Simulation methods for generating pulse patterns

Abstract

Simulation method for generating pulse patterns defined by switching angles for a synchronous PWM modulation method for controlling an inverter that powers an electric motor, wherein the simulation method comprises the following steps: a. Providing an inverter model that reflects inverter losses when using the synchronous PWM modulation method to operate the electric motor; b. Providing a motor model of the electric motor that reflects motor losses of the electric motor when implementing the synchronous PWM modulation method to operate the electric motor; c. Determine at least one switching angle of the synchronous PWM modulation method to define a start pulse pattern and determine the inverter losses and motor losses taking into account the start pulse pattern based on the inverter model and the motor model; d. Searching for suitable switching angles for operating the electric motor by changing at least one switching angle to obtain varied pulse patterns and evaluating the inverter and motor losses determined for each of the varied pulse patterns; and e. Storing the switching angles found by searching for controlling the inverter corresponding to the inverter model to operate the electric motor corresponding to the motor model, wherein a plurality of operating points of the motor corresponding to the motor model is defined, and for each operating point of the plurality of operating points, steps c. and d. are performed for each clock cycle (p) of a plurality of clock cycles, wherein, for a considered operating point of the plurality of operating points, in step c. a number of switching angles are determined as a function of the clock cycle (p), then in step d. the switching angles determined in the number are changed to obtain varied pulse patterns, and after iterating through the changes according to step d., steps c. and d. are repeated for another clock cycle of the plurality of clock cycles.

Description

[0001] The invention relates to a simulation method for generating pulse patterns defined by switching angles, and a corresponding computer program.

[0002] Prior art approaches exist for optimizing switching angles. For example, a genetic algorithm combined with selective harmonic elimination (SHE) is used to select which order of harmonic voltages should be eliminated with regard to minimizing distortion. With this approach, the number of possibilities increases exponentially with an increasing number of switching angles, and the algorithm requires a large number of iterations, resulting in long computation times. Furthermore, the position of the switching angles cannot be influenced, as no starting angles are required. Therefore, this approach is not suitable for delivering realistic results. The publication "Inverter and motor efficiency increase with FPCU implementing optimized pulse pattern methods," Douzane, Khaled et al., discusses this further.The paper "34th International Electric Vehicle Symposium and Exhibition, June 25-28, 2021" describes minimizing switching, copper, and iron losses during inverter operation by selecting suitable modulation strategies. The subsequently published German patent application DE 102023 202 008 A1 proposes generating optimized pulse patterns based on switching angles to reduce losses, noise, and torque fluctuations.

[0003] Furthermore, approaches to switching angle optimization using particle swarm optimization exist. However, due to poor convergence to a global optimum and the high number of iterations, this approach also fails to yield the desired results.

[0004] Against this background, the object of the invention is to create a simulation method that delivers realistic results for determining optimized pulse patterns. At the very least, the object of the invention is to provide an alternative to the prior art.

[0005] This problem(s) is solved by a simulation method according to claim 1. Preferred embodiments are the subject of the dependent claims.

[0006] The simulation method according to the invention for generating pulse patterns defined by switching angles for a synchronous PWM modulation method for controlling an inverter that feeds an electric motor comprises the following steps: a. Providing an inverter model that reflects inverter losses when using the synchronous PWM modulation method to operate the electric motor; b. Providing a motor model of the electric motor that reflects motor losses of the electric motor when implementing the synchronous PWM modulation method to operate the electric motor; c. Determine at least one switching angle of the synchronous PWM modulation method to define a start pulse pattern and determine the inverter losses and motor losses taking into account the start pulse pattern based on the inverter model and the motor model; d. Searching for suitable switching angles for operating the electric motor by changing at least one switching angle to obtain varied pulse patterns and evaluating the inverter and motor losses determined for each of the varied pulse patterns; and e. Saving the switching angles found by searching for controlling the inverter corresponding to the inverter model in order to operate the electric motor corresponding to the motor model.

[0007] The simulation method is, in particular, a computer-implemented method that is executed using a computer.

[0008] The simulated motor is, for example, a permanent magnet synchronous motor.

[0009] The search performed in step d. is preferably implemented using an algorithm that operates according to a system corresponding to the "patternsearch" algorithm in MATLAB. This allows the algorithm to find globally optimized switching angles.

[0010] The storage performed in step e. takes place in a motor control system, preferably in a modulation block for controlling an inverter corresponding to the inverter model, which in turn feeds the motor that corresponds to the motor model.

[0011] The number of switching angles determined in step c depends on the clock frequency p, with the respective switching angles being changed in step d. Additionally, a starting edge of the switching angle can be set to positive or negative in step c.

[0012] The switching angles determined by the simulation method and the pulse patterns defined thereby constitute a mid-pulse modulation or edge-pulse modulation. Reference is made to the corresponding details in the preferred embodiment.

[0013] The simulation method is preferably designed such that in step c. the at least one switching angle of the synchronous PWM modulation method is determined for a quarter or half period of the start pulse pattern and / or in step d. the switching angle for maintaining the varied pulse patterns is changed for a quarter or half period of the respective pulse pattern; wherein the inverter model has a switching angle block that maps the switching angle of the quarter or half period to a full period.

