Method for controlling a multi-phase two-point inverter and multi-phase two-point inverters

Abstract

Method for controlling a two-point inverter (1) with at least two phases (P), wherein the two-point inverter (1) has a switching arrangement (2) with - Input terminals (8) for the two poles of an input DC voltage (U ZK ), - one output terminal (A) per phase (P) P ), - a bridge circuit (6) with one half-bridge per phase (P), wherein the half-bridges are each circuit breakers (S P V ) have, via which the respective output port (A) P ) is connected to the input terminals (8), and - a relief network (7) with -- a bidirectional relief switch (B P ) per phase (P) and -- an inductance (L) for the at least two phases (P) that are connected via the respective bidirectional load shedding switch (B) P ) with at least two output terminals (A P) is connected, the procedure comprising the following steps: - Performing multiple switching cycles to generate an alternating current (i P ) at the output terminals (A P ), where each of the output terminals (A P ) the potential (ϕ) applied to at least two phases (P) P ) the input DC voltage (U ZK ) by switching the respective circuit breakers (S P V ) is reversed, - Activating the relief switches (B P ) to reduce switching losses when switching the respective circuit breakers (S P V ), preventing the relief switches (B P ) different phases (P) are activated simultaneously, while the output terminals (A P ) the respective phases (P) at different potentials of the input DC voltage (U ZK ) lay, - where switching times ( t P , t P ' ) the circuit breaker (S P V ) two phases (P) are shifted relative to each other in such a way that the respective relief switches (B P ) are activated with a time delay to relieve the switching process, - where an averaging of the relative shift of the switching times ( t P , t P ' ) the phases (P) occur over several switching periods.

Description

[0001] The invention relates to a method for controlling a multi-phase two-point inverter, for example a three-phase two-point inverter. The invention also relates to a two-point inverter with a switching arrangement.

[0002] Two-level inverters are used in power electronics to convert direct current (DC) voltage into alternating current (AC) voltage. They are used, for example, in automotive drive converters or photovoltaic (PV) inverters. The inverter's power switches convert the applied DC voltage into a pulsed voltage through appropriate control, thus replicating an AC voltage. The switching period usually differs from the period of the fundamental frequency of the output voltage. The power switches must therefore switch under voltage and current load, which leads to switching losses. To reduce these losses, load balancing networks or auxiliary circuits are used, enabling soft switching, such as zero-voltage switching (ZVS).One proven effective method is the so-called Auxiliary Resonant Commutated Pole (ARCP) configuration, as described, for example, in EP 0 500 818 B1. This load-suppression network requires only one bidirectional switch and one inductor per output phase. When the bidirectional switch closes, a load-suppression current can build up in the inductor, which counteracts the load current and thus ensures low-loss charging of the circuit breakers being switched during the phase switching process. A disadvantage of this is that one inductor is required per phase, which leads to increased weight and volume of the inverters configured accordingly.

[0003] To reduce the number of passive components, W. Gong et al., in "A Novel Auxiliary Resonant Commutated Pole Soft-Switching Inverter" (2021 IEEE Applied Power Electronics Conference and Exposition (APEC), pages 2166 to 2170), proposed combining the load-sensing network for different polarities of the voltage change, i.e., for different switching edges. This approach requires an inductor for each direction of the switching edge reversal. A further reduction in passive components cannot be achieved this way. Furthermore, the described control method does not allow for adjustment of the average neutral point voltage.

[0004] CN 1 01 018 023 A discloses a three-phase two-point inverter with a load balancing network. DE 10 2019 135 718 A1 discloses an inverter and a method for operating it.

[0005] The object of the present invention is to improve a two-point inverter with switching relief, in particular to reduce the number of its passive components.

[0006] This problem is solved by a method for controlling a two-point inverter with at least two phases according to claim 1 and a multi-phase two-point inverter according to claim 7.

[0007] The method according to the invention serves to control a two-phase inverter with at least two phases, which has a switching arrangement with input terminals for an input DC voltage, one output terminal per phase, and a bridge circuit with one half-bridge per phase, wherein the half-bridges each have power switches via which the respective output terminal is connected to the input terminals. The multi-phase two-phase inverter to be controlled also has a load shedding network with a bidirectional load shedding switch for each phase and an inductor for the at least two phases, wherein the inductor is connected to the at least two output terminals via the respective bidirectional load shedding switch.Several switching cycles are performed to generate an alternating current at the output terminals, whereby the potential of the input DC voltage at the output terminals of at least two phases is reversed by switching the respective power switches. The load shedding switches are activated to reduce switching losses during the switching of the respective power switches, preventing the load shedding switches of different phases from being activated simultaneously while the output terminals of the respective phases are at different potentials of the input DC voltage.

[0008] The core of the method consists of the fact that the inductance of the load balancing network is connected to multiple output terminals via the respective load balancing switches. When the load balancing network is activated, coupling of several phases can occur, which can cause short circuits and damage to the two-phase inverter and connected devices. Couplings leading to short circuits are referred to here and in the following as collisions. According to the invention, it has been found that safe and collision-free operation of the circuit breaker is ensured, despite the single inductance serving different phases, by controlling the two-phase inverter in a way that prevents the load balancing switches of different phases from being active simultaneously when their respective output terminals are at different potentials.This reduces the switching load in the two-phase inverter, as only one inductor is required for multiple phases, and in particular, exactly one inductor for all phases of the two-phase inverter. The method enables the realization of two-phase inverters with a simple load-relief network, in which the number of passive components, especially the number of inductors, is reduced. This allows for a significant reduction in the weight and volume of the two-phase inverters. Additionally, the component costs of the two-phase inverter are reduced. The two-phase inverter according to the invention particularly enables bidirectional power flow.

[0009] A two-phase inverter has at least two phases. The input DC voltage is converted into at least two AC phases. Such a two-phase inverter is also called a multi-phase two-phase inverter. Specifically, a two-phase inverter has three phases.

[0010] The generated alternating current is also referred to as load current. The direction of the load current is defined in particular by the current flowing out of the two-level inverter.

[0011] The switching cycles for generating alternating current can be carried out in a known manner, for example, using known modulation techniques. For each switching cycle, a modulation state is determined that defines the switching points of the circuit breakers of the at least two phases. The modulation state, in particular, establishes a switching sequence for the circuit breakers. Space vector modulation has proven especially effective, whereby a target space vector is simulated in each switching cycle. The switching can, for example, be symmetrical around a midpoint of the switching cycle.

[0012] The switching of the circuit breakers can be carried out, in particular, with fundamental frequency switching, where the switching frequency corresponds to the fundamental frequency of the load current. It is especially preferred that the switching frequency differs from the fundamental frequency of the generated alternating current, and is particularly preferably significantly higher than the fundamental frequency of the load current. Switching preferably takes place at a high frequency, for example, in the kilohertz range. At high frequencies, the load current can be considered approximately constant in the respective phase via a switching frequency.

[0013] The process of relieving the switching load on individual circuit breakers by activating their respective load relief switches is a known concept. This relief can be achieved, in particular, by zero-voltage switching of the circuit breakers. Activation of the load relief switch means that it is closed to provide a current path from the output terminal via the inductor. Activating the load relief switches is particularly effective in reducing switching losses during switching operations where the load current of the respective phase has the same polarity as the potential change due to the switching. For example, if the load current is positive and the output terminal is switched from the negative potential of the input DC voltage to the positive potential of the input DC voltage, this leads to high switching losses without load relief.If, however, the load current has the opposite sign to the potential difference, it can cause the circuit breakers to recharge, thus reducing switching losses and eliminating the need to activate the corresponding load relief switch. If the opposite load current is small in magnitude, however, activating the load relief switch can still be advantageous even for such switching edges.

