mh and 3.3 F. The top printed board contains the gate drives, overvoltage protection, and analog measurement circuitry. The described implementation of the RIVM has been implemented as a short C-Code into a Texas-Instruments 320C40 signal processor connected to an FPGA containing the modu- lators and commutation control. Results of voltage and current at the input and output are given in Fig. 14, demonstrating the advantageous input power factor and sinusoidal current shape at the input and output. A series connection of a resistor of 5.6 and an inductor of 5 mh has been used as load. The efficiency has been measured with the same output current using a resistor of 11.2 to increase the output voltage to 80% of the maximum sinusoidal output voltage. At an output power of kw, the losses of the power circuit excluding the line inductor have been 305 W, resulting in an efficiency of 95.6%. At higher output voltage levels with the same output current, even higher efficiencies can be expected as the losses are mainly dependent on the output current. Further, the MC has been operated at a 7.5-kW induction motor using a sensorless field-oriented control. Fig. 15 shows the regenerating operation during the deceleration of the drive. This measurement proves the dynamic capability of the MC as well as the full braking capability by regenerating the mechanical energy. IX. CONCLUSION The advantageous features of the MC, for example, in decentralized drive systems, have led to a great deal of research effort in the past few years. Some of the results have been presented. A closed mathematical description of the MC s transfer function enables the use of conventional modulation methods and

6 406 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL 2002 [9] J. Mahlein and M. Braun, A matrix converter without diode clamped over-voltage protection, in Proc. 3rd IPEMC, Beijing, China, [10] M. Hornkamp, M. Loddenkötter, M. Muenzer, O. Simon, and M. Bruckmann, EconoMAC the first all-in-one IGBT module for matrix converters, in Proc. Drives and Controls and Power Electronics Conf., London, U.K., Mar , 2001, pp Fig. 15. frequency, P = output power). Operation of an induction motor (Iq = torque current, f = estimated Olaf Simon received the Dipl.-Ing. and Dr.-Ing. degrees from the University of Karlsruhe, Karlsruhe, Germany, in 1993 and 1998, respectively. In 1998, he joined Siemens AG, Erlangen, Germany, where he has been working in the Department for Innovation and Technology for electrical drives. His areas of interest are power electronic devices, new converter topologies, and the impact of control on converter system performance. gives a theoretical solution for the overmodulation mode to improve the voltage gain. Control methods to ensure a robust commutation even under heavily disturbed line voltages have been described and experimentally verified. Overvoltage protection without clamp capacitors and a new MC IGBT module have enabled the realization of a very compact design with an efficiency of 95%. Future demands of compact energy-saving drives providing clean power can be solved using the MC topology. REFERENCES [1] A. Alesina and M. Venturini, Intrinsic amplitude limits and optimum design of 9-switches direct PWM AC AC converters, in Proc. IEEE PESC 88, vol. 2, Kyoto, Japan, April 11 14, 1988, pp [2] L. Huber, D. Borojevic, and N. Burany, Analysis, design and implementation of the space-vector modulator for forced-commutated cycloconverters, Proc. Inst. Elect. Eng., pt. B, vol. 139, no. 2, pp , Mar [3] O. Simon and M. Braun, Theory of vector modulation for matrix converters, presented at the 9th European Conf. Power Electronics and Applications, Graz, Austria, Aug , 2001, Paper L1a-3. [4] M. Ziegler and W. Hofmann, A new two steps commutation policy for low cost matrix converters, in Conf. Proc. PCIM 2000 Europe, Nuremberg, Germany, [5] O. Simon and M. Bruckmann, Control and protection strategies for matrix converters, in SPS/IPC/Drives, Nuremberg, Germany, 2000, p [6] J. Mahlein, O. Simon, J. Igney, and M. Braun, Robust matrix converter commutation without explicit sign measurement, presented at the 9th European Conf. Power Electronics and Applications, Graz, Austria, Aug , 2001, Paper DS2.5. [7] J. Mahlein, O. Simon, and M. Braun, A matrix converter with space vector control enabling overmodulation, presented at the 8th European Conf. Power Electronics and Applications, Lausanne, Switzerland, Sept. 7 9, [8] C. Klumpner, P. Nielsen, I. Boldea, and F. Blaabjerg, A new matrix converter-motor (MCM) for industry applications, in Conf. Rec. IEEE-IAS Annu. Meeting, Rome, Italy, Jochen Mahlein received the diploma engineer title from the University of Karlsruhe, Karlsruhe, Germany, in Since 1996, he has been an Assistant Professor in the Electrotechnical Institute, University of Karlsruhe. He teaches courses in Control of Electrical Drives and Practice of Electrical Drives. His research areas are direct converters, soft-switching converters, and control of electrical drives. Mark N. Muenzer received the Dipl.-Ing. degree in electrical engineering from the Rheinisch-Westfaelische Technische Hochschule (RWTH) Aachen, Aachen, Germany, in Since then, he has been with Eupec GmbH + Company KG, Warstein, Germany, first as a Marketing Engineer, and then, since 1998, as a Development Engineer. His area of interest is the development of power semiconductor modules. Manfred Bruckmann received the Dipl.-Ing. degree in electrical engineering from Friedrich Alexander University, Erlangen, Germany, in Since 1991, he has been with the Department for Innovation and Technology, Siemens AG, Erlangen, Germany. His research interests are new power semiconductor devices, high-voltage IGBT applications, and new converter topologies.