Model predictive control for converter with filter cells based on off-line computated pulse pattern and

Abstract

An electrical converter (10) comprising: a main stage (12) adapted to convert a DC voltage to an intermediate voltage comprising at least two voltage levels; and a filter cell stage (14) having, for each phase of the main stage (12), a filter cell (30), each filter cell (30) being adapted to add or subtract a cell voltage of said filter cell (30) to or from an intermediate voltage. A method for controlling an electrical converter (10) comprises a pattern determination section (44) comprising the steps of: selecting and reading an off-line calculated optimized pulse pattern (OPPbase) from a database, where the selected optimized pulse pattern (OPPbase) comprises switching timings of a primary stage (12) and a filter unit (30) on a next calculation window (alphawin) of a predetermined width (alphaw); the optimization pulse pattern (OPPbase) is adjusted by shifting the switching time of the optimization pulse pattern (OPPbase) in such a way that an average output voltage generated in the filter unit (30) is shifted towards an average output voltage reference (Uba1, abc) of the filter unit (30), which average output voltage reference (Uba1, abc) is determined as a function of a measurement in the filter unit (30), and the adjustments are compensated by a corresponding modification of the switching time of the main stage (12). The method further includes a model predictive control section (46) including steps performed several times during the next calculation window ([alpha] win): modifying the adjusted optimization pulse pattern (66) by moving at least one transition time of the switching time, such that a flux error determined from the difference between the estimated flux (74) of the electrical converter (10) and the reference flux trajectory (psi ref) is minimized; and applying at least a next switching time to the electrical converter (10) according to the modified adjusted pulse pattern.

Description

Technical Field

[0001] The present invention relates to the field of high power converter control. In particular, the present invention relates to a method, a computer program, a computer readable medium and a controller for controlling an electric converter. The present invention also relates to an electric converter. Background Art

[0002] Especially in medium voltage applications, the use of a 3L(A)NPC+AF topology gives a cost-effective electrical converter topology with high power quality. Here, 3L means 3-level, (A)NPC means neutral point clamping and / or active neutral point clamping, and AF means active filter. 3L(A)NPC is also referred to below as the primary stage of the converter. This topology has been proposed for several years. However, optimized hardware design, reliable control and fail-safe operation remain challenges.

[0003] One of the main advantages of the topology is the small voltage steps that can be generated. Small voltage steps reduce machine side dv / dt and overvoltage as well as current THD (Total Harmonic Distortion) and involved harmonic losses in the machine.

[0004] Balancing the active filter cell voltage is one of the main challenges when controlling this system. The active filter cell has no external power supply, so a zero average voltage needs to be injected during the fundamental cycle, otherwise the cell will charge / discharge. In order to fully benefit from the improved power quality, the control scheme needs to cover the entire range of operating points of the machine from low speed to high speed, and the active filter cell voltage needs to be balanced in all operating points.

[0005] The control scheme may work using an optimized pulse pattern (OPP) calculated offline and stored in a lookup table or database. For example, as described in EP 2 469 692 A1, the optimized pulse pattern may be modified online based on model predictive control, wherein the pre-calculated optimized pulse pattern switching angle is modified online to follow a reference flux trajectory. Typically, the optimized pulse pattern may be a sequence of switching transitions that has been optimized with respect to a certain optimization objective. For example, the optimized pulse pattern may be calculated offline for all modulation indices and switching frequencies of a motor or any other physical system acting as a load of the converter, and may be optimized to minimize the total current distortion for a given switching frequency.

[0006] In particular, for a control scheme based on model predictive pulse mode control (hereinafter also denoted as MP3C core algorithm in this article), the balancing scheme must also be compatible with the entire operating range. A previously developed balancing concept for active filter cells is based on common mode injection. For example, it comprises the following steps: first, the cell voltages are equalized, and then the cell voltages of all (e.g., three) phases are adjusted together to a reference. However, this approach has the disadvantage that it does not operate as expected in certain operating points and may result in oscillations of the filter cell voltages.

[0007] Existing control schemes for nominal speed operation of a 3L(A)NPC+AF converter can be summarized as follows.

[0008] One possibility is conventional real-time carrier-based pulse width modulation (CB-PWM) for the main stage of the converter (3L(A)NPC) and for the filter cell stage (AF). An error signal derived from the difference between the main stage switching state and a sinusoidal voltage reference is fed to the PWM modulator for the filter cell. A balancing signal is added on top to stabilize the filter cell. This scheme uses a different top-level control scheme such as vector current control and may not be compatible with the MP3C core. Furthermore, if the ratio between the filter cell voltage and the main stage half-dc-link voltage is below 40%, the modulator of the filter cell may be repeatedly driven into the overmodulation range, which may result in difficulties in balancing the cell voltages.

[0009] Another possibility is an OPP-based MP3C for a 3L(A)NPC+AF converter with common mode injection to balance the filter cells. As mentioned above, this control scheme may not perform as expected at certain operating points, which may result in low frequency oscillations of the cell voltage. In addition, the cell voltage cannot be controlled independently. Summary of the invention

[0010] It is an object of the invention to improve the control of a multilevel converter comprising a filter unit.It is a further object of the invention to reduce the computational requirements of such a control method and to reduce the total harmonic distortion and losses of the electrical converter.

[0011] These objects are achieved by the subject-matter of the independent claims. Further exemplary embodiments are evident from the dependent claims and the following description.

[0012] One aspect of the invention relates to a method for controlling an electrical converter. The electrical converter may be a medium voltage converter or a high voltage converter adapted to handle voltages up to 6.6 kV or higher. The method may provide an algorithm for operating the converter at a medium to nominal fundamental frequency of an electric machine or any other physical system acting as a load for the converter.

[0013] According to an embodiment of the invention, an electrical converter comprises a main stage adapted to convert a DC voltage into an intermediate voltage comprising at least two voltage levels. The main stage may comprise one or more (A)NPC half-bridges connected in parallel with the DC link. The output of the half-bridge may be an intermediate voltage, which may be a multi-phase voltage, in particular a three-phase voltage. Depending on the topology of the main stage, the output voltage may have two, three, five or more voltage levels. In the case of a 3L (A)NPC half-bridge, there may be three output voltage levels.

[0014] According to an embodiment of the invention, the electrical converter further comprises a filter cell stage (hereinafter also referred to as "cell stage") having a filter cell for each phase of the main stage, wherein each filter cell is adapted to add a cell voltage of the filter cell to the intermediate voltage or to subtract the cell voltage from the intermediate voltage. Each filter cell may comprise a cell capacitor providing the cell voltage, the cell capacitor being connected in parallel with two half-bridges providing the input and output of the filter cell. The filter cell stage may be regarded as an active filter (AF) of the electrical converter.

[0015] According to an embodiment of the present invention, the method includes a mode determination part, which includes the following steps: selecting and reading an optimized pulse pattern (OPP) calculated offline from a database (i.e., a lookup table generated offline), wherein the selected optimized pulse pattern includes the switching moments of the main stage and the filter unit level on the next calculation window of a predetermined length or width within the fundamental period of the main stage; adjusting the selected optimized pulse pattern by moving the switching moment of the selected optimized pulse pattern based on measurements in the filter unit, so that the average output voltage generated in the filter unit is offset toward the average output voltage reference of the filter unit; and compensating for these adjustments to the unit level by corresponding modifications to the switching moment of the main stage.

[0016] Therefore, the average output voltage reference of the filter unit can be used to balance the filter unit capacitor voltage. Then, the nominal unit pulse pattern of each filter unit, which can be determined according to the selected optimized pulse pattern, is adjusted so that the filter unit will be better balanced. In order to achieve this, the switching moments of the unit pulse pattern are moved so that this purpose is achieved. Balancing means that the unit capacitor voltage moves towards the unit capacitor voltage reference. The unit capacitor voltage reference can be the desired voltage of the unit capacitor.

[0017] For example, a typical configuration of an electrical converter with a 3L(A)NPC+AF topology (see above) foresees a rather low cell voltage of the filter cells, typically 1 / 3 of half of the DC link voltage of the main stage or even lower, such as 1 / 5 of half of the DC link voltage of the main stage. The pulse pattern for the offline generation of the filter stage (AF) is generated so that the fundamental component is zero. However, this does not mean that the average filter cell voltage will be zero over the calculation window. Therefore, a complex balancing method is required, which ensures that the combined output voltage average value of the main stage (3L(A)NPC) and the cell stage (AF) does not change over the calculation window. To this end, in the present method, a time-domain based average value adjustment algorithm as described in this article has been implemented, which is able to implement an arbitrary average value of the filter cell output voltage over a given calculation window.

[0018] To determine the average output voltage reference, the capacitor voltage of the filter cell may be measured, compared to a capacitor voltage reference, and manipulated to generate a filter cell average output voltage reference signal for cell balancing. Furthermore, the injected filter cell balancing voltage may be compensated by modifying the switching moments of the main stage, such that the sum of the main stage and filter cell stage voltages (e.g., for each of three or another number of phases) is constant and does not have an impact on the flux trajectory or torque control of the converter.