[0014] The load balancing network is typically arranged between the output terminals and the input terminals, particularly between a center tap for the DC input voltage and the output terminals. The common inductance of several phases can be directly connected to the center tap. The center tap is typically connected to a DC link, which supplies the input DC voltage to the two-level inverter. For example, the center tap is located between two DC link capacitors of a DC link.

[0015] The components of the bridge circuit and the load balancing network, in particular the power switches and load balancing switches, respectively, can be designed in a known manner. Suitable power switches are, in particular, switchable in the forward direction and conducting in the reverse direction. The forward direction is understood to be the conventional current direction with respect to the polarity of the input DC voltage. Suitable power switches can, in particular, comprise semiconductor switching elements, for example, MOSFETs, IGBTs, and / or bipolar transistors. A diode can be connected antiparallel to these for the conducting connection in the reverse direction. The power switches have a parasitic capacitance. To increase the capacitance, the power switches can also have capacitors connected in parallel.

[0016] The bidirectional load shedding switches can, for example, consist of two series-arranged, oppositely oriented semiconductor switching elements, each with an antiparallel diode. This allows the bidirectional switch to be implemented using simple semiconductor switching elements.

[0017] The switching points of the circuit breakers of two phases are shifted relative to each other such that the respective load-relieving switches are activated with a time offset. In particular, switching points that do not comply with the minimum time interval can be shifted relative to each other so that the switching points comply with the minimum time interval. Preferably, the relative shift is applied to both switching edges of the respective phase. In this way, the average voltage applied over the switching period is not changed. The shift does not affect the implementation of the specifications by the modulation method, in particular the target space vector.

[0018] For example, the switching points of the switching edges of one phase can be shifted relative to the potentially colliding phase. This results in the switching of the circuit breakers being offset from the midpoint of the switching period while maintaining a constant average voltage.

[0019] In some cases, shifting the switching times can lead to a change in the sequence of switching operations. Before implementing the shift, it is advantageous to check whether the control system can handle the shifted switching times.

[0020] Shifting the switching points implements a further collision avoidance measure, enabling control using modulation states, particularly target space pointers, which in themselves do not adhere to the minimum time interval. This collision avoidance measure expands the range of usable target space pointers without introducing collisions due to the simplified relief network.

[0021] The relative shift of the phase switching times is averaged over several switching periods. This allows, in particular, the simulation of average switching times that correspond to the target switching times specified by the target space vector. Advantageously, the target characteristics of the output voltages can be reliably maintained despite a shift in the switching times.

[0022] For example, it may be possible to shift the switching edges of different phases differently relative to each other in different switching periods. For example, in successive switching periods, the switching edges of two phases can be shifted alternately relative to each other.

[0023] According to a preferred aspect of the method, only those modulation states are considered during the execution of the switching periods in which the switching times of the circuit breakers of the at least two phases preclude simultaneous activation of the load shedding circuit breakers of different phases, while the at least two phases are at different potentials of the input DC voltage. In particular, only those modulation states can be considered in which the switching of the circuit breakers of different phases occurs essentially simultaneously, while maintaining a minimum time interval and / or with the same sign of the potential change. This enables collision-free switching by selecting suitable modulation methods, in particular suitable target space vectors, for controlling the circuit breakers.During control, the modulation states, especially target space vectors, can be analyzed to determine whether their switching times could cause a collision. If necessary, the modulation states can be rejected as impermissible. To achieve the desired load current, a sequence of other permissible modulation states can then be selected.

[0024] Simultaneous activation of the load-sensing circuit breakers for different phases, while these phases are at different potentials, is impossible if the switching times of the phase breakers coincide when the potential change is of the same sign. Therefore, switching operations that occur essentially simultaneously when the potential change is of the same sign are unproblematic. Modulation states whose switching times are spaced so far apart that simultaneous activation of the respective load-sensing circuit breakers is impossible are also unproblematic. This is ensured, for example, if the switching of the circuit breakers of different phases occurs with a minimum time interval that is greater than or equal to the sum of half the activation times during which the respective load-sensing circuit breakers of the phases are activated to relieve the switching of the respective circuit breakers.A collision-avoiding modulation method can be limited in particular to modulation states in which switching of the power switches of different phases occurs either with the minimum time interval and / or, if the potential change has the same sign, essentially simultaneously.

[0025] The activation time of the load shedding circuit breaker is determined primarily by the sum of the time required to build up and then release a suitable discharge current in the inductor, and the charging time during which the circuit breakers are recharged using this discharge current. The discharge current corresponds specifically to the load current (measured in the direction of the potential change) plus a commutation current, which ultimately recharges the circuit breakers. The current build-up and drop-down times are therefore proportional to the discharge current and the inductance. The charging time depends on the capacitance of the circuit breakers, the input DC voltage, and is inversely proportional to the commutation current. Given the load current, the components of the two-level inverter, and a specified commutation current, the activation time can be determined.A suitable commutation current can be determined, for example, by specifying a desired recharging time based on the capacity of the power switches and the input DC voltage.

[0026] Since the switching operations typically only occupy a small fraction of the period of each switching cycle, the minimum time interval to be maintained is also small compared to the period itself. Therefore, excluding switching states that do not comply with the minimum time interval only leads to a minor restriction of the possible modulation states, particularly the available target space vectors.

[0027] According to a preferred aspect of the method, a minimum time interval exists between the switching of circuit breakers of different phases. This enables a simple and reliable method of preventing simultaneous activation of the load-sensing circuit breakers, regardless of the potential at which their respective output terminals are located. This constitutes a collision avoidance measure based on the modulation states used for controlling the circuit breakers, whereby only those states are used that implement a time interval of the switching of circuit breakers of different phases by the minimum time interval. Switching operations that could potentially lead to collisions are not considered during the control process. This can be achieved, for example, by specifying fixed switching patterns or by excluding certain spatial vector ranges in which the minimum time interval is not met.

[0028] For example, the minimum time interval is greater than or equal to the maximum activation time, which results when the load current is at its maximum magnitude, i.e., corresponding to the maximum permissible current of the phases of the two-phase inverter. Specifying a minimum time interval at least equal to the maximum activation time ensures that collisions are reliably avoided, regardless of the respective load currents and modulation states. The corresponding minimum time interval provides a safety buffer that prevents potential collisions from occurring in the first place.

[0029] According to a preferred aspect of the method, the potential applied to the output terminals of at least two phases is reversed essentially simultaneously, provided the potential change has the same sign. Short circuits due to the simultaneous activation of the load-sensing switches of different phases only occur if the output terminals are at different potentials. By switching the two phases essentially simultaneously with the same sign of the potential change, the presence of different potentials at these phases is avoided. Simultaneous activation of the load-sensing switches does not result in a collision.

[0030] The switching of circuit breakers for two phases occurs essentially simultaneously, particularly when the time interval between the switching points is on the order of the duration of typical switching operations. This essentially simultaneous switching can be achieved, for example, by switching the relevant circuit breakers simultaneously, within the accuracy limits of the control system.

[0031] Particularly preferred is the essentially simultaneous switching of the circuit breakers of two phases, achieved by shifting their switching times relative to each other to avoid a time interval that could lead to a collision, i.e., one that does not comply with the minimum time interval. To ensure the average voltages per switching period, the shift is preferably made for both switching edges of the respective phase. For example, the switching edges of both phases can be shifted so that one of the switching times coincides with one of the switching times of another phase.

