Distortion spectrum control through space vector modulation

Abstract

Method for controlling a distortion spectrum (520, 610) generated by space vector modulation in a switch-based inverter (306) that modulates a plurality of output variables, characterized in that, for each subsequent switching operation, various switch positions are determined which cause the modulation of the plurality of output variables (510) that approximately follow a given reference signal (112, 122, 208, 310), and a respective distortion spectrum (520, 610) of deviations between the reference signal (112, 122, 208, 310) and the plurality of output variables (510) that would result from each realization of these different switch positions is calculated, wherein a modulator (114, 124) is used for modulating the plurality of output variables of the inverter (306), by means of the modulator (114,124) within a space vector diagram of the space vector modulation, a predetermined number of vectors (418) are determined which are in the closest vicinity (210, 212, 214) of a vector (208) of the reference signal to be located within the space vector diagram of the space vector modulation, wherein each vector of the space vector modulation corresponds to a switch position, and the switch position is selected and implemented in the inverter (306) for which the associated distortion spectrum (520, 610) is most suitable according to predetermined properties.

Description

[0001] The present invention relates to a method and a system for controlling switching distortions in an inverter by means of space vector modulation, in particular when the inverter is used in an electric vehicle.

[0002] In electric vehicles, an inverter, a power electronics unit generally consisting of several semiconductor switches, is used to control an electric motor. An output variable, i.e. output voltage or output current, which is usually modulated sinusoidally by the switches and is also provided by the inverter in several phases, e.g., to operate a three-phase drive motor, is primarily generated using pulse width modulation. In order to replicate a sinusoidal shape as accurately as possible, a large number of switching cycles are necessary. In a frequency spectrum, these lead to so-called switching harmonics, i.e. fundamental oscillations and their overtones caused by switching frequencies, which manifest themselves as spectral peaks with an associated high energy density.

[0003] An introductory overview and a method for influencing the frequency spectrum is provided by C. Tao, AA Fayed, “PWM control architecture with constant cycle frequency hopping and phase chopping for spur-free operation in buck regulators”, in IEEE Transactions on very large scale integration (VLSI) systems, 21(9), 1596-1607 (2013).

[0004] The applicant discloses a method from DE 10 2016 106 472 A1 that compares the advantages of an otherwise high switching frequency with the disadvantages of this approach. Switching distortions and switching losses are thus evaluated against each other and attempted to be minimized. However, the disclosed method is a binary control system that can only be used to control a single-phase output variable. The possibilities of space vector modulation for controlling multilevel systems cannot be utilized in this case.

[0005] The paper by Kartick Chandra Jana and Sujit K. Biswas, "Generalized switching scheme for a space vector pulse-width modulation-based N-level inverter with reduced switching frequency and harmonics," IET Power Electron., 2015, Vol. 8, Issue 12, pages 2377-2385, presents a generalized switching scheme for space vector modulation of an inverter. Controlling the switching frequency reduces switching losses and the power content of harmonics.

[0006] The publication V. Staudt: “Raumzeiger”, Ruhr University Bochum, Energy System Technology and Power Mechatronics provides a compact insight into the process of space vector modulation.

[0007] The following publications all concern the control of switching distortions in the inverter using space vector modulation, particularly in applications in electric vehicles.

[0008] In the document DE 10 2011 088 242 A1, voltage vectors are suitably adjusted using space vector modulation in order to compensate for a dead time in the inverter without measuring phase currents and / or phase voltages.

[0009] The document DE 10 2012 210 658 A1 describes a method which prevents an excess temperature of at least one of the switches used to provide a current by space vector modulation within an inverter.

[0010] The document US 2008 / 0 252 250 A1 addresses the treatment of dead time in the inverter and discloses a method that keeps an output voltage at a defined value during the dead time.

[0011] Document US 2008 / 0 297 100 A1 deals with the reduction of inverter voltage losses when a high torque is generated in the electric motor at a low (rotational) frequency, which is particularly the case during the starting process of the electric motor.

[0012] The reduction of switching losses in the inverter when determining a rotor angle position in the electric motor is disclosed in document US 2011 / 0 012 544 A1, whereby the waveform of the pulse width modulation is modified for this purpose.

[0013] Voltage losses are also to be reduced by the method described in the document US 2012 / 0 075 892 A1, which involves monitoring the switching states in the inverter.