[0019] In general, the optimized pulse pattern may include two pulse patterns: a main pulse pattern for the main stage and a unit pulse pattern for the filter unit stage. Otherwise, i.e., before making adjustments, if the optimized pulse pattern is provided as a total (i.e., combined or accumulated) pulse pattern for both the main stage and the unit stage, it may be parsed as a main pulse pattern for the main stage and a unit pulse pattern for the unit stage. Each pulse pattern is or includes one or more sequences of switching moments over a specific time interval (i.e., the next calculation window).

[0020] The switching moment indicates a point in time and / or a transition time and the switch position to be switched to at that point in time. In other words, for each switching moment, ie for each transition, the following data can be stored in the pulse pattern: transition angle, final position at the end of the transition and position change during the transition.

[0021] It has to be noted that the optimized pulse pattern or respectively the main pulse pattern and the unit pulse pattern may be multi-phase in number, ie there may be different switching moments for each phase of the intermediate voltage.

[0022] The optimized pulse pattern can be provided as a multiphase pulse pattern from the outset. Otherwise, ie if the optimized pulse pattern is provided as a single-phase pulse pattern, it can be divided into individual phases so that a multiphase pulse pattern is obtained in order to adjust each phase of the intermediate voltage individually.

[0023] According to an embodiment of the present invention, the method also includes a model predictive control part, which includes the following steps which are executed several times during the next calculation window: modifying the adjusted optimized pulse mode (or the corresponding main pulse mode and the unit pulse mode) by moving at least one transition time of the switching moment so that the flux error determined according to the difference between the estimated flux and the reference flux trajectory of the electric converter is minimized; and applying at least the next switching moment to the electric converter according to the modified adjusted pulse pattern.

[0024] The model predictive control part as well as the model predictive control may be performed as in EP 2 469 692 A1 incorporated by reference. Furthermore, model predictive control of pulse patterns created based on pulse width modulation is described in the chapter "Pattern Controller" of EP 3 729 637 A1 incorporated by reference.

[0025] The model prediction method attempts to optimize the control objective which can be modeled using an objective function. The objective function is a function that depends on parameters and quantities that are determined from an adjusted optimized pulse pattern (or two adjusted pulse patterns for the main stage and for the unit stage, respectively) using time-shifted switching moments. The pulse pattern switching moments are time-shifted so as to minimize the objective function.

[0026] The objective function is modeled to minimize the flux error between the estimated flux of the electrical converter and a reference flux trajectory. The actual estimated flux may be determined from measurements in the electrical converter.

[0027] When the adjusted pulse pattern has been optimized (i.e. modified) in this way, at least one next switching instant is applied to the electrical converter, i.e. the main stage and / or the filter unit stage. This is achieved by switching the switches of the main stage and / or the filter unit stage to the position required by the switching instant.

[0028] The model predictive control portion may be repeated several times using the same pulse pattern and reference flux trajectory as those determined in the mode determination portion. The same adjusted pulse pattern (e.g., the same main pulse pattern and the same unit pulse pattern) determined by the mode determination portion may be used again in the model predictive control portion at a later time instant for again calculating the next switching instant.

[0029] The model predictive control part may be executed more frequently than the mode determination part, meaning with a higher execution frequency. In the mode determination part, the adjusted optimized pulse pattern or two adjusted pulse patterns for the main level and the unit level are provided several times, for example, as a lookup table, in order to be used as an input to an optimizer in the model predictive control part, which optimizer offsets the transition time.

[0030] The model predictive control part can select the appropriate offline calculated OPP (to be adjusted in the mode determination part) based on an outer control loop, which can generate the required modulation index and flux reference angle, for example. To account for magnetic field weakening, the flux reference amplitude may need to be further scaled.

[0031] According to an embodiment of the present invention, the next calculation window may be a fixed angle range relative to the optimized pulse pattern. The length, start angle and / or end angle of the next calculation window may be fixed relative to the optimized pulse pattern.

[0032] According to an embodiment of the present invention, the next calculation window may be a time period of a predetermined length within the fundamental wave period of the primary stage, or a time period of a predetermined length of the reference output voltage of the electrical converter. For example, the calculation window may be a time period corresponding to 1 / 6 of the fundamental wave period (or an angular width of 60°).

[0033] If the fundamental frequency of the main stage ranges from, for example, 25 Hz to 60 Hz, the execution period of the mode determination part of the algorithm with an update frequency corresponding to 60° of the fundamental period may be 6.7 ms to 2.8 ms. This may leave enough time for the calculations required by the algorithm without excessive computational burden. Since the pulse pattern of the main stage can have a very low pulse number, it often happens that only two of the three phases are switched. In order to ensure that there are always at least two phases available, the modulation index-dependent offset of the next calculation window can be set accordingly.

[0034] In particular, the mode determination part algorithm proposed in this article is compatible with the model predictive pulse mode control (also referred to as MP3C core algorithm) using the OPP calculated offline. As already mentioned, according to the selected optimized pulse mode calculated offline, the mode determination part can repeatedly calculate the corresponding nominal pulse mode for the main level and (active) filter unit level for the upcoming calculation window online and / or generate a multi-phase OPP according to the original single-phase OPP. Subsequently, the nominal pulse mode of the filter unit level is adjusted to achieve unit balance, and other adjustments can be made as described above and below in this article. Then, based on the adjusted pulse mode, an accurate flux reference trajectory can be pre-calculated for the upcoming calculation window. Finally, the adjusted pulse mode and the corresponding flux reference trajectory can be written into a lookup table and passed to the model predictive control part. The model predictive control part ensures closed-loop control by tracking the provided flux reference trajectory using an online modification of the provided adjusted optimized pulse mode.

[0035] Compared to other methods used to control a converter comprising a main stage and a filter unit stage, the proposed mode determination part of the method may specifically have the following advantages:

[0036] It is compatible with the model predictive control part of the approach and it can be aligned with the low speed flux trajectory and pulse pattern generation approach. Thus, only one control scheme is required for both low speed and nominal speed operation and no complex switching logic is required.

[0037] It combines a time-domain based approach to filter cell voltage balancing. Since the adjustment of the optimized pulse pattern is done by shifting the switching moments in the time domain, it does not rely on power transfer at the fundamental frequency and can therefore even operate at zero fundamental frequency (stationary). The approach is therefore more straightforward and simple to implement and can also operate well during transients. In particular, the cell voltage can be stabilized very effectively.

[0038] It can be calculated using a reduced update frequency (e.g., equivalent to 60° of the fundamental period). If a fundamental frequency range of 20 Hz to 60 Hz is assumed, this corresponds to an update rate of 8.3 ms to 2.8 ms or a maximum update frequency of about 360 Hz. Therefore, the algorithm of the mode determination part can be run as a background task, leaving sufficient computing resources for other fast control blocks such as the model predictive control part.

[0039] ●Each phase active filter unit voltage can be independently balanced in the adjustment step.

[0040] • It can easily handle synchronization pulses between the main stage and the filter unit stage as may be required for voltage step reduction (VSR).

[0041] ● Using the offline calculated nominal flux trajectory as the initial condition for each calculation window, the online reference flux trajectory can be easily calculated.

[0042] According to an embodiment of the present invention, each phase of the average output voltage reference corresponding to the filter unit is based on the measured capacitor voltage signal of the filter unit and the difference of the reference capacitor voltage of the filter unit. The capacitor voltage signal can be filtered using a low-pass filter to reduce ripple, and the capacitor voltage reference can be subtracted subsequently. For example, the unit capacitor voltage of the filter unit in three (or any other number of) phases of the intermediate voltage can be measured individually, and the unit capacitor voltage can be optionally processed using a low-pass (moving average) filter. The difference obtained by subtracting the reference capacitor voltage of the filter unit can be processed using a proportional (P) controller or a proportional integral (PI) controller. The output signal can be interpreted as a non-zero differential voltage reference, which needs to be injected by the filter unit to accumulate charge in the filter unit capacitor, and thus the unit capacitor is charged or discharged.

[0043] If the cell (capacitor) voltage matches the reference (capacitor) voltage, the average voltage produced by the filter cell over one calculation window should not change, since the OPP is generated so that the fundamental component of the cell pulse pattern is zero. The fundamental component power is provided only by the primary stage. Further, the sign of the phase current can be interpreted so that the differential voltage command for the filter cell drives the cell voltage in the correct direction, i.e., the PI controller output signal is multiplied by the sign of the corresponding phase current. In order to compensate for delays in the current measurement and avoid applying the wrong sign over the calculation window, the phase angle of the fundamental current relative to the flux reference angle can be used as a reference from which the current sign can be advanced so that the correct sign is applied over the upcoming calculation window.