[0014] The inverter model preferably considers, for the purpose of determining the inverter losses, a type of switching element of the inverter corresponding to the inverter model and corresponding switching and / or conduction losses of the switching elements, wherein the inverter losses determined in step d. relate to switching and / or conduction losses during switching operations that occur taking into account currents flowing at that time, and / or the inverter losses determined in step d. relate to switching and / or conduction losses during switching operations that occur taking into account averaged currents averaged over a corresponding period resulting from the quarter or half period.

[0015] Examples of switching elements include transistors, such as IGBTs or MOSFETs.

[0016] The features of the preferred embodiment of the invention relating to the inverter model can be considered in isolation in connection with the inverter model.

[0017] In the simulation method according to the invention, the motor model preferably takes into account iron, magnet and / or copper losses to determine the motor losses, preferably divided between the rotor and stator of the motor.

[0018] The features of the preferred embodiment of the invention relating to the engine model can also be considered in isolation in connection with the engine model.

[0019] The simulation method is preferably designed such that a plurality of operating points of the motor corresponding to the motor model are defined, and for each operating point of the plurality of operating points, steps c. and d. are performed for each clock cycle (p) of a plurality of clock cycles, wherein, for a considered operating point of the plurality of operating points, in step c. a number of switching angles are determined as a function of the clock cycle (p), then in step d. the switching angles determined in the number are changed to obtain varied pulse patterns, preferably by calculation, and after the changes according to step d. have been carried out, steps c. and d. are repeated for another clock cycle of the plurality of clock cycles.

[0020] The clock numbers p are in particular odd, p = 3, 5, 7, ..., 23.

[0021] Each operating point is preferably defined by the motor's rotational speed n and torque Tq. Alternatively, each operating point can also be defined by the modulation level m and torque Tq. Each of the PWM pulse patterns defined by the switching angles found in step d. and stored in step e. preferably corresponds to exactly one operating point.

[0022] The simulation method according to the invention can be applied to, for example, 120V drive architectures with Si-MOSFETS, HV inverters with IGBT, railway technology, wind power, energy supply, photovoltaic inverters, micromobility, or also to high-voltage vehicle drives with a nominal voltage of at least 400 V or 800 V or a nominal power of at least 100 kW, 200 kW or more.

[0023] Furthermore, it is proposed to operate an electric drive according to the results of the simulation procedure. Specifically, it is proposed that an electric drive be operated which comprises an electric motor and an inverter. The switching angles determined and stored as described herein will be retrieved from a memory. The inverter will then operate according to these switching angles. The inverter will thereby generate a (multiphase) current that is supplied to the motor. The current supplied to the motor is based on the switching angles. The inverter corresponds to the inverter model. The electric motor corresponds to the motor model.

[0024] A preferred embodiment of the simulation method according to the invention is explained below with reference to the accompanying figures. The figures serve to illustrate exemplary embodiments of the methods described herein and the corresponding computer program. Fig. Figure 1 shows the conceptual structure of a control loop for controlling an electric motor, wherein the essential part of the invention relates to the control of the inverter supplying the motor by means of a synchronous PWM modulation method; Fig. 2a and Fig. Figure 2b shows a circuit diagram of the inverter and a corresponding equivalent circuit diagram; Fig. Figure 3 shows an example PWM pulse pattern of a phase U, V, W. Fig. Figure 4 shows a block diagram of a simulation model used to carry out the simulation method according to the invention. Fig. 5a and Fig.Section 5b shows the calculation of inverter losses and motor losses in detail. Fig. 6a and Fig. Figure 6b shows two possibilities for mid- and edge pulse modulation.

[0025] Fig. Figure 1 shows a conceptual design for the control of an electric motor 1, such as a permanent magnet synchronous motor, which can preferably be operated in motor mode but also in generator mode. The electric motor 1 is preferably an electric drive unit of a motor vehicle or motorcycle.

[0026] The simulation method according to the invention, which is described in detail below, serves to generate loss-optimized synchronous PWM pulse patterns for the subsequent control of the motor 1.

[0027] The control of motor 1 will now first be explained for a better understanding of the invention, with particular attention to the synchronous PWM pulse patterns used.

[0028] The control loop of the shown control system detects current intensities of alternating currents i at a first block T1. u , i v , i w , the phases U, V, W supplying motor 1 flow.

[0029] Block T1 preferably captures in Fig. 1. Only the currents of phases U and V are recorded. The remaining alternating current value of phase W is calculated from the recorded values. The alternating currents have a sinusoidal waveform and are preferably phase-shifted by 120° relative to each other, with the same being true for the corresponding phase voltages.

[0030] Block T1, for example, transforms the three detected alternating quantities of the currents flowing in phases U, V, W into stator- and / or rotor-fixed current quantities using Clarke transformations and / or Park transformations. These transformations are known.

[0031] The obtained current quantities can be compared with externally specified reference quantities Ref at a comparator VG and then modeled in a desired way via the control block RB shown, which is, for example, a PI controller.

[0032] For example, the externally specified reference values ​​Ref correspond to acceleration requirements given by a driver.

[0033] The received output variables of the control block RB form an operating point of the motor 1, which is to be approached, and input variables of a modulation block 3.

[0034] The operating point is defined in particular by a desired rotational speed n and a desired torque T. q 1. Alternatively to the speed, a modulation level can be used to define the operating point of the motor. Preferably, further parameters can also be used to define the operating point.