[0032] The essentially simultaneous switching of the circuit breakers of two phases is particularly relevant when the load currents of both phases have the same sign, especially when the sign of the two load currents is the same as the sign of the respective potential change. In such cases, the same load relief network can support both phases simultaneously. However, this also means that the load relief current that builds up in the inductor must correspond to the sum of the two load currents and twice the commutation current. This, in particular, increases the current flowing through the inductor.

[0033] A particularly preferred aspect of the method is to check whether the discharge current flowing through the inductor exceeds a permissible maximum current of the inductor. This prevents damage to the inductor. If there is a risk of exceeding the maximum current of the inductor, the corresponding collision avoidance measure can be discarded and another collision avoidance measure selected. Alternatively or additionally, an inductor with a correspondingly high maximum current can be selected, in particular with a maximum current greater than or equal to twice the maximum load current plus twice the commutation current.

[0034] By essentially switching the circuit breakers of two phases simultaneously, in particular by relative shifting of the switching times, a collision avoidance measure is implemented which allows control even with space vectors that do not in themselves comply with the minimum time interval.

[0035] According to a preferred aspect of the method, the load currents of the two phases to be reversed essentially simultaneously have opposite signs. When the load relief networks are activated simultaneously, the load currents with opposite signs are coupled, resulting in a smaller differential load current. The load relief current to be generated therefore only needs to correspond to the sum of the differential load current and the commutation current. The load relief current flowing through the inductance of the load relief network is reduced. This reduces the activation time required for switching load relief. Losses due to the load relief current are minimized.

[0036] The relief switch is particularly often activated for a phase that wouldn't actually require switching relief. By additionally activating the relief switch, a reduced relief current can be achieved due to the resulting differential load current, with the advantages mentioned above.

[0037] According to a preferred aspect of the method, the load-sensing circuit breaker of one phase is not activated to relieve the switching of the corresponding circuit breakers as long as the load-sensing circuit breaker of another phase is activated. Specifically, the switching of the circuit breakers of one phase occurs without activating the respective load-sensing circuit breaker while the load-sensing circuit breaker of another phase is active. This allows the switching of the circuit breakers to occur independently of any time interval between switching points and without risk of collisions. For the phase where the load-sensing circuit breaker is not activated, the switching of the circuit breakers is carried out, in particular, by so-called hard switching, i.e., lossy switching. To avoid collisions, individual hard switching operations and thus minor switching losses can be accepted.

[0038] According to a preferred aspect of the method, the phase whose respective load relief switch is not activated carries a smaller load current than the phase whose load relief switch is activated, provided that the load relief switch of another phase is activated. In particular, the phase whose switching occurs without activating its respective load relief switch carries a smaller load current than the phase whose load relief switch is activated. In this way, switching losses due to occasional hard switching can be further reduced.

[0039] Hard switching of one of the phases is a collision avoidance measure that is applicable regardless of the dimensioning of the components of the switching arrangement and regardless of the choice of a modulation state, in particular a target space vector.

[0040] According to a preferred aspect of the method, the relief switches are activated sequentially. A relief switch that is initially not activated to avoid collisions is activated sequentially after a previously activated relief switch of another, potentially colliding phase. In cases with insufficient time intervals between switching points, this leads to incomplete switching relief operations. Sequential activation reduces the overall switching losses in the circuit breakers and relief switches.

[0041] For example, it is possible to complete the relief process for one of the phases, i.e., to wait for the relief current in the inductor to dissipate. After the relief current in this phase has dissipated and the corresponding relief switch has closed, the relief switch for the other phase can be activated. Due to a small time interval between the switching points of the two phases, the sequential activation of the relief switch may not occur in time to fully build up the relief current required for switching relief. Zero-voltage switching may therefore not be possible. Advantageously, however, at least a portion of the required relief current can build up in the relief network, so that the switching of the circuit breakers is at least supported by this partial relief current.The switching process does not need to occur for the entire load current, but only for the difference between the load current and the proportional relief current. This reduces switching losses compared to hard switching of the circuit breakers.

[0042] It is also possible to perform a hard switch on the load relief circuit breakers of the two phases. This means that the load relief circuit breaker of the first phase is deactivated before the load relief current has completely dissipated. A hard switch on the load relief circuit breaker of the other phase then takes over the incomplete load relief current. This is particularly advantageous when the load currents of the respective phases have the same polarity. The load relief current remaining after the hard switch can then also be used to relieve the charging process for the other phase. Hard switching of the load relief circuit breakers results in lower switching losses, as they are typically designed to be smaller than circuit breakers.

[0043] The collision avoidance measures described above are each suitable and sufficient on their own to ensure the safe operation of the two-point inverter. Particularly preferable are combinations of the collision avoidance measures provided in the methods to allow the selection of one or more of the collision avoidance measures depending on the application.

[0044] A switching arrangement for a two-phase inverter with at least two phases comprises input terminals for the two poles of an input DC voltage, one output terminal per phase, and a bridge circuit with one half-bridge per phase. Each half-bridge includes a power switch through which the respective output terminal is connected to the input terminals. The switching arrangement includes a load shedding network consisting of one bidirectional load shedding switch per phase and exactly one inductor for the at least two phases, which is connected to the at least two output terminals via the respective bidirectional load shedding switch. The switching arrangement features a simpler load shedding network geometry, in particular a minimal number of passive components. This reduces the cost, weight, and volume of the multi-phase two-phase inverter.The switching arrangement may in particular have one or more of the features that have been discussed above with regard to the switching arrangement.

[0045] According to a preferred aspect of the circuit arrangement, the inductor of the load balancing network is an air-core inductor. Air-core inductors enable high-frequency charging dynamics, allowing for particularly fast charging processes. This is especially advantageous because the inductor must carry a load balancing current for at least two phases, and in particular for all phases. Furthermore, air-core inductors exhibit a high maximum current, making them particularly well-suited for the collision avoidance strategies discussed above. By eliminating magnetic core materials, magnetic core losses and saturation effects are avoided.

[0046] The multiphase two-point inverter according to the invention comprises the switching arrangement described above and a control unit, the control unit being configured to carry out the control method according to the invention. The inverter offers the advantages discussed with regard to the method and the switching arrangement.

[0047] Preferably, the control unit is configured to implement the advantageous aspects of the method discussed above. For example, the control unit can be configured to calculate different collision avoidance measures and to select a suitable collision avoidance measure for the respective application, in particular for the respective switching period.