[0044] As mentioned above, the balancing voltage to be injected by the filter unit can be compensated by the main stage so that the sum of the main and unit level (three-phase) voltages remains unchanged and does not affect the flux trajectory or torque control. Since the pulse pattern of the main stage can have a very low number of pulses, it is often the case that only two of the three phases are switched. In order to ensure that there are always at least two phases available, the modulation index-related offset of the calculation window can be set accordingly. In order to properly compensate the voltage injected by the filter unit, the differential signal can be assigned to the main phase being switched. Then, the expected volt-second offset of each filter unit phase in the filter unit phase can be calculated taking into account the possible pre-offset of the synchronization angle of each filter unit. The result can be a total effective volt-second offset reference for each unit pulse pattern in the unit pulse pattern. Then, the initially selected optimized pulse pattern (e.g., the total multi-phase pulse pattern of the main stage and the filter stage) is updated (i.e., adjusted) using the modified switching moment.

[0045] In order to prevent the model predictive control part from modifying the adjusted optimized pulse pattern so that the balance of the filter unit is destroyed, the reference flux trajectory can be determined according to the adjusted optimized pulse pattern in the next calculation window. This can not only ensure that the unit voltage balancing algorithm works as expected, but also simplify the balance of the neutral point potential of the primary stage.

[0046] Therefore, according to an embodiment of the present invention, the mode determination part further comprises: determining a reference flux trajectory on a next calculation window according to the adjusted optimized pulse mode. This can be done by integrating the output voltage as indicated by the total adjusted optimized pulse mode or by integrating the sum of the output voltage of the main stage as indicated by the adjusted main pulse mode and the output voltage of the filter unit stage as indicated by the adjusted unit pulse mode, depending on the specific situation.

[0047] The reference flux trajectory may be, for example, a vector quantity having a component for each phase or a vector quantity in an orthogonal reference frame described using two components (referred to below as α-β components).

[0048] A reference flux trajectory may be determined for a particular angular range corresponding to the next calculation window. Using the reference flux trajectory may be beneficial because it includes the flux ripple and therefore matches the adjusted pulse pattern. Thus, in the absence of external interference, the adjusted pulse pattern is applied without modification.

[0049] It is also possible to use the circular flux trajectory directly as reference flux trajectory. However, in this case the mode switching moments may be modified by the model predictive controller to better match the circle, thus overriding the tuning and possibly disturbing the cell balancing target.

[0050] If the optimized pulse pattern calculated offline contains a nominal flux trajectory also calculated offline, as is usually the case, the initial flux trajectory angles and α-β components can be easily interpolated and used as starting points for accurately calculating the ripple in the reference flux trajectory using a simple piecewise integration of the adjusted optimized pulse pattern. The newly generated reference flux trajectory corner points can consist of a unique angle in the α-β plane plus the flux α-β component over the calculation window. In the case where neutral point balancing is also considered as further described below, the nominal flux trajectory can be used as an initial condition, for example, only at each second or third update. Otherwise, the endpoint of the previously calculated flux trajectory segment can be used as the starting point of the next calculation window. This can ensure that the neutral point balancing adjustment occurs at 120° or 180°.

[0051] According to an embodiment of the present invention, the main stage of the electric converter includes a DC link, and the method further comprises: adjusting the optimized pulse mode in the mode determination part by moving the switching moments of the optimized pulse mode so as to offset the voltage difference between the upper DC link voltage and the lower DC link voltage of the measured DC link to the reference neutral point voltage (nominal zero). This part of the adjustment process is also denoted herein as neutral point (NP) potential balance. For example, the difference between the upper DC link voltage and the lower DC link voltage as obtained from the measurement can be calculated, and the difference can be optionally processed using a low-pass (moving average) filter. Then, the result is compared with the reference neutral point voltage value (which can be nominally zero, for example), and the difference obtained according to the comparison can be processed, for example, using a P controller or a PI controller. For example, the neutral point error can be compensated by adjusting the main stage pulses accordingly. For example, the correction can be added to the main stage pulse mode as a common mode component, but it can also be added by offsetting the angle or other parameters of the switching moments of the total optimized pulse mode in other ways. As mentioned above, the neutral point balance adjustment can occur at 120° or 180° of the fundamental period.

[0052] According to an embodiment of the present invention, the method comprises: determining whether the flux reference angle is within the calculation window; generating a trigger signal if the flux reference angle leaves the calculation window; and triggering the mode determination part of the method by the above-mentioned trigger signal. The trigger signal can be generated, for example, based on the flux reference angle and the modulation index provided by the external loop controller and the angle representing the delay until the next control cycle. If the trigger signal is true, the nominal pulse mode provided by the mode determination part, for example, a 3-phase mode set, can be updated for the next calculation window. For example, the offset of the calculation window can be determined by the modulation index, and it can be determined from this whether the flux reference angle is within the current window. Determining this can also take into account any delays in the microcontroller to ensure that the pulse mode for the next calculation window is always available. This can also include double buffering of the mode so that it is always possible to switch to the next mode when necessary. This ensures that there is always a mode available to feed into the model predictive control part of the method.

[0053] If it is determined as described above that an update is required, the currently available offline calculated optimized pulse pattern, e.g., single-phase OPP, can be read from the model predictive control portion and divided into three (or another number of) phases on the next calculation window. The resulting new multiphase pattern can then be parsed into corresponding switching moments for the main stage and for the filter unit level, e.g., including the angles of the main stage and the filter unit level and the corresponding switch positions and changes in the switch positions. The first angle may always be zero, and the last angle may always be the length of the calculation window. The switch angle can be normalized, for example, by subtracting the starting angle of the calculation window and noting that the span from 360° to 0° is interpreted. A flag can also be generated for each unit level angle synchronized with the main stage angle, which can be used to reduce the combined voltage step by switching in the opposite direction using a double step.

[0054] According to an embodiment of the present invention, the steps in the mode determination part are performed by a mode determination controller, and the steps in the model prediction control part are performed by a model prediction controller. The execution of the model prediction control part can be independent and / or generally requires a higher update frequency for closed-loop control. The two parts of the method executed at different speeds can be executed by two different controllers. The model prediction controller can have a faster execution speed than the mode determination controller. For example, the speed of the model prediction controller can be at least 10 times faster than the speed of the mode determination controller.

[0055] After making adjustments as described above and below herein, the resulting updated optimized pulse pattern (which may be, for example, an overall multi-phase pulse pattern) may be stored in a lookup table during the mode determination portion and may be made available to the model prediction portion for use in the next calculation window. According to an embodiment of the present invention, the adjusted optimized pulse pattern may be stored in a lookup table together with the newly generated reference flux trajectory. The table may be provided in a common memory of the mode determination controller and the model predictive controller. In other words, the model predictive controller uses the contents of the lookup table provided by the mode determination controller.

[0056] As mentioned above, online loading of the selected offline calculated optimized pulse pattern can be performed at a reduced update frequency (such as 360 Hz) (i.e., for a long period of time), followed by filter unit balancing and other mode adjustments as described herein, and calculation of the corresponding reference flux trajectory. It can be aligned with a calculation window of, for example, 60° equivalent to the fundamental period. The execution of the model predictive control part is independent and generally requires a higher update frequency for closed-loop control.

[0057] According to an embodiment of the invention, the algorithm calculates the adjusted optimized pulse pattern and the corresponding flux reference for the next two calculation windows and stores them in a lookup table. This can simplify the implementation on a microcontroller because the algorithm can run in a lower priority interrupt and calculate the nominal pattern for the next calculation window while the model prediction part works on the active calculation window using the corresponding adjusted optimized pulse pattern that has been previously calculated in the pattern determination part.

[0058] As mentioned above, the model predictive control portion executes a nominal pulse mode and controls the flux (which may be a fundamental flux and a ripple, for example) to follow a reference flux trajectory.

[0059] According to an embodiment of the invention, when the stator flux reference angle (e.g., as commanded and / or requested by an external loop controller) is outside the angular range covered by the pre-calculated reference flux trajectory based on the adjusted optimized pulse pattern, the flux error is determined from the difference between the estimated flux of the electrical converter and the circular flux trajectory, rather than from the difference between the estimated flux and the pre-calculated reference flux trajectory. For example, in the presence of a torque step or any other disturbance that causes the flux reference angle to be outside the angular range corresponding to the calculation window, the reference flux trajectory stored in the lookup table may be skipped and a circular flux reference may be used alternatively.

[0060] In summary, the proposed active filter cell balancing method for 3L(A)NPC+AF converter topology is seamlessly integrated into the OPP-based model predictive pulse mode control (OPP-MP3C), which provides high dynamic control response and excellent power quality simultaneously. In particular, the newly developed mode determination part algorithm can be based on a per-phase approach. The original optimized pulse patterns of the main stage and the filter cell stage are manipulated (i.e., adjusted) online over a fixed pre-calculation window (e.g., 60°) to achieve the desired balancing action, and the flux trajectory reference can also be recalculated. The mode determination part can be performed at a reduced update frequency (e.g., every time period equivalent to 60° of the fundamental period) so that the computational burden is low. The proposed control scheme can, for example, include the following four main steps: reading the offline calculated optimized pulse pattern of 3L(A)NPC+AF; manipulating the pulse pattern over the next calculation window to inject the filter cell average voltage that is conducive to rebalancing the cells; pre-calculating the reference flux trajectory over the calculation window; creating a lookup table containing the adjusted pulse pattern and the resulting flux reference, and executing it through the model predictive control part.