[0014] In the documents US 2012 / 0 139 461 A1 and US 2015 / 0 077 025 A1, the switches of an inverter are controlled by a control unit in such a way that dead time and switching distortions are reduced.

[0015] US Pat. No. 6,088,246 A proposes broadening the spectra over a predetermined frequency range in a high-frequency component of each of the inverter's output voltages. This is intended to reduce the energy density of the electromagnetic noise generated by the inverter, which can potentially also interfere with the control of the electric motor.

[0016] Im Stand der Technik sind ebenfalls Raumzeigermodulationsverfahren für Multilevelkonverter beschrieben, z.B. in W. Yao, H. Hu and Z. Lu, „Comparisons of Space-Vector Modulation and Carrier-Based Modulation of Multilevel Inverter“, in IEEE Transactions on Power Electronics, 23(1), 45-51 (2008), Jae Hyeong Seo, Chang Ho Choi and Dong Seok Hyun, „A new simplified space-vector PWM method for three-level inverters“, in IEEE Transactions on Power Electronics, 16(4), 545-550 (2001), oder auch in B. P. McGrath, D. G. Holmes and T. Lipo, „Optimized space vector switching sequences for multilevel inverters“, in IEEE Transactions on Power Electronics, 18(6), 1293-1301 (2003).

[0017] A central multilevel converter in this sense is the modular multilevel converter MMSPC, described by SM Goetz, AV Peterchev and T. Weyh, “Modular Multilevel Converter With Series and Parallel Module Connectivity: Topology and Control”, in IEEE Transactions on Power Electronics, 30(1), 203-215 (2015).

[0018] However, none of these methods is capable of controlling or regulating the distortion spectrum. Instead, all of these methods produce sharp switching harmonics in the distortion spectrum.

[0019] Against this background, it is an object of the present invention to provide a method that discloses a modulation method for the output variables of an inverter that utilizes the advantages of space vector modulation and controls the distortion spectrum. Furthermore, it is an object of the present invention to provide a corresponding system for implementing such a method.

[0020] To achieve the above-mentioned object, a method is provided for controlling a distortion spectrum which arises through space vector modulation in a switch-based inverter which modulates at least one output variable. In this case, various switch positions are determined for each switching operation which is closest in time, which effect a modulation of the at least one output variable which approximately follows a predetermined reference signal. Furthermore, a respective distortion spectrum of deviations between the reference signal and the at least one output variable which would arise from a respective implementation of these various switch positions is calculated, and the switch position is selected and implemented in the inverter at which the associated distortion spectrum is most suitable according to predetermined or predefined properties, i.e. the predetermined or predefined properties are best implemented.

[0021] In one embodiment of the method according to the invention, an output variable of the inverter corresponds to a respective phase of an alternating current and / or an alternating voltage.

[0022] In order to keep the switching losses of the inverter limited, in one embodiment of the method according to the invention, an average switching rate can be selected as at least one of the predetermined properties for selecting the most suitable switch position.

[0023] In a further embodiment, the distortion spectrum associated with the switch position to be selected, which has a spectral gap in a predetermined range, is selected as at least one of the predetermined properties. This is particularly important when systems that dynamically change their sensitive spectral range, such as car radios that perform a station sweep, are used simultaneously. A spectral gap in the desired distortion spectrum that runs parallel to the car radio's station sweep is conceivable. Likewise, standards applicable at a location where the electric vehicle is used (e.g., ISO, national standards, CISPR limit values) can lead to spectral gaps in the distortion spectrum being specified. This may require a new selection of the predetermined properties if national borders are possibly exceeded during operation of the electric vehicle.The current location of the electric vehicle can be provided from the on-board electronics via GPS information.

[0024] The specification of properties according to which the most suitable distortion spectrum according to the properties is established when the method according to the invention is carried out is a central advantage of the invention, since this avoids a method of system clock variation known from the prior art, which influences other vehicle systems and complicates control.

[0025] The method according to the invention is preferably implemented by a modulator that modulates the output of the inverter. The task of the modulator is to approximately adapt a stepwise quantized output of the inverter to the continuous, at least partially sinusoidal reference signal, while taking predetermined properties of the distortion spectrum into account. Both are achieved by numerically processing the distortion spectrum. In an extension of the method according to the invention, at least one switching signal for the inverter is generated directly by the modulator.