[0048] Further advantages and details of the invention are explained in more detail below with reference to the drawings. These show: Fig. 1 a schematic representation of a multi-phase two-point inverter, Fig. 2 a circuit diagram of a switching arrangement of the inverter according to Fig. 1, Fig. 3. A space vector diagram to illustrate possible switching states for controlling the inverter. Fig. 4. A plot of the potentials applied to the output terminals of the inverter and of a current flowing through an inductor of a relief network over the course of an exemplary switching period, wherein relief of switching processes takes place at individual switching edges. Fig. 5 a Fig. 4. Corresponding representation of an exemplary switching period, in which a collision occurs during the activation of the relief network for two switching edges, Fig. 6 a Fig. 5. Corresponding representation of a switching period with a collision avoidance measure to prevent the collision of two switching edges, wherein the switching times of the circuit breakers are shifted relative to each other in such a way that no simultaneous activation of the relief network occurs. Fig. 7 a schematic procedure for implementing the collision avoidance measure according to Fig. 6, Fig. 8 a Fig. 5. Corresponding representation of a switching period with a further collision avoidance measure to avoid a collision of the switching edges, wherein the switching times of the switching edges are shifted in such a way that simultaneous switching takes place. Fig. 9 a schematic procedure for implementing the collision avoidance measure according to Fig. 8, Fig. 10 one Fig. 5. Corresponding representation of a switching period with a further collision avoidance measure, whereby activation of the relief network for one of the switching edges is omitted. Fig. 11 a schematic procedure for implementing the collision avoidance measure according to Fig. 10 Fig. 12 one Fig. 5. Corresponding representation of a switching period with a variant of the collision avoidance measure according to Fig. 10, Fig. 13 one Fig. 5. Corresponding representation of a switching period with a further variant of the collision avoidance measure according to Fig. 10, and Fig. 14 one Fig. 5 corresponding representation of a switching period, wherein an additional activation of the relief network takes place for a switching edge in order to reduce a current flow in an inductance of the relief network.

[0049] In Fig. Figure 1 schematically shows a multiphase two-phase inverter 1. The two-phase inverter 1 has a switching arrangement 2 and a control unit 3. The switching arrangement 2 has a DC voltage input 4 and a multiphase AC output 5. The DC voltage input 4 is connected to the AC output 5 via a bridge circuit 6. The bridge circuit 6 is designed as a bridge inverter. A load shedding network 7 is connected between the DC voltage input 4 and the bridge circuit 6. The control unit 3 controls the half-bridges of the bridge circuit 6 and the load shedding network 7, as described below.

[0050] In Fig. Figure 2 shows a circuit diagram of the switching arrangement 2 of the two-point inverter 1. The DC voltage input 4 has two input terminals 8 for the two poles +, - of an input DC voltage U. ZK on. The input DC voltage U ZKThe DC voltage is provided via a DC link comprising two DC link capacitors 9 and is also referred to as the DC link voltage. A center tap 10 is located between the two DC link capacitors 9 for the midpoint of the input DC voltage U. ZK formed. The central tap 10 is located at a medium potential ϕ. M .

[0051] The two-point inverter shown has three output phases P, where P = 1, 2, 3 denotes the respective phase. Each phase P has one output terminal A. P available. The output connections A P Together they form the multi-phase AC output 5. To the output terminals A P An exemplary symmetrical inductive load 11 is connected. The two-point inverter 1 is generally suitable for any load, including capacitive and / or asymmetrical loads. The output terminals A PThe load 11 is assigned to the respective phases P and the load currents i P provided.

[0052] The bridge circuit 6 has one half-bridge per phase P, via which the respective output terminals A are connected. P are connected to input terminals 8. The half-bridges each have a circuit breaker S. P V on, where V = + , - denotes the pole of the DC voltage input 4, with which the output terminal A P via the circuit breaker S P V is connected. The circuit breakers of a phase P are also collectively referred to by the reference symbol S. P designated via the circuit breakers S P V can the output terminals A P The output terminals can be selectively connected to the respective poles + and - of the DC input 4. This allows the potential at which the output terminal A is set. PPhase P lies between the potentials of the input DC voltage U ZK Reverse polarity. The potential of the respective phase P is defined by ϕ. P This is designated, for example, as circuit breaker S1. + closed and the circuit breaker S1 - When open, the output terminal A1 is at the potential of the positive terminal of the input DC voltage U. ZK : ϕ1 - ϕ M = U ZK / 2. By switching the switches S1, the potential ϕ1 can be reversed to: ϕ1- ϕ M = - U ZK / 2 .

[0053] The circuit breakers S P V They are all identical in design. They enable switching, i.e., locking and unlocking, in the forward direction, i.e., in the conventional current direction. In the opposite direction, the circuit breakers S P Vconductive. The power switches have a switching element 12 and an antiparallel diode 13 for this purpose. The switching element 12 is a semiconductor switching element, for example a MOSFET. Other, comparable switching elements are also possible, for example IGBTs and / or bipolar transistors. The power switches S P V exhibit a medium capacity C s The capacitance can be a parasitic capacitance of the components, especially the switching element 12. To increase the capacitance C s In the illustrated embodiment, a capacitor 14 is connected in parallel to the switching element 12 and the diode 13.

[0054] Relief network 7 is located between output ports A P and the central tap 10. The relief network 7 consists of a bidirectional relief switch B. P each phase P and a common inductance L, which is connected via the respective load-sensing switches BP with the corresponding output terminal A P is connected. The load balancing network 7 thus has a common inductance L for all phases P. The inductance L is between the center tap 10 and the output terminals A. P switched on.

[0055] The inductor L is an air-core coil. This allows for high current flows and rapid charging and discharging processes. The current flow in the coil is subsequently referred to as the discharge current i. L designated.

[0056] The bidirectional relief switches B PIn the illustrated embodiment, two switching elements 15 are connected in series, each with an antiparallel diode 16. The series-connected switching elements 15 and the antiparallel-connected diodes 16 are oriented in opposite directions. The switching elements 15 are semiconductor switching elements, for example, MOSFETs, IGBTs, or bipolar transistors. In comparison to the power switches S P is the average current load of the relief switches B P low, so that the relief switches B P They only need to have a correspondingly lower current carrying capacity.

[0057] The following refers to the Fig. 3 and Fig. 4. The function of the two-point inverter 1, in particular its load balancing network 7, is described by way of example. For generating the output alternating currents i P High-frequency switching of the circuit breakers S takes place. P , where the potential ϕ PThe phases P are reversed. The switching frequency f s is many times higher than the fundamental frequency of the output current i P This allows for a wide range of output current and output voltage. The two-point inverter 1 enables, in particular, bidirectional energy flow. The corresponding operating principle, especially the switching of the power switches S, P is generally known and will only be roughly outlined below.

[0058] The switching operations of the two-point inverter 1 are in switching periods with period length T. s subdivided. This is necessary to achieve a desired output current i P The modulation methods required for phase P can be represented, for example, in a space vector diagram, as shown in Fig. Figure 3 shows how the phases P are represented. In the space vector diagram, the phases P are represented by the letters R, S, and T on axes offset by 120° from each other. In each switching period, a target space vector u is defined. By projecting the target space vector u onto the axes of the respective phases, an average voltage assigned to the respective phase P in the switching period can be determined. From switching period to switching period, the target space vector u can be varied so that it traces a curve within the space vector diagram. A dashed circle 17 in Fig. Figure 3 shows a typical operating range of space vector modulation with constant mean neutral point voltage in common two-point inverters.

[0059] In Fig. Figure 4 shows an example of switching behavior for one switching period. The time course of the respective potentials ϕ is shown. P - ϕ M the phases as well as the relief current iL The signal is plotted over a switching period. Based on the corresponding target space vector u, target voltages P for the individual phases are determined. To achieve the target voltages, the polarity of the respective phase P is reversed from the potential -U. ZK / 2 of the negative terminal of the input DC voltage U ZK to the potential U ZK / 2 of the positive terminal and back. Depending on the time interval T P In a switching cycle where the polarity of the respective phase P is reversed, the average voltage of phase P during the switching period corresponds to the target voltage. The switching normally occurs symmetrically around a center point T. M of the respective switching period. Therefore, a double polarity reversal occurs per switching period, namely from the negative potential to the positive potential of the input DC voltage U. ZKon a first switching edge and vice versa on a second switching edge. The switching time of the first switching edge is denoted by t in the figures. P , the second switching edge tP' The time interval between the switching edges, in which the respective phase P is reversed, is denoted by T. P designated.