[0061] The control scheme proposed herein allows and / or is based on independent control of three (or any other number of phases) filter cell voltages. The control scheme starts from the OPP calculated offline and can generate an independent mode for each phase, which ensures the balance of the active filter cells and optionally also the balance of the neutral point potential of the primary stage of the converter. It is compatible with a wide range of ratios between the filter cell voltages and the primary half DC link voltage (especially also covering low ratios of about 20%), which may be particularly important for 6kV drive lineups.

[0062] Other aspects of the invention relate to a computer program for controlling an electric converter, which, when executed by a processor, is adapted to carry out the steps of the method as described herein, and to a computer readable medium in which such a computer program is stored. Both the mode determination part and the model predictive control part can be executed as software on one or more processors.

[0063] The mode determination controller may have a processor, and the mode determination part may be executed on the processor. The mode determination part may be stored in a memory of the mode determination controller.

[0064] The model predictive controller may also have a processor, and the model predictive control portion may execute on the processor. The model predictive control portion may be stored in a memory of the model predictive controller.

[0065] The computer readable medium may be a floppy disk, a hard disk, a USB (Universal Serial Bus) storage device, a RAM (Random Access Memory), a ROM (Read Only Memory), an EPROM (Erasable Programmable Read Only Memory) or a FLASH memory. The computer readable medium may also be a data communication network, such as the Internet, which allows the program code to be downloaded. In general, the computer readable medium may be a non-transitory medium or a transient medium.

[0066] Another aspect of the invention relates to a controller for controlling an electric converter. The controller is adapted to perform a method as described herein. The controller may include a mode determination controller for performing the steps in the mode determination part of the method. The controller may include a model predictive controller for performing the steps in the model predictive control part of the method.

[0067] The controller may include one or more processors, wherein the method is executed as software. The method may also be implemented at least partially or completely in hardware such as a DSP or FPGA.

[0068] Another aspect of the invention relates to an electrical converter which is controlled in a manner such as described herein.

[0069] According to an embodiment of the present invention, the electrical converter comprises a main stage, adapted to convert a DC voltage into an intermediate voltage comprising at least two voltage levels; a filter unit stage, having a filter unit for each phase of the main stage, wherein each filter unit is adapted to add a cell voltage of the filter unit to the intermediate voltage or to subtract the cell voltage from the intermediate voltage; and a controller, such as the controller described herein.

[0070] In general, it should be understood that features of the method as described above and below may be features of the computer program, computer readable medium, controller and electrical converter as described above and below, and vice versa.

[0071] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The subject matter of the invention will be explained in more detail in the following text with reference to exemplary embodiments illustrated in the drawings.

[0073] Figure 1 A schematic circuit diagram of an electrical converter according to an embodiment of the invention is shown.

[0074] Figure 2 A schematic block diagram of a controller according to an embodiment of the present invention is shown.

[0075] Figure 3aand Figure 3b Shown is a diagram of the fundamental current, its sign and its angle over several cycles illustrating how the sign of the current changes in an electrical converter according to an embodiment of the invention.

[0076] Figure 3c A diagram illustrating how the sign of the current is carefully interpreted in a filter cell balancer according to an embodiment of the invention is shown.

[0077] Figure 4 Shown is a diagram illustrating the switching angles over the full range of the modulation index of the primary stage of an electrical converter according to an embodiment of the invention.

[0078] Figure 5a A single-phase total optimized pulse pattern for a control method according to an embodiment of the present invention is shown.

[0079] Figure 5b The control method according to an embodiment of the present invention is shown in FIG. Figure 5a The pulse patterns of the individual phases are obtained from the overall optimized pulse pattern of the three-phase.

[0080] Figure 6 A table is shown for decoding the overall switch position into switch positions of the main stage and the filter unit stage in a control method according to an embodiment of the invention.

[0081] FIG. 7 shows a diagram illustrating adjustment of each phase of a unit pulse mode, a main pulse mode, and a total pulse mode in a control method according to an embodiment of the present invention.

[0082] Figure 8a A diagram showing the flux ripple of an updated reference flux trajectory compared to an original reference flux trajectory in a control method according to an embodiment of the present invention.

[0083] Figure 8b Shows Figure 8a The amplitude of the reference flux trajectory.

[0084] In principle, identical components are provided with the same reference symbols in the figures. DETAILED DESCRIPTION

[0085] Figure 1 An electrical converter 10 is shown having a main stage 12 and a filter unit stage 14. The electrical converter 10 is adapted to convert a first AC voltage provided by an electrical grid 15 into an output voltage to be supplied to an electrical load 16.

[0086] The primary stage 12 includes a rectifier 18, which may be a passive diode rectifier and is configured to transform an AC voltage from the grid 15 into a DC voltage, which is supplied to a DC link 20. The DC link 20 may be a split DC link.

[0087] Furthermore, the main stage 12 includes three output converters 22, wherein each output converter 22 is adapted to transform a DC voltage in the DC link 20 into a phase of an intermediate voltage provided at an output 24 of the main stage 12. The output converters 22 are connected in parallel with the DC link 20, for example, via a clamping inductor and / or resistor 28 or directly. Each of the output converters 22 may include a clamping circuit consisting of a clamping inductor and / or resistor 28 and a clamping capacitor 26.

[0088] As shown, each of the output converters 22 may include a 3-level neutral point clamped (3LNPC) half-bridge 23, which may be based on IGCT, but other topologies and semiconductor types are also possible. The half-bridge 23 may also be an ANPC half-bridge. Furthermore, other topologies are possible for the entire main stage 12. In general, the main stage 12 may be adapted to provide a two-level intermediate voltage or a multi-level intermediate voltage at the output 24, optionally with more than one phase. As shown in FIG. Figure 1 The half-bridge 23 shown is suitable for generating three different voltage levels for respective phases of the intermediate voltage.

[0089] The clamping inductor and / or resistor 28 together with the clamping capacitor 26 provide an IGCT di / dt clamping circuit. In the case of other types of semiconductor switches such as IGBTs, the clamping circuit may be omitted. The clamping circuit 26 may be used to control the device voltage during switching.

[0090] For each phase of the main stage 12, the filter unit stage 14 comprises a filter unit 30, which may be regarded as an active filter, and an optional passive output filter 32, which are connected in series between the corresponding output 24 of the main stage 12 and the output 34 of the electrical converter 10. More than one filter unit 30 may be connected in series and / or in parallel between the output 24 and the output 34 of a phase. Each of the passive filters 32 may comprise an inductor, a resistor and / or a capacitor.

[0091] Each filter unit 30 comprises two half bridges 36 connected in parallel with a unit capacitor 38. A first midpoint of one of the half bridges 36 is connected to the output 24 of the main stage 12. A second midpoint of the other half bridge 36 provides an output 40 of the filter unit 30 and is connected to the corresponding output 34 of the electrical converter 10 via an optional passive filter 32.

[0092] Instead of the passive filter 32, a dv / dt filter may also be used.

[0093] The filter unit 30 comprises three different switching states, which can be realized by correspondingly switching their semiconductor switches (two semiconductor switches per half bridge), which semiconductor switches can be, for example, IGBTs. In the first switching state, the filter unit 30 connects the output 24 directly to the output 40. There may be two redundant states here: two upper switches are open or two lower switches are open. Usually, the two redundant states alternate in the switching cycle. In the second switching state, the converter unit 30 connects the cell capacitor 38 between the outputs 24, 40, so that the cell capacitor voltage of the cell capacitor 39 is added to the intermediate voltage provided at the output 24. In the third switching state, the filter unit 30 connects the cell capacitor 38 between the outputs 24, 40, so that the cell capacitor voltage of the cell capacitor 38 is subtracted from the intermediate voltage provided at the output 24. In this way, the intermediate voltage from the main stage 12, which is usually shaped as a step function due to the limited number of levels of the main stage, can be converted into a voltage (i.e., the output voltage) that better approaches a sinusoidal and / or continuous output voltage reference.

[0094] Figure 2 The block diagram of the controller 42 is shown, which is composed of a mode determination controller 44 and a model prediction controller 46. The controller 42 can be used to control Figure 1 The blocks of the controller also indicate method steps of the control method performed by the controller. The blocks of the mode determination controller 44 can also be seen as the mode determination part 44 of the control method. The blocks of the model prediction controller 46 can also be seen as the model prediction control part 46 of the method.

[0095] Mode determination section 44 may be executed at a time period equivalent to 60° of the fundamental frequency of main stage 12 (MC) of electric converter 10. If the main stage fundamental frequency ranges from 25 Hz to 60 Hz, the execution cycle may be 6.7 ms to 2.8 ms.