[0026] According to the method according to the invention, a predetermined number m of vectors is determined within a space vector diagram of the space vector modulation, which vectors are in the closest neighborhood to a vector of the reference signal to be located within the space vector diagram of the space vector modulation. For example, in a so-called three-phase two-level inverter, there are always m=3 vectors in the closest neighborhood to the vector to be assigned to the reference signal (see Fig. 2).

[0027] In an embodiment of the method according to the invention, a zero vector located in the immediate vicinity is selected such that switching losses are minimized when the switch position is changed to adjust it. Generally, there are several equivalent implementations of the zero vector in inverters, especially in bridge circuits. For example, in a two-level inverter (see Fig. 2) The zero vector is generated either by closing all so-called low-side switches or, alternatively, by closing all so-called high-side switches. To minimize switching losses, the implementation that requires the fewest number of switch changes from the previous switch position would be chosen. Advantageously, the invention also chooses the implementation of the zero vector that minimizes the current required to change the switch position.

[0028] According to the invention, for each of the m determined vectors, a respective difference to the vector of the reference signal is determined, from this respective difference and respective differences between vectors of the at least one output variable and vectors of the reference signal from a plurality of immediately preceding time steps, a temporal profile of the differences is created, and this temporal profile is numerically compared with a temporal profile designed according to the predetermined properties (see Fig. 3).

[0029] In a variant of the method according to the invention, differences that lie further back in time are given less weight, and the further back in time they lie, the less weight the more weakly they are given.

[0030] In an embodiment of the method according to the invention, the temporal course of the differences of the at least one output variable is transformed into the frequency domain and forms there the so-called distortion spectrum, with which the numerical comparison with a frequency spectrum designed according to the predetermined properties is carried out.

[0031] In a further embodiment of the method according to the invention, a filter corresponding, for example, to the inverse of a predetermined distortion spectrum is applied to the distortion spectrum obtained by the transformation into the frequency domain. If, for example, there is a spectral gap in the predetermined distortion spectrum, this region remains unchanged in the distortion spectrum obtained from the above transformation by applying the filter with the inverse of the predetermined distortion spectrum, while other spectral regions are reduced, for example.

[0032] According to the invention, a norm formed from the distortion spectrum obtained from the transformation is traced back to a quality number, with the aid of which the most suitable distortion spectrum with associated switch position, which corresponds, for example, to a smallest numerical value of the quality number, is selected and passed on to the inverter.

[0033] In an embodiment of the method according to the invention, a p-norm is formed as the norm, for example, with p=2, the square root of the sum of the squared differences. When using a 2-norm (which corresponds to a power), the value is identical for most frequency transformations due to the energy equality of the time and frequency domains according to Plancherel's theorem.

[0034] According to the invention, an overall quality number is obtained from the summation of the respective quality numbers of an output variable, for example, by forming a p-norm. This, in turn, is used to select the most suitable switch position.

[0035] In a further embodiment of the method according to the invention, the difference between the determined vectors and the vector of the reference signal is calculated exclusively in phase space. Since conversions to a time domain always involve additional computational effort, and any numerical operations lead to rounding error propagation due to finite computational accuracy, it is advantageous to limit the calculation to a single phase space representation by amplitude and phase for all outputs of a three-phase system, e.g., three phase outputs in a three-phase drive motor. This also minimizes the memory required to store the temporal profile of vectors previously obtained by executing the method according to the invention.

[0036] As above, in this embodiment, each candidate i=1,...,m to be tested consists of the previous switching sequence and one or more future steps, whereby the respective candidate is now purely in two-dimensional vector representation v K,i (t)=(A(t), φ(t)) T , with amplitude A(t) and phase φ(t) of a single space vector quantized to the points of a space vector diagram at time t. Instead of converting the deviation from a given reference signal into a voltage or current waveform at one or more phase outputs for at least one candidate and then determining the distortion spectrum, the distortion spectrum is estimated directly in the phase space. First, at a time t+t predict a two-dimensional vectorial deviation between the candidate vector v K,i and the reference vector v representing the reference signal refThis is done over a certain window with n points in time in the recent past, in which all candidates are similar or correspond to a previous switching history, or in the case of predictive methods, a certain time in the future in which the candidates differ. The reference signal usually corresponds to a continuous vector v ref (t)=a0e r +φ(t)e φ in angular coordinates, or v ref (t)=a0 cos(φ(t))e x +a0sin(φ(t))e x in Cartesian coordinates, and represents, for example, a sinusoidal waveform of phase voltages or phase currents with a certain frequency ω0, from which the phase φ(t)=ω0t results. The calculation of the deviation {[v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)]⋅e→x}2+{[v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)]⋅e→y}2, is carried out over all two dimensions of the phase space and for all times τ, and is mapped to a scalar function over time, preferably by a vector norm: ‖v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)‖.