[0035] Modulation block 3 controls an inverter 2 to generate target alternating voltages (with amplitudes ûs oll ) on the phases U, V, W, which are phase-shifted relative to each other and supply the motor 1 for starting up to the operating point. The modulation block 3 shown takes over the control of the inverter 2 by applying each of the target AC voltages (U, V, W) with amplitudes ûs oll ) is replicated or modeled by synchronous PWM pulse width modulation.

[0036] Modulation block 3 employs a synchronous modulation method, which is explained in more detail below. This method does not calculate the required synchronous PWM pulse pattern in real time (online), but rather accesses a stored PWM pulse pattern that is generated offline for the respective operating point. For this purpose, modulation block 3 has a memory in which the necessary PWM pulse patterns for possible operating points are stored.

[0037] The present invention is aimed at obtaining the aforementioned necessary synchronous PWM pulse patterns by means of a simulation method and storing them in the memory of the specific control loop corresponding to the simulation.

[0038] Fig. Figure 2a shows a circuit diagram of inverter 2.

[0039] Inverter 2, which powers motor 1, plays a central role. In the... Fig. In the control loop shown in section 1, it converts a DC voltage U. dfrom a battery (high-voltage battery) not shown into the required three target AC voltages U, V, W (with amplitudes ûs oll ) for engine 1.

[0040] The three phases U, V, W of motor 1 are in Fig. 2a, for example, can be connected together in a star configuration. A delta configuration is also possible.

[0041] For the conversion of the DC voltage U d Inverter 2 operates using pulsed switching states. Instead of a continuous energy flow, the energy is transferred in individual packets using synchronous PWM modulation. A high-frequency sequence of these packets transforms the segmented energy flow into an approximately continuous output signal that powers the corresponding phase of motor 1.

[0042] The in Fig.Inverter 2 shown in Figure 2a is preferably a B6C bridge circuit with three bridge branches, each bridge branch corresponding to one of the phases U, V, W. Each of the three bridge branches contains two switches 20, 21 and 22, with a tap for the respective assigned phase U, V, W being made between the corresponding switches 20, 21 and 22.

[0043] Switches 20, 21, and 22 are controlled using the aforementioned synchronous PWM modulation method to generate the alternating quantities in phases U, V, and W. Each of the three bridge branches thus supplies one of the phases U, V, and W, whereby the synchronous PWM modulation method makes it possible to generate the three-phase alternating voltage for motor 1 with variable frequency and amplitude.

[0044] This allows any operating point to be defined by rotational speed n or modulation level m and torque T. q - start within the operating range of motor 1.

[0045] To achieve the required high switching frequencies, semiconductor switches such as IGBTs (Insulated-Gate Bipolar Transistors) or MOSFETs (Metal-Oxide-Semiconductor Field-Effect Transistors) are typically used for switches 20, 21, and 22. However, these power semiconductors limit the switching frequency because their switching losses increase with increasing switching frequency, and there is a limit to the dissipated power.

[0046] A compromise must therefore be found between good modulation of the phase voltages, and thus low losses due to harmonics of the phase currents, and not excessively high switching losses. The simulation method explained later achieves this compromise.

[0047] The in Fig. The bridge circuit shown in 2a, corresponding to inverter 2, can be represented in an idealized form. Fig. Figure 2b shows the corresponding equivalent circuit diagram, in which switches 20, 21, 22 are replaced by simple switches.

[0048] A virtual center point of the intermediate circuit is introduced, but not connected to a load neutral point N, as this would otherwise prevent it from being a symmetrical three-phase system. This applies to both the aforementioned star connection and the delta connection.

[0049] In the equivalent circuit diagram, the battery voltage U d The voltage is split into two identical voltage sources. The three phases U, V, and W can thus be switched to the two available voltages +Ud / 2 and -Ud / 2 to replicate the desired sinusoidal voltage waveform. The resulting currents i U , i V and i W Ideally, they exhibit the same sinusoidal shape. The same relationship applies to magnetic flux.

[0050] Furthermore, the equivalent circuit diagram according to Fig. 2b Central stresses U U0 , U V0 and U W0shown, which can be derived from the respective switch position. These result as follows: UU0=sU⋅Ud2. UV0=sV⋅Ud2, UW0=sW⋅Ud2

[0051] The neutral point voltage U is also used to calculate the phase voltages. N0 used. For a symmetrical load from motor 1 without a common-mode component, this results in: UN0=(sU+sV+sW)⋅Ud6

[0052] Due to the symmetrical voltage system without a neutral conductor, the sum of the phase currents i must also be U , i V and i W The result is zero. Therefore, a phase quantity can be calculated from the remaining two. iU+iV+iW=0

[0053] From the three bridge branches, 2 can therefore be derived. 3 = 8 discrete switching states can be implemented. The switching states and the corresponding voltages are summarized in the following table. Condition s U s V s W u U0 u V0 u W0 u U u V u W u N0 0 Z -1 -1 -1 −Ud2 −Ud2 −Ud2 0 0 0 −Ud2 1 Z 1 -1 -1 Ud2 −Ud2 −Ud2 2Ud3 2Ud3 −Ud3 −Ud3 2 Z 1 1 -1 Ud2 Ud2 −Ud2 Ud3 −2Ud3 −2Ud3 −Ud6 3 Z -1 1 -1 −Ud2 Ud2 −Ud2 −Ud3 2Ud3 −Ud3 −Ud6 4 Z -1 1 1 −Ud2 Ud2 Ud2 −2Ud3 Ud3 Ud3 Ud6 5 Z -1 -1 1 −Ud2 −Ud2 Ud2 2Ud3 2Ud3 2Ud3 −Ud6 6 Z 1 -1 1 Ud2 −Ud2 Ud2 Ud3 −2Ud3 Ud3 Ud6 7 Z 1 1 1 Ud2 Ud2 Ud2 0 0 0 Ud2