[0096] The method is based on an optimized pulse pattern OPP generated offline in advance and stored in a database. base In the graphical determination part 44 and the model predictive control part 46 of the method, the optimized pulse pattern OPP can be selected and read from the database base For further online adjustment and modification. For example, the MP3C (Model Predictive Pulse Mode Control) core 47 of the model predictive controller 46 can select an appropriate optimized pulse mode OPP base , the optimized pulse pattern has the desired modulation index M as determined in block 48 i and flux reference angle To account for field weakening, the flux reference amplitude may need to be further scaled.

[0097] Based on the flux reference angle provided by the external loop controller and modulation index M i and an angle value representing the delay until the next control cycle, a trigger signal 50 is generated in a block 52 of the mode determination portion 44. If the trigger signal 50 is true (meaning "need to update"), then the next calculation window α will be updated as described herein. win The 3-phase nominal pulse pattern OPP is updated in the pattern determination section 44 abc Set, the 3-phase nominal pulse mode OPP abc The collection is based on the original single-phase optimized pulse mode OPP base Direct calculation (see below). Calculation window α within the fundamental period win The appropriate deviation angle α off It can also be determined by the modulation index Mi, thereby determining the flux reference angle Is it still in the current window? win The window will be left in the next controller cycle. If not, an update is required.

[0098] This can also take into account any delays in the microcontroller so that the next calculation window α win mode is always available. Furthermore, this can also include double buffering of the pulse mode, that is, for the next two calculation windows α win The pulse pattern is provided by the mode determination part 44 so that it is always possible to switch to the next mode when necessary. This ensures that there is always a mode available to feed into the MP3C (Model Predictive Pulse Pattern Control) core 47 of the model predictive controller 46.

[0099] If an update is required, that is, if the trigger signal 50 generated in box 52 is true, the currently available single-phase OPP selected by the MP3C core 47 is read in box 54. base And appropriately offset it to the next calculation window α win Then, in box 56, the new three-phase mode OPP abc Resolved to the main pulse mode OPP for the main stage 12 MCabc (also denoted MC or main converter) and a cell pulse pattern OPP for the filter cell stage 14 AFabc (Also denoted as AF or active filter).

[0100] Main pulse mode OPP MCabc and unit pulse mode OPP AFabc Including for the next calculation window α win The switching moments of the main stage 12 or the filter unit stage 14. This can include the main stage angle and the filter stage angle α AF, α MC And the corresponding switch position U MC and U AF and switch position change ΔU MC , ΔU AF The first angle is always zero, and the last angle is always the window α win The switching angle can be calculated by subtracting the window α win The starting angle α start (Also denoted as α st , see below) and note that the normalization is performed to account for the span from 360° to 0°. MC Synchronize each filter stage angle α AF Generates a flag that can be used to reduce the combined voltage step by switching with double steps in opposite directions.

[0101] To achieve cell balancing of the filter cell 30 , an active filter (AF) cell voltage balancer 58 and subsequent adjustment blocks 60 , 62 , and 64 of the mode determination portion 44 are operated.

[0102] In the AF balancer 58, the cell capacitor voltages V AFC,abc The cell capacitor voltage V is measured (ie, each phase is measured individually) and optionally filtered using a low pass (moving average) filter 59. AFC,abc The difference obtained by subtracting the predetermined unit reference capacitor voltage is processed using a P controller or a PI controller. The output signal can be interpreted as a non-zero differential voltage reference U bal,abc , the non-zero differential voltage reference U bal,abc needs to be injected by the (active) filter unit 30 in order to accumulate charge in the corresponding cell capacitor 38 and thus charge or discharge the cell capacitor 38. AFC,abc Matching with the reference voltage, the unit pulse mode OPP is generated due to the OPP AFabc The fundamental component of is zero, so the filter unit 30 is in a calculation window α win The average voltage generated on the MOSFET should remain unchanged. The fundamental component power is provided by the main stage 12 only.

[0103] In the AF balancer 58, it is necessary to adjust the phase current I abc The signs of the differential voltage commands for the filter unit 30 are interpreted so that the differential voltage command for the filter unit 30 drives the unit voltage in the correct direction, that is, the PI controller output is consistent with the corresponding phase current I abc To compensate for delays in the current measurement and avoid imposing an erroneous sign on the calculation cycle, the fundamental current is multiplied with respect to the flux reference angle The phase angle can be used as a reference for the current sign, so that in the upcoming calculation window α win Apply the correct sign.

[0104] In blocks 60, 62 and 64 described in more detail below, the total or combined OPP for the main stage 12 and filter stage 14 of the electrical converter 10 is then updated to the filter stage angle α AF The resulting total (or resolved) adjusted optimized pulse pattern 66 includes or consists of switching states, changes in switching states and switching angles for the three phases of the main stage 12 and the filter stage 14 of the electrical converter 10 .

[0105] In particular, in blocks 60, 62 and 64, the injected balancing voltage for the filter unit 30 must be compensated by a corresponding adjustment of the main stage 12, so that the sum of the main and filter stage (three-phase) voltages remains unchanged and does not affect the flux trajectory or torque control. MCabc It is possible to have a very low pulse count, so that usually only two of the three phases are switched. To ensure that there are always at least two phases available, the calculation window α is set accordingly. win The modulation index of the related shift α off In order to properly compensate the voltage injected by the filter unit 30, a differential balanced signal must be distributed to the main phase being switched. Then, the synchronization angle α of each active filter unit in the active filter unit 30 must be considered. AF,abc The expected volt-second offset of each active filter unit phase is calculated based on the possible pre-offset of the active filter unit phase. The result is the expected volt-second offset for the active filter unit mode OPP AFabc The total effective volt-second offset reference for each active filter unit mode in is then updated using the modified angles for the originally generated total 3-phase mode OPP of the main stage 12 and the unit stage 14 abc .

[0106] Optionally, a neutral point (NP) potential balancer 68 may be operated in addition to the active cell balancer 58. In the NP balancer 68, the upper DC link voltage and the lower DC link voltage V measured at the DC link 20 in the primary stage 12 of the electrical converter 10 are calculated. DC+ / - The calculated difference is optionally processed using a low pass (moving average) filter 69. The difference compared to a predetermined reference NP voltage (nominal zero in this example) is then processed using a P controller or a PI controller and the resulting NP balance control signal 70 is fed to blocks 60, 62 and 64 for maintaining the nominal pulse mode OPP. MCabc and OPP AFabcFor example, the NP balance control signal 70 can be configured so that the pulse mode OPP of the main stage 12 is used. MCabc The MC pulse in the compensation of NP error. For example, NP error can be added to the main mode OPP as a common mode component MCabc , but it can also be added by offsetting the angle in other ways. This NP error must usually be added over 120° or 180°.

[0107] In box 72, the calculation window α is then calculated based on the modified angle according to the overall adjusted optimized pulse pattern 66. win The new (updated) reference flux trajectory Ψ on ref Fortunately, the OPP calculated offline base Typically a nominal flux trajectory is included which is also calculated offline. Therefore, as described in more detail further below, the initial flux trajectory angle and the α-β component Ψ can be easily nominal,αβ is interpolated and used as a starting point for accurately calculating the reference flux trajectory Ψ using a simple piecewise integration of the total adjusted optimized pulse pattern 66 ref The newly generated reference flux trajectory corner point is composed of the unique angle in the α-β plane plus the calculation window α win The flux α-β components on the surface are composed of. When the NP balance is also considered, the nominal flux trajectory Ψ nominal is used as the initial condition only at each second or third update. Otherwise, the endpoints of the previously calculated flux trajectory segment are used as the next calculation window α win This ensures that the NP balance adjustment occurs at 120° or 180°.

[0108] The updated 3-phase pulse pattern 66 and the newly generated reference flux trajectory Ψ ref are stored as a lookup table. Therefore, they are available to the model predictive controller 46 for the next calculation window α w i n .

[0109] The mode determination controller 44 can also calculate the next two windows α win The total adjusted optimized pulse pattern 66 and flux reference trajectory 76 are calculated and stored in, for example, a lookup table. Because the method can be run in a lower priority interrupt and can be used when the model predictive controller 46 is in the previously calculated active calculation window α win While working on the above, the next calculation window α win An adjusted optimized pulse pattern 66 is calculated so that this can simplify implementation on a microcontroller.

[0110] In the model predictive control part 46 of the method, the core 47 of the model predictive controller 46 executes the adjusted optimized pulse pattern 66 and controls the stator flux to follow the reference flux trajectory Ψ ref To this end, the core 47 reads the adjusted optimized pulse pattern 66 and the reference flux trajectory Ψ from the look-up table ref , and receives an estimated flux 74 , which is calculated based on measurements in the electrical converter 10 .

[0111] The model predictive control part as well as the model predictive control may be performed as in EP 2 469 692 A1 incorporated by reference.

[0112] Generally speaking, in the calculation window α win During this period, the tasks and / or method steps of the model predictive control part are executed several times.