[0037] If the norm is a 2-norm, hence a Euclidean norm, the result corresponds, as already mentioned above, to a power addition, which is preserved even after a Fourier transformation due to Plancherel's theorem.

[0038] From the scalar function (2) for the deviation of each candidate from the reference for a certain time course, the inventive method, as described above, calculates the respective associated distortion spectrum in the frequency domain using a Fourier transformation (FT), which can be implemented numerically as a fast Fourier transformation. However, in this embodiment of the inventive method, regardless of the number of output phases of the inverter, only one Fourier transformation is always performed, thereby minimizing the computational effort, which is particularly advantageous in a commonly used implementation of the method on an FPGA IC.

[0039] In addition, the respective spectrum can be weighted with a predefined filter, which ultimately amplifies the spectral profile of the resulting distortion to varying degrees according to the desired specifications or highlights it for subsequent evaluation using a quality number. For example, if the candidate v K,i with the smallest deviation given a filter F(ω), then arg mini‖F(ω)FT‖v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)‖ ‖p to calculate, where, as mentioned above, a p-norm was formed over the entire weighted distortion spectrum. Choosing a large p-value primarily involves the most prominent peak from the weighted distortion spectrum at any given time point in the further evaluation and is thus suppressed according to equation (3). A small p-value minimizes all components of the weighted distortion spectrum more evenly. Choosing p=2 approximately optimizes the performance of the weighted distortion spectrum across all frequencies.

[0040] Furthermore, when using a complex Fourier transform, an absolute value can be applied before normalization to reduce complex or negative numbers to their absolute value.

[0041] The candidate with the lowest scalar value from equation (3) generates a distortion spectrum that, according to the inventive method, is closest to the specified distortion spectrum of the filter. Accordingly, this candidate is converted into the switching signals of the semiconductors and adopted as the state for the next switching cycle(s) in the inverter.

[0042] In an alternative to this embodiment of the method according to the invention, the deviations of the candidate vector from the reference signal or from the reference vector from equation (1) before the Fourier transformation are determined over a time window consisting of a part in the past with duration w up to time t+t predict integrated: ∫0w capitaldτ({[v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)]⋅e→x}2+{[v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)]⋅e→y}2).

[0043] Using the Fourier transformation (FT) mentioned above, the weighting with a frequency-dependent filter F(ω), and a p-norm over all weighted frequency components, the calculation rule for determining the quality number for a candidate vector v is K,i : arg mini‖F(ω)FT[|∫0w bigdτ‖v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)‖|]‖p.

[0044] However, the integration leads to a low-pass filtering of the distortion, which should be compensated in the filter by making the filter approximately proportional to the inverse of the frequency.

[0045] In a further embodiment of this alternative of the method according to the invention, the time integral in the Fourier transform is weighted inversely to the frequency ω so that the filter F(ω) can continue to correspond to the desired frequency response of the distortion spectrum: arg mini‖F(ω)a+ωFT[|∫0w bigdτ‖v→K,i(t+tpredict−τ)−v→ref(t+tpredict−τ)‖|]‖

[0046] The constant a prevents the divergence of the DC components and allows an adjustment of the accuracy with which the reference signal (voltage or current specification) is represented.

[0047] Although the method according to the invention has been described repeatedly in this application using a two-level three-phase inverter as an example, it is generally not limited to this and can be used for any multilevel inverter. In particular, vectors in the space vector diagram of the respective multilevel inverter can be determined for any number of levels.

[0048] Furthermore, a system is claimed which comprises an inverter, a current and / or voltage source and a modulator controlling the inverter, wherein the system is designed to carry out the method according to the invention.