[0060] Due to finite switching times, switching losses can occur at the switching edges. These losses can be mitigated by adding the load balancing network 7 by activating the corresponding bidirectional load balancing switch B. P to be reduced. Whether relief using the relief network 7 is necessary depends in particular on the direction of the load flow i flowing from the respective half-bridge. Pdependent. As a rule, activation of the relief network 7 is only necessary during switching operations where the sign of the potential change at output terminal A is different. P and the sign of the respective load current i P are the same. In the case of opposing currents i P A recharging of the circuit breakers S takes place P using the respective load current. In Fig. 4 are shown in a diagram for the current i flowing through the inductance L L additionally the load flows i P shown. Due to the high switching frequencies of the switching periods, the load current i P essentially constant within a switching period. From a comparison of the sign of the load current i P and a potential change occurring at the respective switching edge can determine which of the switching edges activates the respective relief switch B. Prequire. The corresponding switching edges are in Fig. 4 marked with an *.

[0061] For the switching edges marked with *, the load current i P the capacities C s the respective circuit breaker S P not, so that switching losses would occur. To reduce these switching operations, the circuit breaker S is switched before switching. P the corresponding bidirectional relief switch B P of the relief network is activated, so that the bidirectional relief switch B P and the inductance L a path for the respective load current i P This results in the following: The discharge current i builds up in the inductor L. L up. As soon as the relief flow i L the load current i P to ensure a sufficient commutation current ΔI ZVS If the commutation current ΔI exceeds the specified value, it will be increased by the following values: ZVS to a recharging of the circuit breakers SP , so that they can switch off without voltage. The relief current then builds up. L off again, whereupon the bidirectional relief switch B P can be closed.

[0062] In Fig. Figure 4 is an enlarged representation of the time course of the relief flow i L Included. After closing the bidirectional relief switch B. P The relief flow builds up i L over a period of time t AP up. After the relief flow i L the load current i P to determine the commutation current ΔI specified for recharging ZVS If the threshold is exceeded, the respective circuit breakers S are switched off. P During a commutation period t K A transfer of the circuit breakers S takes place P for voltage-free switching. The relief current then builds up. L again over the period t AP away.

[0063] The time for the current rise or fall depends on the load current i P and the magnitude of the commutation current ΔI ZVS The time for the increase or decrease in current is calculated as follows: tAP(iP)=|iP|+ΔIZVSUZK2⋅L.

[0064] The commutation time can be estimated as follows: tK≈2⋅∫0UZKCS(uS)duΔIZVS≈2⋅CSUZKΔIZVS.

[0065] Overall, this results in an activation time T. AP , in which of the corresponding bidirectional relief switches B P to relieve the switching of the circuit breakers S P A phase P is activated to: TAP(iP)=2⋅tAP(Imax).

[0066] The activation time T AP Given the design of the two-point inverter, in particular the inductance L and the capacitance C s , as well as for a given input DC voltage U ZKessentially dependent on the selectable commutation current ΔI ZVS and the respective load current i P The activation time T AP is at maximum with maximum load current I max: TAmax=TAP(Imax).

[0067] It can also be set to a desired maximum activation time. TAmax=TAP(Imax). be specified and the design of the two-point inverter 1 be adapted accordingly.

[0068] The following are example values ​​for a two-level inverter. The input DC voltage is U. ZK = 800 V. The maximum phase current is I max Assume a current of 900 A. The switching frequency is f s = 10 kHz, which corresponds to a period of T s = 100 µs. Furthermore, the following specifications are given: The commutation time t K , which describes the duration of the switching process, let t K= 200 ns, which corresponds to a slow switching speed of 4 V / ns. The average capacitance of a half-bridge with two power switches S P let 2 · C s = 4 nF. The maximum permissible activation time becomes TAmax=TAP(Imax)=2 μs The value in µs is set. This determines the maximum time duration for the current rise and fall at t. AP = 900 ns. The required commutation current for achieving voltage-free switching is therefore: ΔIZVS=2⋅CS⋅UZKtK=16 A.

[0069] The inductance must be chosen accordingly: L=UZK2⋅tAPImax+ΔIZVS=393 nH.

[0070] The numerical values ​​mentioned above are purely exemplary and not to be understood as restrictive. With the selected parameters, it becomes apparent that the maximum activation time T A (I max ) only a few percent of the period T sIn the specific embodiment shown, this amounts to 2%.

[0071] At the in Fig. In example 4, the relief network 7 is only activated for switching edges where the load current i P has the same sign as the potential difference of the respective switching process. With opposite load current i P A transfer of load currents takes place via the load current i P , so that the relief flow i L This is not required for this purpose. For small load currents i P However, can the load current i P the recharging of the circuit breakers S P This does not always have a sufficient effect. Therefore, with small load currents, it is advantageous to also consider the other switching edge, in which the load current i PThe potential difference has a different sign and can be relieved using the relief network 7. In principle, therefore, all switching edges can be relieved by activating the relief network.

[0072] The basic procedure for relieving the switching processes is known. The special feature of the two-point inverter 1 arises from the fact that the use of a single inductor L for all phases P leads to a coupling of the current paths of the relieving network 7 for different phases P. This is unproblematic as long as there is no change between the activation of different bidirectional relieving switches B. P , i.e., an activation of the relief network 7 for switching the circuit breakers S P different phases P have a finite time interval t*, as is the case in Fig. As indicated in section 4, there is a temporal overlap or the time interval between different activation times T. AP The values ​​of different phases P and P' can be calculated as follows: t*=|tP−tP'|−tAP−tAP'−tK, where t* > 0 describes a time interval and t* < 0 describes a time overlap.

[0073] It was recognized that there was a temporal overlap in the activation of the relief switches B. P This can lead to a short circuit for different phases P if the respective phases P are activated simultaneously by the relief switches B. P at different potentials of the input DC voltage U ZK lie. Such a case is also called a collision. In Fig. Figure 5 shows an example collision. Here, the switching times t1 and t2 of phases 1 and 2, respectively, are so close to each other that the activation of the respective bidirectional load shedding switches B1 and B2 overlap in time (t* < 0). During the time overlap, phase 1 is already at the potential of the positive terminal of the input DC voltage U. ZK , while phase 2 is still at the potential of the negative pole.

[0074] The time interval t* can be determined by the control unit 3 in each switching period for the respective modulation state. This allows it to be determined for each switching period, particularly in real time, whether the modulation state potentially poses a collision risk.

[0075] The following describes control methods that prevent collisions. These control methods enable the simplified design of the load balancing network 7 with a single inductor L. This allows for a simple, cost-effective, and space- and weight-saving design of the two-point inverter 1. The corresponding control methods are implemented by the control unit 3.

[0076] A first collision avoidance measure is implemented by ensuring that the modulation method only considers modulation states in which simultaneous activation of the relief switches B is possible. P Different phases P are excluded, while the respective phases P are at different potentials of the input DC voltage U. ZKModulation states, in particular target space vectors u, that do not meet this condition are rejected as inadmissible. Control is then carried out using suitable, admissible modulation states, in particular target space vectors u.

[0077] For example, it is possible to use pre-calculated modulation states that do not cause collisions. Optimized switching patterns can be calculated, stored, and played back for this purpose. The well-known concept of synchronous clocking offers an exemplary method for calculating switching patterns.