[0113] In the model predictive control part 46, the adjusted optimized pulse pattern 66 received from the mode determination controller 44 is further modified by moving at least one transition time and / or time point of the switching instant so that the estimated flux 74 of the electric converter 10 and the reference flux trajectory Ψ ref The flux error determined by the difference between is minimized. For this purpose, an objective function is used, into which the transition times and / or the time points of the switching instants are input.

[0114] Then, at least the next switching instant from this modified pulse pattern is applied to the electrical converter 10. This means that the switches of the main stage 12 and the filter unit stage 14 are switched to the switching positions as indicated by the next switching instant at the transition time of the next switching instant.

[0115] In the presence of a torque step or when the stator flux reference angle is within the pre-calculated calculation window α win In case of any other disturbance outside the corresponding angular range, the flux reference trajectory Ψ is skipped ref And instead a circular flux reference trajectory at the stator flux reference angle is used.

[0116] The following is Figure 2 Possible embodiments of the frame are described in more detail.

[0117] Filter unit balancer 58

[0118] The active filter (AF) cell voltage balancer 58 (also denoted as AF balancer) can be implemented as a P controller or a PI controller. It controls the capacitor or DC link voltage V AFC,x and the reference capacitor (or DC link) voltage V refA low pass filter 59 or a moving average filter can be used to filter the capacitor or DC link voltage V over half a fundamental cycle. AFC,x The output of the filter unit balancer 58 is multiplied by the sign of the current I in the corresponding phase x:

[0119] U bal，AF，x =G PI (s)·(V ref -G LP (s)·V AFC，x )·sign(I x )

[0120] Balance voltage U bal Or average output voltage reference U bal will be injected by the filter unit 30 as the original optimized pulse pattern OPP base deviation so that the cell capacitor or DC link voltage V AFC,x Control to the unit reference voltage V ref .

[0121] like Figure 3a to Figure 3c As shown, it is also necessary to calculate the entire upcoming window α win The current in the upper phase I x Explain the symbols carefully. Figure 3a shows the fundamental current I x , its instantaneous symbol sign(I x ) and the offset sign function sign(I adv ). Figure 3b The angle θ of the offset sign function is shown over several fundamental cycles. sign，adv and the instantaneous symbol sign(I x ). From these diagrams, it can be appreciated that it is easy for the sign of the current I to change immediately after sampling. This may result in win The upper has a balanced signal U bal Error symbol.

[0122] To avoid this, you can use Figure 3c The estimated displacement angle φ of the current I defined in ph To calculate Figure 3b The angle θ sign，adv The sign of the current I can be used to estimate the displacement angle φ ph Sufficient filtering is performed to provide a smooth reference. This can be easily accomplished, for example, by an external control loop together with the flux reference angle provided by the core 47 of the model predictive controller 46 Provided together. The displacement angle φ of the current I phIt is defined relative to the fundamental voltage angle, which is assumed to lead the flux reference by π / 2. Figure 3c In , both the fundamental current I and the fundamental voltage V are expressed as vector quantities with two vector components in an orthogonal reference frame (α-β plane).

[0123] In addition, the sign of the current I is based on a sine function rather than a cosine function, so it must also be advanced by π / 2. adv It can then be used to shift the estimated angle forward, for example by calculating the window α win half of the value to ensure that the symbol is win At least half of it is correct.

[0124]

[0125] If we ensure that 0≤θ sign，adv ≤2π, we can simply compare the current I of each phase in the following way adv Angle:

[0126]

[0127] Other methods, such as vector rotations, may also be used, but these methods may be more computationally expensive.

[0128] Trigger signal generator 52

[0129] The purpose of the trigger signal generation block 52 is to determine the flux reference angle Is the window α still being calculated? win and if not, triggers a recalculation of the mode in the mode determination section 44. One way to do this is to calculate a reference angle containing the current flux The calculation window α win The starting angle α st , and check whether it is equal to the current calculation window α win The starting angle α st For example, we can consider three inputs to perform this calculation: the calculation window width ( α w ), the delay until the newly generated pattern is available (α Ts ), and depends on the modulation index M i The offset α off The offset angle dependence is taken into account to ensure that at least two phases are win Because in order to fully compensate for the balanced volt-seconds injected into the three filter unit phases, at least two phases of the main stage 12 must be switched in any calculation window α win So this is necessary.

[0130] Figure 4 The modulation index M of the main stage 12 is shown as 1. i Graph of the switching angle over the entire range of . It can be seen that the minimum calculation window α win Must be at least 60° wide and according to the modulation index M i , we can choose two possible offsets α off The modulation index M can be used i The hysteresis function is used to determine the calculation window α win The deviation angle α off To avoid changing α too frequently off , a larger overlap of the offset regions can be used as shown. Once α off Is determined, the calculation window α can be calculated using the following function win The starting angle α st :

[0131]

[0132] where α w is the calculation window α win Width. Angle α Ts represents the microcontroller sampling time T due to the pattern generation task s The delay caused by this delay depends on the fundamental frequency f fund :

[0133] α Ts =2πf fund T s

[0134] In the case of microcontroller delays and cycle times, the sampling angle α Ts Can be used to plan upcoming sampling periods to determine the flux reference angle Will the window α still be calculated? win If the starting angle α st If the starting angle is different from the current one, the mode needs to be updated. Once the new starting angle α is determined st , the following function provides a flag to indicate whether an update is needed:

[0135]

[0136] Here, the value 1 means that the trigger signal 50 generated in block 52 is true, which triggers updating of the pulse pattern by the pattern determination portion 44 .

[0137] Multiphase Pattern Generator 54

[0138] The multi-phase mode generator 54 reads, for example, the overall single-phase mode OPP currently selected by the MP3C core 47base ,like Figure 5a As shown (FoppA). According to the total single-phase mode OPP base , the generator 54 compares the total single-phase mode OPP base The switching angles of the switching states defined in the calculation window limits are used to generate the total optimized pulse pattern OPP for each phase A, B, C of the intermediate voltage at the output 40. total,abc :For each phase, (α st , α st +α w ),like Figure 5b As shown; and the selected mode segment is stored in the array OPP total,abc In other words, the obtained three-phase mode OPP total,abc All three phases A, B, C are included, wherein for each transition (ie, switching instant), the following data are stored in the corresponding phase: transition angle α i , the final position U at the end of the transition i , the position change at the transition point ΔU i Here, i is an integer indicating the number of transitions in the corresponding pulse pattern sequence.

[0139] Care must be taken that the angles for each phase are in the correct order, with the first entry being in the calculation window α win The beginning of the window α win This also applies when spanning 360° / 0°, so that higher angles are first in the OPP base Then, the three-phase total optimized pulse mode OPP total,abc It is passed to the next functional block of the mode determination section 44 .

[0140] Main-level and unit-level pattern parsers 56

[0141] In the analysis block 56, the total optimized pulse pattern OPP total,abc Resolved to the main pulse mode OPP for the main stage 12 MCabc (also denoted MC or main converter) and a cell pulse pattern OPP for the filter cell stage 14 AFabc (also denoted AF or active filter). The filter cell voltage balancer 58 determines the volt-second injection required by each active filter cell 30 to balance its capacitor voltages. At the same time, the main stage must compensate for this volt-second injection so that the sum of the MC+AF voltages does not change. To do this, the overall optimized pulse pattern OPP must be resolved total,abc The switching times of each phase of are shifted individually:

[0142] OPP total，abc = > OPP MC，abc, OPP AF，abc

[0143] Then, by subtracting α from the resolved mode angle st And note that the span from 360° to 0° is interpreted, normalizing the switch angle to the calculation window α win In this way, the first angle is always zero, and the last angle is always the calculation window α win The length α w .

[0144] Total pulse mode OPP total,abc The switch position U can be adjusted according to Figure 6 The table shown is decoded or parsed to correspond to the main pulse pattern OPP MC,abc and unit pulse mode OPP AF,abc Location of UMC and UAF:

[0145] α st The initial switch position U0 at is determined by the following formula:

[0146] U 0 =U 1 -ΔU 1 ,

[0147] Among them U i = [-4, ..., +4] is the position, and ΔU i =[-1, +1] is the angle α i The final position U end is the calculation window α win The value of the last position in the pattern (that is, U end (α end ), where α end =lastα i <α st +α w ).

[0148] To identify a main-stage transition, the main-stage position U must be monitored when the previous position is different from the current position. MC In this case, for the main pulse mode OPP MC,abc ΔU i With OPP for total pulse mode total,abc ΔU i If so, it can be used with the main level transition U MC Synchronous cell-level transition U AF Generate flag. Applicable to unit pulse mode OPP AF,abc ΔU i The direction is the same as for the main pulse mode OPP MC,abc ΔU iThe direction is opposite, and ΔU i The nominal step size is doubled to minimize the voltage step size of the combined transition. The cell-level behavior for the synchronized transition can be considered in two different ways: 1) as a single transition where the step size is multiplied by 2; or 2) as two single-step transitions at the same angle. Storing the transitions as single steps of double size will reduce the storage requirements.