[0049] In one embodiment of the system, the modulator has at least the following modules: a nearest neighbor module, which is designed to determine, within a space vector diagram of a space vector modulation, a predetermined number of vectors that are in the closest neighborhood to a reference vector of a reference signal to be located within the space vector diagram of the space vector modulation; a history module, which is designed to form a history of reference vectors and a vector of the at least one output variable from each time step of a number of previous time steps; and a selection module, which is designed to select the most suitable vector of the at least one output variable according to predetermined properties using the nearest neighbor module and the history module.

[0050] In a further embodiment of the system, the selection module has a separate pipeline for each of the determined vectors, wherein the selection module is designed to process a plurality of pipelines in parallel.

[0051] Furthermore, a modulator is claimed which is designed to control switches of an inverter using a reference signal as an input variable, and in doing so to carry out the method according to the invention, wherein the modulator has at least the following modules: a nearest neighbor module which is designed to determine, within a space vector diagram of a space vector modulation, a predetermined number of vectors which are in the closest neighborhood to a reference vector of a reference signal to be located within the space vector diagram of the space vector modulation; a history module which is designed to form a history of reference vectors and a vector of the at least one output variable from each time step of a number of previous time steps;and a selection module configured to select the most suitable vector of the at least one output variable according to predetermined properties using the nearest neighbor module and the history module;

[0052] Finally, a modulator is claimed that controls an inverter whose power electronics use a wide-bandgap semiconductor. For example, the method according to the invention can be implemented using a modulator for controlling an inverter that is operated with a gallium nitride semiconductor and has voltages in the range of 100 V, currents in the range of 30 A, and switching rates in the range of 500 kHz.

[0053] Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawings. Fig. Figure 1 shows a schematic representation of two possible tasks of a modulator controlling an inverter. Fig. 2 shows a space vector diagram for selecting nearest neighbors to a vector. Fig. 3 shows a schematic representation of a possible embodiment of a modulator according to the invention. Fig. 4 shows a schematic representation of a possible embodiment of a module of a modulator provided according to the invention for selecting a suitable vector of an output signal. Fig. 5 shows exemplary switching states and a respective associated distortion spectrum according to an embodiment of the method according to the invention. Fig. Figure 6 shows a schematic representation of an environment in which an electric vehicle moves and which contributes to a predetermination of the properties of the distortion spectrum. Fig. 7 shows a schematic representation of a further embodiment of a module of a modulator provided according to the invention for selecting a suitable vector with difference formation to the reference signal purely in the phase space. Fig. Figure 8 shows an example of a distortion spectrum in which the method according to the invention leads to a tunable spectral gap.

[0054] In Fig. 1 illustrates a schematic representation of two possible tasks of a modulator 114 and 124 controlling an inverter. On the one hand, drawing 110 represents a modulator 114 which receives, as an input signal 112, an at least partially sinusoidal reference signal, adapts this to a plurality of quantized output options of the inverter by means of space vector modulation, and whose output signals 116 consist of respectively quantized switching signals for a plurality of phases for controlling the inverter. On the other hand, drawing 120 represents an extended modulator 124 which receives, as an input signal 122, an at least partially sinusoidal reference signal, adapts this to a plurality of quantized output options of the inverter by means of space vector modulation, and whose output signals 126 consist of direct switching signals for controlling the inverter.

[0055] For the example of a two-level inverter, Fig. 2 shows a space vector diagram for the inventive selection of three vectors 210, 212, 214 of the space vector modulation, which are located, for example, in close proximity to a possible vector 208 of the reference signal. Axis 202 forms the angle 0 or 2 Pi to the right and the angle Pi to the left. Axis 204 forms the angle Pi / 2 upwards and the angle 3 Pi / 2 downwards. Vector 208 of the reference signal has an amplitude A and a phase angle φ with reference symbol 206. A respective vector 210, 212, 214, 216, 218, 220, 222 of the space vector modulation results in switch positions to be implemented directly by the inverter. The so-called zero vector 210 has a special position, in which all switches in the inverter are either open or closed. Further switch positions for the vectors 212, 214, 216, 218, 220, 222 are derived from the state of the art.According to the invention, for a vector 208 of the reference signal, a number m of next vectors in the space vector diagram is determined in each time step, which in the hexagon shown for the example of a two-level inverter consists of m=3 vectors that include the zero vector 210 and two corners of the hexagon. For a phase angle 206 of the reference signal greater than 0 and less than or equal to Pi / 3, these are 210, 212, 214; between greater than Pi / 3 and less than or equal to 2 Pi / 3, these are 210, 214, 216; between greater than 2 Pi / 3 and less than or equal to Pi, these are 210, 216, 218; between greater than Pi and less than or equal to 4 Pi / 3, these are 210, 218, 220; between greater than 4 Pi / 3 and less than or equal to 5 Pi / 3, these are 210, 220, 222; and between greater than 5 Pi / 3 and less than or equal to 2 Pi, which also corresponds to an angle of 0, these are 210, 222, 212.