[0054] The eight different states 0 Z to 7 Z can still be put into voltage-generating states ( 1 Z to 6 Z) and zero states ( 0 Z and 7 Z) subdivide. The latter each result in a phase voltage of zero, since the three phases are short-circuited via inverter 2. As can be seen from the table above, in this state 0 Z all switches in position -1 and consequently in the state 7 Z in position +1.

[0055] The in Fig.In this context, modulation block 3, shown in Figure 1, is configured to control switches 20, 21, and 22 according to the read-out synchronous PWM pulse pattern, which is only phase-shifted for phases U, V, and W, in order to generate the corresponding alternating currents in the respective phases. Specifically, modulation block 3 uses the voltage-generating states to generate the respective PWM pulse patterns. In general, the states can be represented as space vectors in the α-β coordinate system according to the table above, and modulation block 3 uses these space vectors, particularly those corresponding to the voltage-generating states, to control inverter 2.

[0056] The modulation method used is, as already mentioned, a synchronous modulation method.

[0057] Here, the selected switching frequency is always synchronous with a fundamental frequency of the output voltage, and the zero crossings are synchronized. A clock ratio nT Here, becomes an integer and a switching frequency is variable. The clock ratio n T gives a ratio of the switching frequency f sw to the frequency of the output voltage, where the former switching frequency f sw a measure of the number of switching cycles of a switch per unit of time.

[0058] The switching frequency's temporal behavior is identical across every electrical period, meaning that no frequencies lower than the fundamental frequency occur during transformation to the frequency domain. Subharmonics, which occur with asynchronous modulation, are thus avoided.

[0059] In addition to synchronization, symmetries of the phase voltages U, V, W to be controlled are employed. Besides half-period symmetry, quarter-period symmetry is used because it reduces current harmonics. This symmetry in the phase voltages prevents even-numbered harmonics from occurring. Furthermore, modulation block 3 only needs to store switching angles for the first quarter period of each PWM pulse pattern. From this, the full PWM pulse pattern can be generated.

[0060] Fig. Figure 3 shows an example of a synchronous PWM pulse pattern for one of the phases U, V, W, which is simply phase-shifted for the others. The second half-period is the inverse switching function of the first half-period.

[0061] Additionally, the switching angle profiles are axially symmetrical to each other, with symmetry axes at π / 2 and 3π / 2. All pulse pattern information is thus contained in the first quarter period and can be used as the basis for developing the full PWM pulse pattern by modulation block 3. The harmonic voltage overtones have the same phase as the fundamental frequency or are out of phase with it. Quarter-period symmetry also requires an odd clock frequency p and a switching edge at ωt = 0 and ωt = π.

[0062] Instead of calculating the switching angles in real time – online – they are calculated in advance based on specific parameters corresponding to the respective operating point and stored in the memory of modulation block 3. The determination and optimization of the switching edges, and thus the pulse patterns, are carried out using the simulation method explained below.

[0063] To reach a specific operating point, a modulator (not shown) is provided, which reads the pulse pattern matching the input parameters (operating point) from the memory of modulation block 3.

[0064] Preferably, for each modulation level m, the PWM modulation patterns with their switching angles and switching states are stored in the aforementioned memory, which are converted into discrete switching times depending on the rotational speed n. The modulation block 3 then forwards the control signals to the inverter 2, which directly controls the motor 1. Due to the dependence of the PWM pulse patterns on the fundamental frequency, the switching frequency increases proportionally with increasing fundamental frequency. Because the maximum switching frequency is limited by the inverter 2, the clock rate p is varied for the entire operating range.

[0065] The different clock frequencies p and the optimization possibilities in the simulation procedure are considered below in order to explain the different possibilities for generating synchronous modulation or synchronous PWM pulse patterns.

[0066] Triple switching is the simplest clock frequency p to implement, since each voltage block per half-cycle is interrupted by only one intermediate pulse. The number of freely selectable switching angles q that occur in a quarter-cycle can be determined according to the following equation. The clock frequency p describes the number of intermediate pulses over one electrical cycle. In the case of triple switching, six on / off cycles correspond to three intermediate pulses. q=p−12 with p=3,5,7..

[0067] Using the example of the aforementioned triple clocking, this results in only one free switching angle in a quarter period.

[0068] With synchronous timing, at least two variations are possible: center-pulse and edge-pulse modulation. They differ in the position of the intermediate pulse, either in the middle or on the edges of a voltage block of length π.

[0069] All switching angles of a pulse pattern together determine the modulation level. The modulation level m describes the normalized length of the voltage space vector and is calculated from the ratio between the amplitude of the output voltage fundamental oscillation and the DC link voltage, divided by the square root of three. In this application, this always refers to the modulation level determined by motor parameters.