[0078] Additionally or alternatively, it is also possible to check the suitability of the modulation states of the modulation method, particularly for potential collisions. This can be done dynamically, for example, by the respective control method. Based on this check, the modulation states can be assessed as permissible or impermissible. For this check, the respective time interval between adjacent switching points of different phases can be evaluated, for example. Additionally or alternatively, a static criterion can be defined against which the permissibility of the modulation state can be checked. This static criterion could, in particular, be a fixed minimum time interval.

[0079] The first collision avoidance measure is exemplified using the space vector diagram in Fig. 3 explained. Here, the modulation method, i.e., the control of the power switches, is selected in such a way that potential collisions are avoided. Target space vectors u, which would lead to collisions, are excluded in space vector modulation. These are all modulation states in which the switching edges of different phases P are so close together that a collision can occur. t*<0 or |tP−tP'|<Δt=tAP+tAP'+tK can be.

[0080] Using the example of space vector modulation, areas of the space vector diagram, i.e., the range of possible target space vectors u, in which collisions can occur, are rejected as impermissible. For this purpose, the control unit 3 can determine the time interval t* for each switching period based on the respective transient load currents i. P dynamically determine whether a potential collision exists and the target space pointer is rejected as inadmissible.

[0081] According to one variant of the first collision avoidance measure, a static minimum time interval can be selected that reliably prevents collisions. The minimum time interval is preferably chosen such that it reliably excludes collisions for all potentially occurring load currents. This reduces the computational effort for control, but may exclude individual modulation states that are otherwise permissible because they do not cause collisions. The minimum time interval can be determined, in particular, based on a worst-case scenario in which switching relief occurs for maximum load currents, i.e., with maximum activation time. TAmax, is assumed.

[0082] To determine a suitable exemplary minimum time interval, it is assumed below that the relief network 7 is activated for each switching edge. Furthermore, it is assumed that for each activation of the relief network 7, it remains active for the maximum activation time. TAmax is activated. To avoid collisions, the switching of the circuit breakers S P Therefore, different phases P must maintain a minimum time interval Δt that is greater than or equal to the maximum activation time. TAmax is: Δt≥TAmax=TAP(Imax).

[0083] Target space vectors u that do not meet this condition have components of two phases that are essentially the same length. Corresponding target space vectors u that do not meet the minimum time interval Δt lie along the phase axes R, S, T or in the outer regions of the space vector diagram. These regions are in Fig. 3 are shown vertically hatched and marked with reference symbol 18. Areas of the space vector diagram lying outside area 18 allow the operation of the two-point inverter 1 with activation of the relief network 7 for all switching operations without any risk of collisions. It is therefore evident that collisions only occur in rare cases. In this regard, it should be noted that, for the sake of clarity, Fig. 3. A disproportionately large representation of the space vector regions that do not meet the minimum time interval Δt was chosen. As explained above in the case of the exemplary design of a two-point inverter, only small fractions of the period T are lost. s on possible activations of the relief network 7. This allows large areas of the space vector diagram to be used for control without fear of collisions.

[0084] It is known that overdriving occurs in the outer regions of the space vector diagram, whereby a constant mean neutral point voltage is not guaranteed. By using the load balancing network 7, the region in which overdriving occurs shifts slightly compared to known two-point inverters. The overdriving regions are described in Fig. Area 3 is hatched with curved lines and marked with reference numeral 19. In these areas, collision-free operation of the two-point inverter 1 is possible if the condition of constant mean neutral point voltages is waived. In the unhatched areas marked with reference numeral 20, collision-free control is guaranteed with a constant mean neutral point voltage.

[0085] The previously described variant of the first collision avoidance measure is based on the fact that target space pointers u, which do not meet the minimum time interval, are excluded from the control of the circuit breakers S. PCollisions are excluded. This collision avoidance measure affects the modulation method itself. Nevertheless, essentially normal operation of the two-point inverter is possible. For example, when traversing a curve of target space vectors, range 18 can be skipped. It is also possible to simulate an average space vector that lies within range 18 by changing the space vectors on average over several switching periods, without risk of collisions. In some cases, skipping a space vector range or an excessively large gap between two target space vectors used in successive switching periods can be detrimental, for example, with regard to excessively large subharmonics in the load current. In such a case, further collision avoidance measures can be applied that enable collision-free operation even for space vectors within range 18.

[0086] With regard to the Fig. 6 and Fig. Section 7 describes a second collision avoidance measure that prevents conflicts due to a temporal overlap of the activation times T. AP of relief switches B P different phases P are avoided. Fig. 6 shows a Fig. 5. Corresponding collision case, in which the switching times t1 and t2 of two switching edges, which require relief for soft switching, are close together in time. Without collision avoidance measures, the activation times T AP the respective bidirectional relief switch B P overlap.

[0087] To avoid collisions, the switching times t1 and t2 are shifted relative to each other, resulting in a finite time interval t* > 0 between the activation times T. APThis is ensured. In the present case, the switching time t2 of phase 2 is shifted, as indicated by arrows 25. To avoid changing the average voltage over the switching period and thus the target space vector u, the switching time is also t2' The switching is shifted accordingly to the second switching edge. The switching therefore no longer occurs symmetrically around the time midpoint T. M However, this does not typically affect the average voltage over the switching period. To avoid impacts on the target output voltage characteristics if a shift over several switching periods is necessary, the shift in switching times can be averaged over these periods.

[0088] The magnitude of the shift results from a calculation of the temporal overlap of the respective activation times T. AP the relief switch B1 or B2: t*=|t1−t2|−tA1−tA2−tK<0.

[0089] From the determined temporal overlap t* < 0, the size of the necessary relative shift of the switching times can then be determined.

[0090] In Fig. Figure 7 shows a schematic procedure for implementing the second collision avoidance measure. Step S1 marks the start of the calculation for one switching period. This is followed by the modulation procedure M, in which step S2 specifies the respective space vector and thus the target values ​​of the average output voltages of the phases P. In step S3, the switching times t are calculated. P , tP' the switching edges based on the target voltages.

[0091] In step S4, the respective load currents i P determined, in particular measured. Based on the load currents i PIn step S5, the switching edges are determined at which the respective relief switches B P to be activated. In step S6, the relief flows required for relief are determined. L calculated. With knowledge of the relief flows i L The activation times T can be determined. AP Determine S7 for the relief network 7 in one step.

[0092] In step S8, a check is performed to see if there are any overlaps in the activation times T. AP and thus collisions exist. If this is not the case, there is no collision problem or a potential collision problem has been resolved (step S9). In step S10, the determination of the control signals is completed, which are then passed on to the driver.

[0093] If a collision is detected in step S8, the overlap period t* for the corresponding switching edges is determined in step S11. Based on the overlap period t*, the required relative shift of the switching times t is calculated in step S12. P , tP' the affected switching edges are calculated, whereby the other switching edges of the affected phases are shifted accordingly.

[0094] In step S13, it is checked whether the implementation of the control system allows for the shifted switching times t. P , tP' can implement this. If this is not the case, the calculation ends in step S14 and another measure for collision avoidance is checked.

[0095] If the check in step S13 reveals no problems with implementation, step S7 is used to recalculate the activation times T. APbased on the shifted switching times t P calculated. A further check is then performed in step S8 to see if the changed activation times T AP No collisions occur in any phase.