[0056] Fig. Figure 3 shows a schematic representation of a possible embodiment of a modulator according to the invention. A current vector v at a time step t ref =(A(t), φ(t)) T of the reference signal 310, consisting of amplitude A(t) and phase angle φ(t), forms an input signal to a nearest neighbor module 302, which can determine, within a space vector diagram of a space vector modulation, a predetermined number m of vectors, m greater than or equal to 1, that are in the closest neighborhood to a reference vector of a reference signal to be located within the space vector diagram of the space vector modulation. The determined m vectors 312 are forwarded to a selection module 304. In addition, the current vector v ref (t)=(A(t), φ(t)) Tof the reference signal 310 is fed to a history module 308, which stores it. In addition, the history module 308 has the values ​​of the vector of the reference signal {v ref (ti)} i=1,...,n Furthermore, the history module 310 also holds the values ​​for a vector of the output variable {v out (ti)} i=1,...,n Both series of values ​​are available as a temporal history 316 of the selected output variable {v ou t(ti)} i=1,...,n and temporal history 318 of the vector of the reference signal including the current value {v re f(ti)} i=0,...,n the selection module 304. The selection module 304 then carries out a selection of the most suitable vector v out(t) for the current output variable 314 and forwards this, or the associated switch position, to a hardware 306, in particular an inverter. In addition, the selected vector v out (t) for the current output variable 314 is stored in the history module 308.

[0057] Fig. Figure 4 shows a schematic representation of a possible embodiment of a module of a modulator provided according to the invention for selecting a suitable vector of an output signal. The m input vectors 418 originate from the nearest neighbor module 302 of Fig. 3 and consist of the m determined vectors that are in close proximity to a reference vector of a reference signal to be located within the space vector diagram of the space vector modulation. For each of these m input vectors 418, there is advantageously a pipeline 404, 406, 408, 410, 412, 414, which are processed in parallel. In a respective module 404, a value of a phase voltage is available that is to be assigned to the respective input vector, which in the example of a three-phase two-level inverter represents one of the three phases. Advantageously, the other two phases are also processed in the same pipeline. Module 402 receives from the history module 308 in Fig. 3 the n values ​​from the past in time for the vector of the reference signal including the current value, as well as the n values ​​from the past in time for the vector of the output variable, converts both series of values ​​into values ​​for voltages and forwards these values ​​to a difference module 406 for difference calculation. Together with the respective current voltage value for an output variable from the respective module 404, the respective difference between the reference signal and the proposed output variable is calculated for n+1 time steps and forwarded to a weighting module 408. In the weighting module 408, which can also optionally be omitted, the differences are weighted differently. For example, it is conceivable to give differences from further back in time less weight than more recent, i.e. younger, differences.It is also conceivable to form an integral over the differences multiplied by a filter and to weight this, for example, against a sum of switching losses arising when implementing the respective vector of the output variable in the inverter. A frequency transformation, preferably a Fourier transformation, is carried out in the frequency transformation module 410. A spectral filter, which corresponds, for example, to an inverse of a desired distortion spectrum, is optionally applied to the resulting frequency spectrum in a respective filter module 412. Finally, in a respective norm module 414, a mathematical norm, for example a p-norm, is applied to the filtered result, which yields a single quality number as the result. If more than one phase was calculated, the three phases, for example, are combined into an overall quality number using a second p-norm during norm formation.In a subsequent evaluation module 416, all (total) quality numbers are available, and the original input vector 418 whose (total) quality number has the smallest value is output 420 to the hardware, or rather, the inverter. In an alternative embodiment, the spectral filter implemented in the respective filter module 412 is applied as a convolution in the time domain, i.e., without frequency transformation, in the frequency transformation module 410.