[0070] For both mid-pulse and edge-pulse modulation, there are two ways to achieve the desired modulation level m, which differ in whether the switching edge is positive or negative at time wt = 0. The following figures illustrate these two possibilities, where the pulse patterns achieve the same modulation level. Type 1 denotes a positive starting edge in phase U, which is associated with a positive sign for the Fourier coefficients of the phase voltage amplitude. Type 2 represents a negative starting edge with a correspondingly negative sign. bv=8vπ[12+∑i=1q(−1)icos(vαi)] bv=−8vπ[12+∑i=1q(−1)icos(vαi)]

[0071] As shown in the preceding illustrations, a positive sign means that the switching pattern starts in the switched-on state. A negative sign means that the switching pattern starts in the switched-off state.

[0072] Therefore, when determining the pulse patterns in the simulation method according to the invention, two solutions are obtained, both of which are calculated. Not all solutions are possible for every modulation level m; some are only available for a sub-range.

[0073] As mentioned previously, in edge pulse modulation, the pulses at which the voltage edge changes are located at the edges of the half-period. Similarly, the corresponding switching angle can be set proportionally to the desired modulation level m. In edge pulse modulation as well, the maximum modulation level is limited by the inverter dead time.

[0074] Overall, center-pulse modulation, due to the placement of the intermediate pulse in the middle of the switching pattern, has a greater influence on the amplitude of the fundamental frequency than edge-pulse modulation. However, edge-pulse modulation offers advantages at high modulation intensities, as the switching angles are in the range of small voltage amplitudes.

[0075] In general, the goal of synchronous modulation is to minimize the amplitudes of harmonics in the current. This primarily reduces harmonic losses in the motor. On the other hand, switching losses in the inverter must be limited, which inherently increase with more complex modulation and the associated higher effective switching frequencies.

[0076] According to the invention and the simulation method, the switching angles (α1, ..., α q ) of the quarter periods directly based on simulated inverter and motor losses P υoptimized. The following boundary conditions apply: Pv(α1,α2,…,αq)=min u^v=1(α1,α2,…,αq)=u^soll 0<α1<α2<…<αq<π2 where the amplitude of the target alternating voltage ûs oll the amplitude of the fundamental wave resulting from all switching angles û υ=1 corresponds.

[0077] Firstly, the simulation method finds the optimum defined from all free switching angles, which in this case is a minimum.

[0078] Secondly, the required target voltage amplitude ûs is derived from all switching angles. oll and thus the desired modulation level m, which correlates with the rotational speed at the operating point.

[0079] Finally, due to the quarter-period symmetry, no switching angle may be less than 0 rad or greater than π / 2, and all switching angles are preferably sorted in ascending order.

[0080] The PWM pulse patterns found are finally saved so that they are available for the operation of the real inverter 2 and motor 1 belonging to the simulation and can be stored in the memory of the modulation block 3.

[0081] The following describes a simulation model used to execute the method according to the invention. The simulation method based on this model yields the optimized PWM pulse patterns, in particular the optimized switching angles of the quarter period, from which the full PWM pulse pattern can be developed.

[0082] According to the simulation method, the switching angles are optimized based on the total losses in order to calculate a fuel-efficient result. For this purpose, a simulation model for motor and inverter losses that is as realistic as possible is used.

[0083] Based on these loss models, various synchronous modulation types, center and / or edge pulse modulation, as well as clock frequencies are simulated at a fixed operating point.

[0084] The switching angles are preferably determined as a function of the modulation level m; this allows, for example, verification of the simulation results on an experimental setup in which the simulation results are compared with an asynchronous space vector modulation (SVPWM) known from the prior art.

[0085] Fig. Figure 4 shows a block diagram of the simulation model used and the integration of the switching angle optimization. The simulation model provides an inverter model (IM) and a motor model (MM) for the simulation procedure, which are used to calculate the total losses.

[0086] The simulation method using the simulation model is, for example, a computer program that is preferably executed on two parallel information processing computers.

[0087] First, the operating points to be simulated are selected and their input parameters are defined in simulation block SB1. The essential input parameters, which are variable for the simulation procedure, are torque Tq and rotational speed n. The other input parameters are preferably kept constant to reduce the computational effort.

[0088] The data is then transferred to the actual switching angle optimization in simulation block SB2.

[0089] SB2 represents the main program, which is structured, for example, as a MATLAB script. All other blocks are also preferably defined as MATLAB functions, since they are accessed multiple times.

[0090] The optimization is performed individually for each operating point based on the calculation of total losses, including inverter and motor losses. In this process, a cycle count p and a switching angle (start angle) are defined for the operating point in SB2.

[0091] The in Fig. The optimization loop shown in Figure 4 refers to an iteration in which the total losses are calculated for a fixed clock frequency p and the switching angle.

[0092] The calculation is then repeated by changing / calculating the switching angle for the operating point while maintaining the clock frequency p. Once all switching angles have been calculated for the clock frequency p, the procedure increments the clock frequency p and recalculates the switching angles for this clock frequency. The number of switching angles varied for each clock frequency depends on the clock frequency. These described loops are preferably executed for p = 3, 5, 7, ..., 23.

[0093] Only when the switching angle has been varied using a solver until the total losses are minimal, is the next operating point optimized.

[0094] For each operating point, the different losses, switching angles and preferably modulation levels are stored in correspondingly large, multidimensional matrices.

[0095] The following sections explain in more detail the blocks used for the simulation procedure.