[0149] For normal asynchronous unit-level transition U AF ΔU i With OPP for total pulse mode total,abc ΔU i same.

[0150] It is also necessary to track the OPP mode that is not in the main pulse mode MC,abc The phases switched in are used for use in the allocation of volt-second compensation.

[0151] Main and unit level pre-shifters 60

[0152] Block 60 is configured to ensure that the balancing process does not change the overall average voltage (i.e., the calculation window α win = the average sum of the output voltage of the main stage 12 and the output voltage of the filter unit stage 14 on the α-axis), in particular as caused by the filter unit balancer 58. This ensures that the flux trajectory is within the calculation window α. win At the end it returns to the original reference. The most efficient way to execute the algorithm is to first offset the main level transitions and the synchronized cell level transitions (if any) and only then proceed to offset the cell level transitions to balance the filter cell voltages.

[0153] As is known from the filter unit balancer 58, the balancing voltage U t to be injected into the corresponding filter unit 30 needs to be subtracted from the main voltage. bal , but only the differential mode (DM) component is subtracted. This can be calculated using the standard Clarke transform:

[0154] U dm，αβ =U bal，AF，αβ =T abc U bal，AF，abc

[0155] in

[0156]

[0157] Since the pulse count of the main stage 12 is very low, typically only two of the three phases will be switching, so a method of allocating the total volt-second compensation to the active phases is needed. In order to only briefly describe the method, the derivation of the transformation will be omitted, but the transformation is shown here for completeness. In the case where all three phases are switching, which can occur at higher pulse counts, the standard inverse Clarke transformation can be used:

[0158]

[0159] In the case where one of the phases is not switching, the following three transformations can be used where appropriate:

[0160]

[0161] Switching phase A, B or C in the change of sign Therefore, according to the required transformation, the main mode OPP can be calculated MC,abc Total offset required:

[0162]

[0163] And the resulting total angular offset is as follows:

[0164] Δα MC，x =-U dm，MC，x

[0165] For each phase x=a, b, c of the main stage 12, the migration algorithm for unit-level migration described in the next section is used. It is important to perform the main-level mode migration separately so that the full range of available migration can be used to maximize the possibility of achieving the desired compensation. The limitations are as follows: adjacent main-level transitions, minimum pulse width and calculation window α win The beginning and end of.

[0166] In addition to applying a transition offset to compensate for the cell level balance offset, the main stage transition offset can also be applied to balance the neutral point (NP) potential of the main stage 12 according to the control signal 70 of the neutral point potential balancer 68. At this time, this can be easily done in the algorithm by adding an additional offset to Δα MC，x to complete.

[0167] Once the necessary and achievable main level offsets are completed, any synchronization angles of the cell level 14 must be offset by the same amount. This is done by copying the modified switching angles of the main level 12 to the synchronization switching angles of the cell level 14 and maintaining the order of transitions (i.e., switching moments) so that the final offset algorithm for the cell level angles can ensure that all cell level limitations are accounted for. Then, when calculating the total volt-second offset for each phase of the cell level 14, the pre-shifted synchronization cell level angles must be taken into account.

[0168] First, the total nominal volt-second offset of the active filter 30 for each phase (x=a, b, c) is calculated as:

[0169] Δα AFN，x =U bal，AF，x / r AF

[0170] where r AF is the ratio of the voltage value of the DC link of the filter unit 30 to half the voltage value of the DC link of the main stage 12 of the same phase. Each synchronization angle is then taken into account by summing the offsets of each main stage angle and multiplying the result by 2. The reasons for this are described in the previous section and the combined voltage steps must be reduced by switching the filter unit 30 in double steps in the opposite direction to the main stage 12. This results in a total effective offset per phase of the filter stage 14:

[0171]

[0172] Filter unit average voltage regulator 62

[0173] The filter unit average voltage regulator 62 is one of the key parts of the algorithm of the mode determination part 44 and is used for the main stage 12 as described above and as explained here for the filter unit stage 14. Ideally, because the load power should be provided only by the main stage 12, the average power injected by the filter unit 30 over the fundamental cycle should be zero. Otherwise, since there is no external power supply, the filter unit 30 will charge or discharge. base The optimization is performed so that the average filter unit output fundamental voltage is zero.

[0174] Therefore, a filter unit average voltage regulator 62 has been introduced to calculate the average voltage of the filter unit within the calculation window α. win The deviation from the average output voltage of the original mode is adjusted to be equal to the voltage command of the filter unit balancer 58 or the average output voltage reference U bal This is done by moving the unit in pulse mode OPP AF,abc The switching angle is adjusted until the desired average voltage change is achieved.

[0175] FIG7 shows an example set of pulse patterns for each phase A, B and C of the main stage 12 (also denoted as MC or 3L(A)NPC, see above), the cell stage 14 (also denoted as AF, see above) and the combination of the main stage 12 and the cell stage 14 (3L(A)NPC+AF). Each pattern shows the original pattern (solid line) and the pulse pattern for balancing the filter cell voltages and compensating for the introduced volt error in the calculation window α. win These adjustments or modifications to the original model are explained in more detail below:

[0176] To decrease the average voltage, positive voltage steps are shifted to the right (delayed), while negative voltage steps are shifted to the left (early). To increase the average voltage, the corresponding pulses are shifted in opposite directions.

[0177] Several strategies for shifting the pulses are possible. One approach aims to shift all available pulses by the same angular shift to achieve a total shift. In order to distribute the total required shift equally to the available (movable) voltage transitions N tr , the angular offset that needs to be applied to each switching transition can be calculated using the following equation:

[0178]

[0179] Apply Δα to each switching transition AF，x，i Ensures that the average output voltage of the filter unit is consistent with the original mode OPP AF,abc The deviation is equal to the voltage command or average output voltage reference of the modified filter unit balancer 58. Constraints can also constrain how much a pulse can be moved, for example, calculating the window α win The ends of the phase shifters, adjacent transitions, etc., or some switching transitions should not be moved at all, for example if they are synchronized with the switching of the main stage 12. Furthermore, the total required angle shift can be distributed to the available pulses in different ways (for example, trying to find the total shift by moving the first pulse as much as possible) or trying to find an adjusted average by executing the algorithm several times. The final result of the adjustment is a modified three-phase mode OPP for the filter unit stage 14 AF,abc,mod or a modified three-phase mode OPP for the main stage 12 MC,abc,mod If the required total volt-second adjustment cannot be made, a transient may occur. If the resulting error in the adjustment is small enough, it can usually be ignored. However, in cases where the resulting error is significant, alternative approaches may be taken, such as simply skipping the adjustment if it cannot be fully applied. Calculation window α win has been chosen to be short enough so that the filter unit capacitor voltage is within the calculation window α win Therefore, if one balancing period is missed, the impact is negligible as far as the active filter voltage control is concerned.

[0180] Total Mode Updater 64

[0181] In the overall mode updater 64, the adjusted three-phase mode OPP for the filter unit stage 14 is updated according to the adjusted three-phase mode OPP for the filter unit stage 14. AF,abc,mod provided in and if necessary also in the adjusted three-phase mode OPP for the main stage 12 MC,abc,mod The modified (re)calculation of the adjusted optimized pulse pattern 66 (in this example, the overall pattern OPP total,abc,mod ).like Figure 2 As shown, the overall mode updater 64 receives the adjusted mode OPP of the filter unit stage 14 from the filter unit average voltage adjuster 62. AF,abc,modand optionally also receives the adjusted mode OPP of the main stage 12 MC,abc,mod .

[0182] Flux Trajectory Generator 72

[0183] In the final step of the mode determination section 44, in box 72, the next calculation window α is used to calculate the win The previously adjusted optimized pulse pattern 66 (in this example the combined (ie, total) pulse pattern OPP for the main stage 12 and the unit stage 14) total,abc,mod ) Pre-calculate the reference flux trajectory Ψ ref This step is Figure 8a and Figure 8b For illustration purposes, Figure 8a The updated (ie, modified or adjusted) reference flux trajectory Ψ calculated in box 72 is shown in an exaggerated manner compared to the original or nominal reference flux trajectory 76 used in the model predictive control portion 46 to control the flux ref of flux ripple. Figure 8b Also shown in an enlarged manner is an updated reference flux trajectory Ψ compared to the original or nominal reference flux trajectory 76 ref The amplitude of .

[0184] As already mentioned, this step is required for the following reasons.

[0185] If the original flux trajectory 76 (e.g., a circular flux trajectory) is used in the model predictive control portion 46 of the present method, the model predictive control portion 46 will shift the pulse pattern OPP provided by the mode determination portion 44 by total,abc,mod The pulses are adjusted to be as close as possible to the original flux reference 76. Therefore, the previously adjusted pulse mode OPP total,abc,mod may be overwritten and therefore not applied as pre-calculated. As a result, modifications introduced to balance the filter cell voltages cannot be applied in a proper way and the cell voltages may drift and cannot be controlled.