[0058] Fig. 5 shows exemplary switching states and an associated frequency spectrum according to one embodiment of the method according to the invention. The output 510 of a three-phase inverter selected as an example appears largely identical to conventional pulse-width-modulated inverters in a time domain 512. The voltage 514 is plotted in volts at the top. However, in a spectral range 520, with power density 524 plotted at the top and frequency 522 in kHz to the right, specific curves can be forced, for example an almost flatly truncated upper edge of the distortion spectrum. For example, the inverse of the distortion spectrum shown in the spectral range 520 approximately corresponds to a rectangle that covers a range from approximately 10 kHz to 500 kHz in frequency 522 and the highest peak of the spectrum in power density 524.

[0059] Fig. 6 shows a schematic representation of an environment in which an electric vehicle 602 moves, which contributes to a predetermination of the properties of the distortion spectrum 610 plotted against the frequency 612. For example, a database 606 carried in the electric vehicle 602, using location information obtained from a GPS system 608, provides specific requirements for the properties of the distortion spectrum, which result from the respective national standards. Electronic devices 604 present in the electric vehicle 602, such as radio tuners, GPS receivers, or any devices that implement any pulse modulation method, also impose requirements on the distortion spectrum 610 caused by an inverter.

[0060] Fig. Figure 7 shows a schematic representation of another embodiment of a module of a modulator provided according to the invention for selecting a suitable vector with difference formation from the reference signal purely in the phase space. The m input vectors 702 originate from the nearest-neighbor module 302 of Fig. 3 and consist of the m determined candidate vectors v K,iwhich are in close proximity to a reference vector of a reference signal to be located within the space vector diagram of the space vector modulation. For each of these m input vectors 702, there is advantageously a pipeline 706, 708, 710, 712, 714, 718, and 720, which are processed in parallel. In a respective module 706, the difference to the reference signal 704 is formed purely in phase space according to equation (1) and fed to a respective module 708, which executes a scalar function, preferably a vector norm, for each time τ from a range of the time profile according to equation (2). The respective result 710 is fed to a frequency transformation 712, which calculates the respective associated spectrum 714 in the frequency domain and weights it with a predetermined filter 716 with module 718, and finally, after execution of a p-norm in module 720, is fed to the determination of the quality number according to equation (3) in module 722.The candidate with the lowest scalar value from equation (3) generates a distortion spectrum that, according to the inventive method, is closest to the specified distortion spectrum of the filter. Accordingly, that candidate is communicated to the inverter as the next switch state 724.

[0061] Fig.8 shows, by way of example, a distortion spectrum in which the method according to the invention leads to a tunable spectral gap 808. In graph 800, the frequency 802 in Hertz is plotted to the right and the power density 804 upwards, as is the case in inset 810. The solid line 806 in both graph 800 and inset 810 indicates a spectral line at 50 Hz, which, for example, is to be avoided when applying the method according to the invention. In general, it is possible to generate spectral gaps in the distortion spectrum by applying the method according to the invention. Due to dynamic control of the spectrum, the spectral gaps 808 can also be adaptively changed, for example, their width can be changed or tuned over a wide spectral range. Modern vehicles, for example, transmit the reception frequency of the radio receiver to various control units.Accordingly, the power electronics can dynamically generate one or more spectral gaps 808 based on such information and, for example, dynamically follow a station search.