[0096] In block SB1, the operating points are defined and their input parameters are specified. The following table contains parameters that remain constant throughout the entire simulation process, as well as the aforementioned variable parameters.

[0097] In addition to the parameters Tq and n, the battery voltage and temperature could also be varied, but this would greatly increase the number of solutions. parameter index Constant parameters Maximum modulation level m max Maximum effective current I eff,max Number of phases m phase Motor and inverter temperature θ Battery voltage U d Number of harmonics of the voltage v Variable Parameter torque Tq speed n

[0098] After the operating points are parameterized, the desired motor model MM and inverter model IM are loaded. Furthermore, the desired output parameters of the motor model can be selected, as only a fraction of the simulated values ​​are needed per iteration. This serves to minimize the computation time.

[0099] The solver, for example, patternsearch in MATLAB, is then initialized. This is described in more detail below. An outer for loop calculates the switching angles from the minimum to the maximum selected clock cycle. Depending on the clock cycle, the corresponding number of switching angles (start angles) is initialized.

[0100] The simulation method is designed to differentiate between mid-pulse and edge-pulse modulation.

[0101] First, the inverter model will be considered.

[0102] The in Fig. Section 4, delimited area IM, models the power electronics of inverter 2, and the properties of the inverter model IM and the simulation of its losses are explained in the following section. In the present application, an inverter 2 with SiC MOSFETs with the following properties was modeled. parameter Value Unit Semiconductor material Silicon carbide - Maximum peak current 600, 10s A rms Voltage class 1200 V Number of MOSFETs 6 - Weight 100 kg Power density 150 kW / 1

[0103] The simulation of inverter 2 can preferably be divided into three large blocks, which are represented by functions. Switching Angles IM1

[0104] Due to the quarter-period symmetry, the optimized angles are defined and optimized only for a quarter period. For further calculations, these are mapped to a full electric period in block IM1.

[0105] Furthermore, in this block IM1, it can be preferably selected whether the switching angles begin with a positive or negative switching edge.

[0106] This block IM1 is simply preferred. Alternatively, the switching angles for the entire period can be stored. Effective switching angles IM2

[0107] The stator current resulting from the operating point is calculated from the DC quantities Id and Iq and preferably converted to an RMS current. Since these DC quantities are motor parameters, a conversion to phase parameters is performed to determine the inverter current.

[0108] If the simulated electric motor is connected in a delta configuration, the RMS current value must be multiplied by the square root of 3. The RMS current serves as the input value for further inverter loss calculations.

[0109] Inverter loss calculations primarily focus on losses during switching operations at the point of maximum current amplitude. For this purpose, the RMS value of the current is converted into a pure sinusoidal waveform. The current amplitudes at the switching times are then determined for this sinusoidal current, taking into account the phase shift of an operating point.

[0110] Block IM2 calculates the average current over one period from the current amplitudes, which would flow at all switching times. This average current—normalized to the maximum occurring current—yields the correction factor. This factor can take values ​​between 0 and 1, where a value of 0 would correspond exclusively to switching times within the current's phase response. A value of 1 corresponds to the case where all switching angles occur at times of maximum current in the inverter. Neither of these extreme cases occurs in reality, so the correction factor value always lies somewhere in between.

[0111] This preferred approximation can be performed in significantly less computation time than the alternative method of comparing the actual current waveform with the switching pattern, since the inverter losses must be recalculated in each iteration. The computation time, which increases sharply with the number of iterations, outweighs the slight inaccuracy of this approximation. Inverter loss calculation IM3

[0112] The preferred inverter loss calculation includes all losses that occur in inverter 2 during the conversion of the intermediate circuit voltage into the multi-phase system.

[0113] The majority of inverter losses are caused by switching and conduction losses, which are the primary focus here. By reducing the number of switching operations in certain operating ranges, synchronous PWM can offer advantages over asynchronous SVPWM.

[0114] The simulation method achieves a short computation time for loss calculations, enabling it to handle the large number of iterations while simultaneously achieving the most realistic result possible. Switching and conduction losses are calculated separately for the MOSFETs and diodes in each bridge branch.

[0115] In the Fig.Figure 5a shows a schematic representation of the calculation of inverter losses according to IM3. eff The initially defined boundary conditions must also be passed to the functions. Switching losses IM31

[0116] Switching losses occur during the MOSFETs' turn-on and turn-off processes when the output current is commutationd. The switching loss energy for both the MOSFET and the diode is interpolated from a table using the maximum current amplitude in Inverter 2, the DC link voltage, and the temperature. The motor temperature is assumed to be the temperature used. Multiplying this value by the number of switching elements yields the maximum switching losses for the entire Inverter 2 across all three phases. These maximum switching losses are then adjusted using the previously determined correction factor and the effective switching frequency. IM32 transmission losses

[0117] The conduction losses are calculated analogously to the switching losses. They occur during current-carrying operation in the semiconductor elements. In the case of the MOSFET, the drain-source voltage of the MOSFET is U. DS The maximum amplitude of the current in inverter 2 and the temperature are interpolated from a table. This voltage—multiplied by the same maximum current—yields the theoretically maximum achievable conduction losses. The maximum conduction losses calculated in this way are then adjusted using the quadratic correction factor, since both the interpolation and the multiplication are performed with the maximum current.