[0186] Therefore, it is necessary to adjust the optimized pulse mode OPP according to the previous total,abc,mod (See Figure 8a to Figure 8b ) pre-calculates an updated reference flux trajectory Ψ including the ripple component ref Fortunately, since the proposed method is based on the pre-computed pulse pattern OPP base , so the nominal or original reference flux trajectory 76 can also be pre-calculated offline and loaded into the lookup table. Based on the pre-calculated original reference flux trajectory 76, the initial conditions can be interpolated to the calculation window α win The starting angle of .

[0187] Due to the average output volt-second offset U of the filter unit bal and U in the primary reference volt-second bal Perform corresponding subtraction (compensation) and calculate the window α win The sum of the average main and filter unit output volt-seconds remains constant and returns to the original total mode (with respect to the differential mode voltage and / or in the orthogonal fixed α-β reference frame). Therefore, in the calculation window α win Afterwards, the flux reference trajectory Ψ ref Finally, the nominal flux trajectory reference 76 is reached again, as Figure 8a to Figure 8b shown.

[0188] Therefore, the window α can be calculated using win A simple forward integration over ref This can be easily done using the nominal flux trajectory 76, which has been generated offline, centered at zero and scaled to unit magnitude. First, by taking the nominal flux trajectory 76 (also denoted as Ψ nominal,αβ ) interpolated to the starting angle α st , we can determine the starting point α st :

[0189] Ψ ref，αβ (α 0 )=Ψ nominal，αβ (α st )

[0190] In the case of neutral point (NP) potential balance (see Figure 2 The last corner point of the flux trajectory from the previous calculation window can be used as the new initial point. Depending on whether the NP balance must be completed over 120° or 180°, this is done in one or two consecutive calculation windows α. win Then, on a second or third iteration, the nominal flux trajectory 76 is again used as the initial point.

[0191] Next, the previously adjusted overall pulse pattern OPP obtained from block 64 total,abc,mod is merged into a single 3-phase switching angle containing each unq,abc The following three phases tot,abc An array of switch positions, where j next is the calculation window α win The index of the next unique angle in end is the calculation window α win The indices of the corners at the ends of:

[0192] α 0 =α st , ..., α j =αunq，abc (j next ), ..., α jend =α st +α w

[0193] Then, for each α j , three-phase switching function U tot,abc will be transformed to the stationary orthogonal (α-β) reference frame:

[0194]

[0195] For the operation of the MP3C core 47, it may be important to scale the fundamental component of the reference flux to unity. To achieve this, the nominal switching function U tot,nom,fund According to the predefined constraints of the offline generated OPP and the fact that the angle refers to the flux trajectory in the α-β plane, the following relationship can be found:

[0196] U tot，nom，fund =-M i sin(θ)

[0197] When this expression is integrated with respect to the angle θ, the fundamental flux is obtained as follows:

[0198] Ψ ref，nom，fund =∫-M i sin(θ)dθ=M i cos(θ)

[0199] It can be seen that the amplitude is the same as the voltage fundamental wave: the modulation index M i Therefore, in order to maintain the average flux reference ratio of 1, the newly calculated flux trajectory must be scaled by 1 / M i Modulation index M i can also be stored in a lookup table to avoid additional calculations. Finally, the switching angle instant α is calculated using the following cumulative sum j The flux trajectory corner at:

[0200]

[0201] Then, for the next calculation window α win , flux trajectory (α j ,Ψ ref,αβ ) together with the modified corner points and angles of the 3-phase mode OPP total,abc,mod Together, they are supplied to the MP3C core 47 of the model predictive controller 46 .

[0202] Although the invention has been illustrated and described in detail in the drawings and the foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments will be understood and implemented by a person skilled in the art and practicing the claimed invention upon study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or controller or other unit may perform the functions of several items recited in the claims. The fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. A method for controlling an electrical converter (10), the electrical converter (10) comprising: a main stage (12) adapted to convert a DC voltage into an intermediate voltage comprising at least two voltage levels; and a filter cell stage (14) having a filter cell (30) for each phase of the main stage (12), each filter cell (30) being adapted to add a cell voltage of the filter cell (30) to the intermediate voltage or to subtract the cell voltage from the intermediate voltage, The method comprises a mode determination section (44), the mode determination section (44) comprising the following steps: Select and read the optimized pulse pattern (OPP) calculated offline from the database base ), where the selected optimized pulse mode (OPP base ) includes a predetermined width (α w ) of the next calculation window (α win ) on the main stage (12) and the filter unit (30); By moving the optimized pulse pattern (OPP base ) to adjust the optimized pulse mode (OPP base ), so that the average output voltage generated in the filter unit (30) is directed toward the average output voltage reference (U baI,abc ) offsets and compensating these adjustments by corresponding modifications of the switching moments of the main stage (12), the average output voltage reference being determined from measurements in the filter unit (30); And the method includes a model predictive control part (46), the model predictive control part (46) includes in the next calculation window (α win ) is performed several times during the following steps: The adjusted optimized pulse pattern (66) is modified by shifting at least one transition time of the switching instants so that the estimated flux (74) of the electrical converter (10) and the reference flux trajectory (Ψ ref ) is minimized; At least a next switching instant is applied to the electrical converter (10) according to the modified adjusted pulse pattern.

2. The method according to claim 1, The average output voltage reference (U baI,abc ) is based on the measured capacitor voltage signal (V AFC,abc ) and the difference between the reference capacitor voltage of the filter unit (30).

3. The method according to claim 1 or 2, wherein the mode determination section (44) further comprises: The next calculation window (α) is determined according to the adjusted optimized pulse mode (66). win ) on the reference flux trajectory (Ψ ref ).

4. The method according to any one of the preceding claims, wherein the primary stage (12) of the electrical converter (10) comprises a DC link (20); and The method further comprises: The optimized pulse pattern in the pattern determination part (44) of the method is adjusted by shifting the switching moments of the optimized pulse pattern so that the measured voltage difference (V DC+ / - ) is offset to the reference neutral point voltage.

5. The method according to any one of the preceding claims, The selected optimized pulse mode (OPP base ) is a single-phase pulse mode; and The mode determination part (44) of the method comprises: According to the single-phase optimized pulse mode (OPP base ) generates a multi-phase optimized pulse pattern and performs the adjustments for each phase individually.

6. The method according to any one of the preceding claims, The selected optimized pulse mode (OPP base ) is an overall optimized pulse pattern comprising accumulated switching moments for the main stage (12) and the filter unit (30); and The mode determination part (44) of the method further comprises: Before making the adjustments, each optimized pulse pattern (OPP base ) is parsed into a main pulse pattern for the main stage (12) and a unit pulse pattern for the filter unit (30) of the corresponding phase.

7. The method according to any one of the preceding claims, further comprising: Determining the Flux Reference Angle Whether in the calculation window (α win )Inside; If the flux reference angle Leave the calculation window (α win ), a trigger signal (50) is generated; and The mode determination part (44) of the method is triggered by the trigger signal (50).

8. The method according to any one of the preceding claims, The calculation window (α win ) is relative to the optimized pulse mode (OPP base )’s fixed angle range; The calculation window (α win ) is a time period corresponding to at least 1 / 6 of the fundamental period of the reference output voltage of the electrical converter (10).

9. The method according to any one of the preceding claims, wherein the steps in the mode determination section are performed by a mode determination controller (44); and The steps in the model predictive control section are executed by a model predictive controller (46), and the model predictive controller (46) has a faster execution speed than the mode determination controller (44).

10. The method according to any one of the preceding claims, When the fundamental flux reference is located at the reference flux trajectory (Ψ ref ) when the angle range provided by the electric converter (10) is exceeded, the flux error is determined based on the difference between the estimated flux of the electric converter (10) and the circular flux trajectory.

11. The method according to any one of the preceding claims, wherein the adjusted optimized pulse mode (66) and the flux reference trajectory (Ψ ref ) is stored in a look-up table during the mode determination portion (44).

12. The method according to any one of the preceding claims, For the next two calculation windows (α win ), the adjusted optimized pulse mode (66) and the flux reference trajectory (Ψ ref ) is calculated during the mode determination portion (44) and stored in a lookup table.

13. A computer program for controlling an electric converter (10), said computer program being adapted to carry out the steps of the method according to one of the preceding claims when executed by a processor.

14. A computer-readable medium having stored therein the computer program according to claim 13.

15. A controller (42) for controlling an electric converter, the controller (42) being adapted to perform the method according to one of claims 1 to 12, the controller comprising: a mode determination controller (44) for executing said steps in said mode determination part of said method; A model predictive controller (46) is used to perform the steps in the model predictive control part of the method.

16. An electrical converter (10), comprising: A main stage (12) adapted to convert a DC voltage into an intermediate voltage comprising at least two voltage levels; a filter cell stage (14) having a filter cell (30) for each phase of the main stage (12), wherein each filter cell (30) is adapted to add a cell voltage of the filter cell to the intermediate voltage or to subtract the cell voltage from the intermediate voltage; A controller (42) according to claim 15.

Images