Claims

[1] Method for controlling a distortion spectrum (520, 610) which arises in a switch-based inverter (306) which modulates a plurality of output variables by space vector modulation, characterized in that for each switching operation which is next in time, different switch positions are determined which cause the modulation of the plurality of output variables (510) which approximately follow a predetermined reference signal (112, 122, 208, 310), a respective distortion spectrum (520, 610) of deviations between the reference signal (112, 122, 208, 310) and the plurality of output variables (510) which would arise from a respective realization of these different switch positions is calculated, wherein a modulator (114, 124) is used to modulate the plurality of output variables of the inverter (306), by means of the modulator (114,124) within a space vector diagram of the space vector modulation, a predetermined number of vectors (418) are determined which are located in the closest vicinity (210, 212, 214) of a vector (208) of the reference signal to be located within the space vector diagram of the space vector modulation, wherein a respective vector of the space vector modulation corresponds to a switch position, and the switch position is selected and implemented in the inverter (306) at which the associated distortion spectrum (520, 610) is most suitable according to predetermined properties. [2] A method according to claim 1, wherein the plurality of output variables are selected as a plurality of phases of an alternating current and / or an alternating voltage. [3] Method according to one of the preceding claims, in which an average switching rate is selected as at least one of the predetermined properties for selecting the respective switch positions. [4] Method according to one of the preceding claims, in which it is determined as at least one of the predetermined properties that the distortion spectrum (510, 610) associated with the switch position to be selected has a spectral gap in a predetermined range. [5] Method according to one of the preceding claims, in which a modulator (114, 124) is used as the modulator, which generates at least one switching signal (126) for the inverter (306). [6] Method according to one of the preceding claims, in which a zero vector (210) located in the immediate vicinity is selected such that a switching loss is minimized when the switch position is changed as necessary to adjust it and / or a current required to change the switch position is minimized. [7] Method according to one of the preceding claims, in which a respective difference to the vector of the reference signal (112, 122, 208, 310) is determined for the determined vectors (418), a time profile (304, 406) of the differences is created from this respective difference and respective differences between vectors of the plurality of output variables (510) and vectors of the reference signal (112, 122, 208, 310) from a plurality of immediately preceding time steps. [8] Method according to claim 7, wherein a difference formation (706) between the determined vectors (702) and the vector of the reference signal (704) takes place exclusively in the phase space. [9] Method according to claim 7 or 8, wherein the temporal course of the differences for the plurality of output variables (510) is transformed into a distortion spectrum (520, 610) in the frequency domain. [10] Method according to claim 9, wherein a filter is applied to the spectrum obtained by the transformation into the frequency domain. [11] Method according to one of the preceding claims, in which a norm formed over the distortion spectrum (520, 610) of the plurality of output variables (510) is traced back to a quality number, with the aid of which the most suitable distortion spectrum (520, 610) and thus the associated switch position are selected. [12] Method according to claim 11, wherein, for the plurality of output variables, a total quality number is obtained from the summation of the respective quality numbers of the respective corresponding plurality of output variables and with the aid of which a most suitable switch position is selected. [13] Method according to one of the preceding claims, in which a multilevel inverter is used as the inverter (306). [14] System comprising an inverter (306), a current and / or voltage source and a modulator controlling the inverter (306), the system being designed to carry out a method according to any one of the preceding claims. [15] The system according to claim 14, wherein the modulator (114, 124) comprises at least the following modules: a nearest neighbor module (302) configured to determine, within a space vector diagram of a space vector modulation, a predetermined number of vectors (312) that are in the closest neighborhood to a reference vector (208) of a reference signal to be located within the space vector diagram of the space vector modulation; a history module (308) configured to form a history of reference vectors and a vector of the plurality of output variables from each time step of a number of previous time steps; and a selection module (304) configured to select the most suitable vector of the plurality of output variables using the nearest neighbor module (302) and the history module (308). [16] System according to claim 15, wherein the selection module (304) has a separate pipeline (404, 406, 408, 410, 412, 414) for each of the determined vectors and is designed to process a plurality of pipelines (404, 406, 408, 410, 412, 414) in parallel. [17] A modulator (114, 124) designed to control switches of an inverter (306) using a reference signal (112, 122, 208, 310) as an input variable, and to carry out a method according to one of claims 1 to 13, wherein the modulator (114, 124) has at least the following modules: a nearest neighbor module (302) designed to determine, within a space vector diagram of a space vector modulation, a predetermined number of vectors (312) that are in the closest neighborhood to a reference vector (208) of a reference signal (310) to be located within the space vector diagram of the space vector modulation; a history module (308) designed to form a history of reference vectors and a vector of the plurality of output variables from each time step of a number of previous time steps;and a selection module (304) configured to select the most appropriate vector of the output signal using the nearest neighbor module (302) and the history module (308); [18] A modulator (114, 124) according to claim 17, which controls an inverter (306) whose power electronics use a wide bandgap semiconductor.

Images