[0118] The conduction losses of the diode result from the interpolated freewheeling voltage U. f The diode's current rating is multiplied by the maximum current. The resulting power loss is also corrected using the quadratic correction factor.

[0119] Additionally, the calculation of dead-time losses is omitted for this block. Since these losses represent a significant portion of the inverter losses, especially at high switching frequencies, their calculation should not be neglected. In the case of a negative phase current, these losses are attributed to the MOSFET's body diode. These losses are added to the total inverter losses. Blocking losses

[0120] Furthermore, switching transistors and diodes exhibit additional losses, such as reverse bias and drive losses. Due to their low leakage currents, reverse bias losses are very small compared to the main losses—conductance and switching losses—and can be taken into account, but do not necessarily have to be. Tax losses

[0121] The drive losses are generally low due to the short pulsed drive currents in IGBTs and MOSFETs and can also be neglected. SNT, PCC and busbar losses IM33

[0122] In addition to switching and conduction losses, the losses of the switched-mode power supply (SMP), busbar, and power capacitor chip (PCC) are also preferably calculated. These together represent a smaller proportion than the switching and conduction losses, but can be given preferential consideration. SMP losses originate in the power supplies that power the MOSFETs. A distinction is made between low-voltage (LV) and high-voltage (HV) switching power supply losses, which together constitute the total SMP losses. Both switching power supplies must provide a constant gate-source voltage. Since they are also dependent on the switching frequency, lower losses result from choosing a lower clock rate at a given operating point. The effective switching frequency fsw,eff is calculated by multiplying the electrical frequency fel by the clock rate p.

[0123] Busbar losses originate from the busbars between the inverter and the motor and are divided into losses caused by the AC component and the DC component of the flowing current. Both components result from the d q Currents for the respective operating point from the machine loss model. For the AC component, the effective current l is used. eff used. The DC component of the current I DC This is calculated using cos(φ) and the respective modulation level m. The individual currents are multiplied by the AC and DC resistance of the busbar of all three phases, which were obtained from a measurement. The sum of the AC and DC losses yields the total busbar losses.

[0124] To calculate PCC losses, the losses occurring in the intermediate circuit capacitor must be determined. First, the current is calculated using the RMS current, cos(φ), and the modulation index m. With the calculated current and the equivalent series resistance (ESR) of the capacitor, the corresponding power loss at the respective operating point can then be determined.

[0125] The following section examines the engine model.

[0126] The in Fig. The delimited area MM represents the simulated electric motor and its loss calculation. The properties of the motor model MM and the simulation of the losses are described in more detail below with reference to... Fig. 5b explained. The following properties were chosen for the electric motor 1. parameter Value Unit Rated power 100 - 250 kW Rated torque 1700 - 4000 Nm Number of pole pairs 4 - Number of phases 3, Triangle - Weight 60 - 80 kg Nominal voltage 400 V Stator length 60 - 180 mm The simulation of the engine can be divided into four large, preferred blocks, which are represented by functions. PWM Pulse Pattern MM1

[0127] In block MM1, the pulse pattern is generated from the input variable of the switching angle, along with information on whether a falling or rising edge is present.

[0128] To simulate the corresponding PWM driver, its parameters are also defined. These include the clock frequency of the installed CPU and the number of periods to be simulated. One switching angle can be implemented per PWM period. Since each period in synchronous PWM is symmetrical with the electrical frequency due to the synchronized zero crossings, simulating one electrical period is sufficient. PWM Spectrum List MM2

[0129] The generated pulse pattern is passed to block MM2, which additionally receives the desired number of harmonic orders.

[0130] For example, the 150 orders with the largest absolute voltage amplitudes are always used.

[0131] Since the harmonic orders are not automatically in ascending order, a vector containing information about the order of the harmonic voltage phasors must be generated. Subsequently, the complex voltage phasors of the harmonic overtones are generated for all three phases. This is done using a Fast Fourier Transform (FFT), which transitions from the image domain to the frequency domain. Because the pulse pattern changes with each iteration, the generation of the complex spectrum must also be performed in each iteration. Udq Spectra MM3

[0132] In block MM3, the complex harmonic voltages are preferably transformed into the dq coordinate system using an inverse Clarke-Park transformation. This is preferred because the motor loss calculation can be performed in these coordinates and the fundamental frequency of the current can also be expressed as I. d or I q is available.

[0133] The most important output parameter of the engine model is its loss (Block MM4).

[0134] For each operating point from block SB1, the complex voltage spectrum is generated from the switching angles via the pulse pattern (MM41). The fundamental oscillation of the currents I dq,fundThe system is determined through optimization according to Maximum Torque per Ampere (MTPA) control. From the complex harmonic voltages, the harmonic overtones of the d- and q-currents are also calculated using machine equations (MM42). These, together with the fundamental frequency, are incorporated into the calculation of harmonic losses. Specifically, the harmonic iron, magnet, and copper losses are calculated, separated for the stator and rotor.

[0135] After adding the harmonic losses, the fundamental losses are added, as these only need to be calculated once for each operating point. In contrast, the harmonic losses differ with every change in the switching angles.

[0136] The simulation method finds a global optimum and enables parallelized computation. Generally, various search algorithms are possible for finding the optimum.

[0137] Preferably, the MATLAB algorithm "patternsearch" enables the finding of the global loss optimum and parallelized computation.