[0096] With regard to the Fig. 8 and Fig. Section 9 describes a third collision avoidance measure. Fig. 8 shows a Fig. 5. A corresponding situation in which the switching times t1 and t2 are so close together that the respective load relief switches B1 and B2 are activated simultaneously. However, the simultaneous activation of the load relief network 7 only leads to short circuits if the corresponding output terminals A P , in this case A1 and A2, to a different potential of the input DC voltage U ZK lie. This is the case with the one in Fig. Example 8 shows the case between switching times t1 and t2. Different potentials of the output terminals A P Different phases P are avoided if switching operations with the same sign of the potential change occur essentially simultaneously. If this is not already the case due to the switching processes required for the simulation of the target space vector u, essentially simultaneous switching of the otherwise colliding phases P, in this case phases 1 and 2, can be achieved by setting the switching times t P are shifted relative to each other in such a way that they coincide.

[0097] At the in Fig. In the embodiment shown in Figure 8, the switching time t2 of the second phase is advanced such that it coincides with the switching time t1 of the first phase. The resulting simultaneous reversal of the polarity of the output terminals A1 and A2 prevents a potential difference at the output terminals and thus a short circuit due to the simultaneously activated load shedding switches B1 and B2.

[0098] To avoid a change in the average target voltages, the other switching edge, in this exemplary embodiment the switching point, is used. t2', The changes are as follows: Fig. 8 marked with arrows 26.

[0099] The relief flow i L , which is used in the inductor for recharging the power switches S P When the simultaneously switching phases P flow, the load currents i must be considered. P Equalize both phases P. The relief current iL is therefore correspondingly increased. This leads to an increased current load on the inductor L. The simultaneous activation of the relief networks and the resulting relief current i L are in Fig. 8 shown as an example with reference number 33.

[0100] In Fig. Figure 9 shows a schematic procedure for implementing the third collision avoidance method. The procedure steps S1 to S7, in particular the calculation of the modulation method M and the activation times T, are shown. AP of the relief network 7, correspond to the procedural steps S1 to S7 of the in Fig. 7 shown procedure sequence.

[0101] In process step S20, it is checked whether the circuit breakers S are suitable for recharging. P When the relief network is activated, the required currents overload the inductance L, i.e., whether the sum of the relief currents i Lgreater than a maximum current I L,max The inductance L is high. If this is the case, the procedure is terminated in step S21 to select a different collision avoidance measure.

[0102] If the relief flow exceeds i L not the maximum current I L,max The inductance L is subjected to a collision check in the procedure already outlined in Fig. The 7 known steps S8. If there is no collision, the known steps S9 and S10 follow, at the end of which the control signals are passed to the driver.

[0103] If a collision is detected in step S8, a time interval t between the switching times is set in step S22. P determined, in this case t2 - t1. Based on the determined time interval, a shift of the switching times relative to each other is carried out in a process step S23, in particular one of the switching times t Pto the previously calculated time interval of the switching times t P postponed. The switching point tP' The opposite flank of phase P is shifted accordingly in order to avoid changing the average target voltage.

[0104] Based on the shifted switching times, steps S6 and S7 are then repeated to determine the charging of the circuit breakers S P required relief flows i L and the corresponding activation times T AP of the relief network 7. Subsequently, the further steps are carried out, in particular step S20, to check whether the changed relief flows i L This leads to an overload of the inductance L, i.e., exceeding the maximum current I. L,max , lead.

[0105] With regard to the Fig. 10 and Fig. Section 11 describes a fourth collision avoidance measure. Fig. 10 is a Fig. Figure 5 shows a corresponding collision case in which the switching times t1 and t2 of two switching edges, which require relief for soft switching, lead to a temporal overlap of the activation of the corresponding relief switches B. P This would lead to a collision. To avoid a collision, the relief of one of the switching edges is omitted in this case, so that switching of the corresponding circuit breakers S P This is achieved through so-called hard shifting. The in Fig. The collision avoidance measure 10 shown therefore accepts individual unloaded switching processes and thus switching losses in order to avoid a collision.

[0106] In the case of the collision avoidance measure according to Fig. 10. Individual switching edges are not relieved of load. This results in only minor switching losses. The reason for this is the previously discussed relationship to the period T. sshort length of maximum activation time T during a switching period AP , which means that collisions only occur in rare cases. In such cases, only one of the switching operations to be relieved is hard-shifted. Overall, this results in hard shifting for a small proportion of the switching operations to be relieved.

[0107] To reduce shifting losses caused by occasional hard shifting in accordance with the collision avoidance measure in Fig. To further reduce the load by 10, preferably the switching operation whose phase P has a lower load current i is not relieved. P wears. In the Fig. In the 10th switching period under consideration, the load current i2 is smaller than the load current i1. Consequently, phase 2 is selected for hard switching at time t2. The hard-switched edge is in Fig. 10 marked with reference 27.

[0108] In Fig. Figure 11 shows a schematic procedure for implementing the fourth collision avoidance measure. Procedure steps S1 to S7 correspond to the previously discussed procedures S1 to S7. Subsequently, the also known procedure step S8 is carried out, in which it is determined whether a collision exists between switching operations of different phases P that are to be relieved. If this is not the case, the known procedure steps S9 and S10 follow, at the end of which the control signals are passed on to the driver.

[0109] In the event of a collision, a comparison of the load currents is carried out in a process step S25. P of the colliding phases P. In a subsequent process step S26, the phase P with the larger load current i is selected. P for the relief. In process step S27, the occupancy of the relief network 7 is determined such that for phase P with the largest load current iP an activation of the corresponding relief switch B P This occurs. In process step S28, a hard switching operation is performed, i.e., a blocking of the activation of the relief switches B. P , for all further colliding phases P with low load current i P The specified control signals are then passed on to the driver in step S10.

[0110] The in the Fig. 10 and Fig. The collision avoidance measure shown in Figure 11 can be implemented without restrictions, in particular for all target space pointers u and regardless of the design of the components of the two-point inverter 1.

[0111] In relation to Fig. 12 will be an advantageous variant of the collision avoidance measure according to Fig. 10 described. Fig. 12 shows a Fig. 10 corresponding collision cases. As in relation to Fig. As described in section 10, the switching edge t1 of phase 1 is relieved. The relief switch B2 of phase 2 is not initially activated. The switching edge of phase 2, for which the relief switch B2 is not initially activated, is in Fig. 12 is marked with reference numeral 28. After the relief process for the switching time t1 of phase 1 has been completed and the corresponding relief switch B1 has been closed, the relief switch B2 of phase 2 is activated. The activation of the relief switch B2 occurs with a finite time interval t* > 0.

[0112] Since the relief switch B2 could only be activated after the relief switch B1 was closed to prevent a collision, only a small relief current i can occur. L build up in the relief network 7. The accumulating relief flow i LThis is insufficient to enable zero-voltage switching of the S2 circuit breakers of phase 2. Nevertheless, the partially generated relief current i ensures L This means that not the entire load current i2 of phase 2 needs to be switched. This reduces switching losses compared to hard switching of phase 2.

[0113] The in Fig. The measure shown in point 12 is particularly useful if the following applies to the time interval between the switching times of the two phases: |t2−t1|>tA1+tK1.

[0114] An available reduced current rise time tA2' then 0 <tA2'≤|t2−t1|−tA1−tK1.

[0115] The current relief compared to a completely unloaded switching process therefore corresponds to the quotient of the reduced current rise time and the current rise time actually required for the relief. tA2' / tA2.