[0138] It is possible to specify initial values ​​for the switching angles to more quickly find a solution with switching angles at the edges or in the middle of the half-periods. The algorithm used creates a pattern of potential solutions around the starting point (switching angle), the so-called mesh. Its size is initialized at the beginning and modified over subsequent iterations.

[0139] The new points are derived from the initial value, added to the size and direction of the grid. The calculated solutions for these points are compared, and the best solution is chosen as the new initial value.

[0140] Results that do not meet the chosen boundary conditions will be disregarded. The boundary conditions are defined by the formulas explained above and shown below. Pv(α1,α2,…,αq)=min u^v=1(α1,α2,…,αq)=u^soll 0<α1<α2<…<αq<π2 used. Additionally, a minimum interval between the switching angles can preferably be inserted, which corresponds to the dead time. This is necessary to ensure that the inverter's switches are not switched on simultaneously.

[0141] For each new potential solution, the process is repeated, increasing the mesh size until no better solution is found. Then, the mesh is reduced in size at each step, for example, by halving its size.

[0142] The solver stops when iterating to a minimum and the termination criteria are met. No information about the gradients is required to perform pattern search, making this solution method suitable even for discontinuous functions.

[0143] Regarding switching angle optimization, the switching angles obtained for each operating point are stored and used as starting values ​​for the next operating point. This facilitates rapid iteration to the optimum, since, with the low discretization used, the optimal switching angles are close to the previous operating point.

Claims

[1] Simulation method for generating pulse patterns defined by switching angles for a synchronous PWM modulation method for controlling an inverter that supplies an electric motor, wherein the simulation method comprises the following steps: a. Providing an inverter model that reflects inverter losses when using the synchronous PWM modulation method to operate the electric motor; b. Providing a motor model of the electric motor that reflects motor losses of the electric motor when implementing the synchronous PWM modulation method to operate the electric motor; c. Determine at least one switching angle of the synchronous PWM modulation method to define a start pulse pattern and determine the inverter losses and motor losses taking into account the start pulse pattern based on the inverter model and the motor model; d. Searching for suitable switching angles for operating the electric motor by changing at least one switching angle to obtain varied pulse patterns and evaluating the inverter and motor losses determined for each of the varied pulse patterns; and e. Storing the switching angles found by searching for controlling the inverter corresponding to the inverter model to operate the electric motor corresponding to the motor model, wherein a plurality of operating points of the motor corresponding to the motor model is defined, and for each operating point of the plurality of operating points, steps c. and d. are performed for each clock cycle (p) of a plurality of clock cycles, wherein, for a considered operating point of the plurality of operating points, in step c. a number of switching angles are determined as a function of the clock cycle (p), then in step d. the switching angles determined in the number are changed to obtain varied pulse patterns, and after iterating through the changes according to step d., steps c. and d. are repeated for another clock cycle of the plurality of clock cycles. [2] Simulation method according to claim 1, wherein in step c. the at least one switching angle of the synchronous PWM modulation method is determined for a quarter or half period of the start pulse pattern and / or in step d. the switching angle for maintaining the varied pulse patterns is changed for a quarter or half period of the respective pulse pattern; and the inverter model has a switching angle block that maps the switching angle of the quarter or half period to a full period. [3] Simulation method according to claim 1 or 2, wherein To determine the inverter losses, the inverter model takes into account a type of switching element of the inverter corresponding to the inverter model and corresponding switching and / or conduction losses of the switching elements, and The inverter losses determined in step d. refer to switching and / or conduction losses during switching operations, taking into account currents flowing at that time. [4] Simulation method according to claim 1 or 2, wherein To determine the inverter losses, the inverter model takes into account a type of switching element of the inverter corresponding to the inverter model and corresponding switching and / or conduction losses of the switching elements, and The inverter losses determined in step d. relate to switching and / or conduction losses during switching operations, taking into account averaged currents averaged over a corresponding period resulting from the quarter or half period. [5] Simulation method according to claim 3 or 4, wherein, in order to determine the inverter losses, the inverter model takes into account the following losses in addition to the corresponding switching and / or conduction losses of the switching elements: - Switching power supply losses (SBT); - Busbar losses; and / or - Power capacitor chip (PCC) losses. [6] Simulation method according to any one of claims 1 to 5, wherein, to determine the motor losses, the motor model takes into account iron, magnet and / or copper losses of the motor corresponding to the motor model. [7] Simulation method according to claim 6, wherein, to determine the motor losses, the motor model takes into account the iron, magnet and / or copper losses divided between the rotor and stator of the motor. [8] Simulation method according to one of the preceding claims, wherein To determine the motor losses, the motor model takes into account the iron, magnet and / or copper losses of the motor and provides these separately for a fundamental frequency and harmonic overtones with respect to the operating point under consideration, and For each cycle, the iron, magnet and / or copper losses of the motor are determined only once for the fundamental frequency in the corresponding steps c. and d., and for the harmonic overtones are determined for each change of the switching angle according to d. [9] Computer program which, when executed on a computer, is configured to perform the simulation method according to any one of the preceding claims 1 to 8. [10] Operating an electric drive comprising an electric motor and an inverter, wherein those switching angles are recalled from a memory which were stored according to the method according to one of claims 1-9, and wherein the inverter is operated according to the switching angles, and generates a current which is supplied to the motor.

Images