[0116] With reference to Fig. 13 will be a further advantageous variant of the collision avoidance measure according to Fig. 10 described. Fig. 13 shows a Fig. 10 corresponding collision cases. As in Fig. 12 takes place in Fig. 13 a sequential activation of the relief switches B1 and B2 of phases 1 and 2 respectively. In contrast to the one in Fig. In the variant shown in Figure 12, the relief process for phase 1 is not fully completed. Instead, the relief switches B1 and B2 are hard-switched while the relief current i L The current, which was used to relieve the switching of the phase 1 circuit breaker S1, has not yet been completely dissipated. In the illustrated embodiment, a hard switching operation is performed by deactivating the relief switch B1 and activating the relief switch B2 at a finite relief current i. LThe relief switches B1 and B2 are therefore subjected to the current load of the not fully dissipated relief current i. L switched. The hard switching of the relief switches B1 and B2 is in Fig. 13, marked with reference numeral 29. The incompletely reduced relief flow i L The load is transferred to phase 2 via the load relief switch B2. After activation of the load relief switch B2, the load relief current i can be transferred. L to rebuild in order to help relieve the switching of the S2 circuit breakers of phase 2.

[0117] The in Fig. Variant 13 shown can be used in particular if the switching direction is identical and the following applies: |t2−t1|>tK1.

[0118] In this variant, a difference between the load currents i P must be taken into account. The relief flow to be achieved i LThis can be achieved, for example, if there is a sufficient time interval between the corresponding switching times. |t2−t1|>tK1+|i2−i1|i1⋅tA1. can be adjusted. Alternatively, the relief flow i L for the switching edge with the higher load current i P be set. The switching edge with the smaller load current i P can occur due to a higher commutation current ΔI ZVS will continue to be relieved of commuting.

[0119] The collision avoidance measures described above are each suitable and sufficient, in themselves, to operate the two-point inverter with a single inductor L of the load balancing network 7. A suitable collision avoidance measure can be selected for each specific application. If, in some applications, it is not possible to apply individual collision avoidance measures, for example, because the maximum current of the inductor would be exceeded, another collision avoidance measure can be used.

[0120] Fig. Figure 14 shows an advantageous control measure with which the relief current i L and thereby reduce losses in the relief network 7, especially at the inductance L. Fig. Figure 14 shows a switching period in which no collisions occur. The switching edges requiring relief, marked with an asterisk (*), are sufficiently spaced apart. The switching times t2 and t3 of the second and third phases are close together. The switching process at switching time t3 does not require relief because the load current i3 has a opposite sign to the potential change caused by the switching process. Since the potential changes caused at switching times t2 and t3 have the same sign, but the corresponding load currents i2 and i3 have opposite signs, simultaneous activation of the corresponding relief switches B2 and B3 would lead to coupling, in which the load current i3, due to its opposite sign, acts as the relief current for phase 2.Therefore, when both relief switches are activated simultaneously, the relief current i. required for the relief of phase 2 can be increased. L and thus those with the relief flow i L associated losses and activation time T A (i P ) can be reduced. However, simultaneous activation of the load-sensing switches B2 and B3 would lead to a collision as long as the output terminals A2 and A3 are at different potentials of the input DC voltage U. ZK To avoid this, the switching times t2 and t3 can be shifted relative to each other, so that a simultaneous switchover occurs. In the case of the Fig. In the embodiment shown in 14, the switching time t2 and, accordingly, the switching time are t2' the two switching edges of phase 2 are shifted, as shown by arrows 30 in Fig. 14 is marked. The reduced relief current iL is in Fig. 14 marked with arrow 31.

[0121] In general, a simultaneous reversal of two phases P, in particular by setting the switching times t P They are shifted relative to each other accordingly, with simultaneous activation of the respective relief switches B. P as well as when the potential difference has the same sign and the load current has a different sign i P This can be achieved by coupling different current paths in the relief network 7, which is advantageous for reducing the relief currents i. L and thus reduce losses. This measure can be implemented in addition to one or more, in particular all, of the aforementioned collision avoidance measures.

[0122] The collision avoidance measures described above, as well as their variants, in particular those in the Fig. 12 and Fig.The 13 variants of the fourth collision avoidance measure shown can also be combined with each other.

Claims

[1] Method for controlling a two-point inverter (1) with at least two phases (P), wherein the two-point inverter (1) has a switching arrangement (2) with - Input terminals (8) for the two poles of an input DC voltage (U ZK ), - one output terminal (A) per phase (P) P ), - a bridge circuit (6) with one half-bridge per phase (P), wherein the half-bridges are each circuit breakers (S P V ) have, via which the respective output port (A) P ) is connected to the input terminals (8), and - a relief network (7) with -- a bidirectional relief switch (B P ) per phase (P) and -- an inductance (L) for the at least two phases (P) that are connected via the respective bidirectional load shedding switch (B) P ) with at least two output terminals (A P) is connected, the procedure comprising the following steps: - Performing multiple switching cycles to generate an alternating current (i P ) at the output terminals (A P ), where each of the output terminals (A P ) the potential (ϕ) applied to at least two phases (P) P ) the input DC voltage (U ZK ) by switching the respective circuit breakers (S P V ) is reversed, - Activating the relief switches (B P ) to reduce switching losses when switching the respective circuit breakers (S P V ), preventing the relief switches (B P ) different phases (P) are activated simultaneously, while the output terminals (A P ) the respective phases (P) at different potentials of the input DC voltage (U ZK ) lay, - where switching times (tP,tP') the circuit breaker (S P V ) two phases (P) are shifted relative to each other in such a way that the respective relief switches (B P ) are activated with a time delay to relieve the switching process, - where an averaging of the relative shift of the switching times (tP,tP') the phases (P) occur over several switching periods. [2] Method according to claim 1, characterized by , that between the switching of the circuit breakers (S P V ) different phases (P) have a minimum time interval (Δt), where the minimum time interval (Δt) is greater than or equal to the sum of half the activation times (T) AP ) is, which is the relief switch (B P ) of the respective phases (P) to relieve the switching of the respective circuit breakers (S) P V ) are activated. [3] Method according to any of the aforementioned claims, characterized by , that a reversal of the polarity at the output terminals (A P ) potential applied to at least two phases (P) (ϕ) P ) occurs essentially simultaneously if the potential change has the same sign. [4] Method according to any of the aforementioned claims, characterized by , that the relief switch (B P ) of a phase (P) to relieve the switching of the corresponding circuit breakers (S P V ) is not activated as long as the relief switch (B) P ) is activated in another phase (P). [5] Method according to claim 4, characterized by , that the phase (P), whose relief switch (B) P ) is not activated, a smaller load current (i P ) leads as the phase (P), whose relief switch (B P ) is activated. [6] Method according to one of claims 4 or 5, characterized by, that the initially inactive relief switch (B P ) of one phase (P) sequentially to the previously activated relief switch (B) P ) of the other phase (P) is activated. [7] Multiphase two-point inverter, having - a switching arrangement (2), comprising -- Input terminals (8) for the two poles of an input DC voltage (U ZK ), -- one output terminal (A) per phase (P) P ), -- a bridge circuit (6) with one half-bridge per phase (P), wherein the half-bridges are each circuit breakers (S P V ) have, via which the respective output port (A) P ) is connected to the input terminals (8), and -- a relief network (7) consisting of --- a bidirectional relief switch (B P ) per phase (P) and --- exactly one inductance (L) for the at least two phases (P) that are connected via the respective bidirectional load shedding switch (B) P ) with at least two output terminals (A P ) is connected, - a control unit (3), wherein the control unit (3) is configured to carry out a control procedure according to any one of claims 1 to 6. [8] Multiphase two-point inverter according to claim 7, characterized by , that the inductance (L) of the relief network is an air coil.

Images