METHOD AND DEVICE FOR CONTROLLING A SYNCHRONOUS MACHINE WITHOUT A POSITION SENSOR BY UNIQUE ASSIGNMENT OF ADMITTANCE OR INDUCTIVITY TO THE ROTOR POSITION

Description

[0001] The invention relates to a method for assigning the inductance or admittance to the rotor position of a synchronous machine, comprising a stator and a rotor with or without permanent magnets, wherein the synchronous machine is controlled via pulsed terminal voltages and the inductance or admittance is calculated from these and the measured current response, as well as a device for controlling and regulating a rotating field machine, which is set up and configured for carrying out such a method.

[0002] Methods that enable efficient control of a synchronous machine without a position sensor (often referred to as "sensorless" or "sensorless" control) are divided into 2 classes: 1. Fundamental wave methods evaluate the voltage induced by motion, deliver very good signal characteristics at medium and high speeds, but fail at lower speeds, especially when stationary. 2. Anisotropy-based methods evaluate the position dependence of the machine's inductance, for which no rotational speed is necessary, but exhibit several problems and hurdles that explain why many applications still require a position sensor (with its disadvantages).

[0003] Anisotropy-based methods can be further divided into two stages. The first stage is anisotropy identification, which calculates the components of the anisotropy vector from the applied voltage and the resulting current response. Examples of methods relating to this stage can be found in [1], [2], [3], [4], [5]. The second stage is rotor position assignment, in which a rotor position is assigned to the measured anisotropy vector. Examples of this can be found in [6] [7] [8] [9].

[0004] Although anisotropy-based methods have existed since 1988 [1]

[10]

[11]

[12] , no proposal with rotor position assignment existed until 2004. Saturation and harmonics were neglected, and the measured / calculated anisotropy angle was directly fed back as the rotor position value for control.

[0005] Since 2005, publications have revealed a load dependency of the anisotropy [7]

[13] . These publications describe, on the one hand, that its orientation shifts relative to the rotor, which has since been compensated for during operation using a corresponding characteristic curve

[14]

[15]

[16]

[17] . Such compensation methods can be considered the first procedures for anisotropy-rotor position assignment. These publications also describe that the magnitude of the anisotropy ( L d - L q ) decreases, which reduces or potentially leads to the loss of location information.

[0006] Since 2007, the anisotropy shift has also been described as a function of the rotor position in the form of harmonics [8], which have since been compensated

[18]

[19] or evaluated [6]

[20] . Both approaches can be considered as an extended anisotropy-rotor position mapping and mainly lead to improved accuracy.

[0007] Regarding the phenomenon of diminishing anisotropy under load described above, approaches have been proposed since 2013 [9]

[21] to modify the efficiency- or performance-optimal target current trajectories (e.g., MTPA) in such a way as to bypass critical points with vanishing anisotropy. However, such approaches promise only moderate success because the actual problem is not the vanishing of the anisotropy, but, as explained in

[22] , a more complex one that persists even after bypassing the critical points.

[0008] Approaches

[23] ,

[24] and

[25] attempt to overcome this problem by additionally considering, based on a conventional rotor position assignment, that the occurrence of an estimation error (as expected) leads to the imprinting of an incorrect current and, based on this, formulating an estimation error correction rule that exhibits improved stability properties. However, even in these methods, the underlying model remains dependent on the rotor coordinates, which are unknown during estimation, and therefore cannot provide a completely unique rotor position assignment.

[0009] German patent application DE 10 2016 102 635 A1 discloses a method for calculating the rotor position using the stator currents as a function of the load. The measured stator currents are transformed into a stator-fixed coordinate system, and components of the admittance vector are determined from this.

[0010] The IEEE publication "Compensation for Rotor Position Estimation Error due to Cross-Coupling Magnetic Saturation in Signal Injection bases Sensorless Control of PM Brushless AC Motors" by ZQ Zhu et al. (IEEE, Piscataway, NJ, USA, May 1, 2007, pages 208-213, XP031114836) also discloses the rotor position estimation taking the load into account, whereby a coordinate transformation into a rotor-fixed coordinate system is performed.

[0011] The following is a general explanation, also concerning optional embodiments of the invention. The following will be shown: Fig. 1 Conventional anisotropy trajectories without load (left), with load (center), and a critical curve (right), each with the 0° point marked (small circle). Fig. 2 Block diagram of the two-stage setup of conventional anisotropy-based methods with current transformation into estimated rotor coordinates. Fig. 3 Conventional anisotropy trajectories with current angle dependence at a fixed rotor position. Fig. 4 Conventional anisotropy trajectories with current dependence at a fixed rotor position and position dependence at a fixed stator current. Fig. 5 Block diagram of the two-stage setup of an anisotropy-based method with SFC-based rotor position assignment, without current transformation into estimated rotor coordinates. Fig. 6 Assignment of a measured anisotropy vector. y ˜ Δ s to the nearest point of the SFC trajectory that corresponds to, for example, the currently measured current vector i 1 s valid. Fig. 7 SFC trajectories for all current directions in stator and dual-current coordinates. Fig. 8 Course of the isotropic component under RFC and SFC conditions with load. Fig. 9 Course of the isotropic component versus current magnitude for different machine temperatures. Fig. 10 Influence of PM temperature on the RFC and SFC trajectories. Fig. 11 Position estimation (top left) and the components of the anisotropy each shown versus time in seconds (center left and bottom left) and as a locus curve (right). Fig. 12 Position estimation (top), q-current ramp relative to rated current (center) and estimation error (bottom) versus time in seconds. Fig. 13 Estimated rotor position (top), weighting ratio (center) and estimation error (bottom) versus time in seconds.

[0012] The term "machine" is used here in the sense of an "electric machine", i.e. an electric motor or an electric generator.

[0013] The admittance Y of a synchronous machine corresponds approximately to the inverse inductance L for high-frequency excitation, both of which can be described as a matrix in the case of magnetically anisotropic behavior. Y s s ≈ L s s − 1 .

[0014] The superscript represents the coordinate system (CS), in this case stator coordinates; the subscript describes the quantity in more detail, in this case its relationship to the stator winding. Admittance describes the transformation of an exciting voltage vector. u s s into a current response Δ i s s Δi s s = Y s s u s s Δt , Δi α Δi β = Y aa Y αβ Y βα Y ββ u α u β Δt , wherein Δ t the time is when the tension u s s facility and Δ i the difference between current values i s s before and after this time. If the wave rotates, further disturbance terms are added in (2), which must be compensated for within the framework of anisotropy identification.

[0015] Anisotropy identification can alternatively also be based on inductance. L s s determine u s s = L s s Δi s s Δt , which contains equivalent anisotropy information and can therefore be assigned to a rotor position in the same way. However, this derivation is based on the example of an identified admittance. Y s s .

[0016] The goal of anisotropy identification is now to determine, from the course of the stimulating voltage, u s s (e.g. injection) and the resulting current response i s s to infer magnetic anisotropy - specifically the so-called anisotropy vector y Δ s , whose entries are a linear combination of certain components of the admittance matrix Y s s are y Δ s = y Δα y Δβ = 1 2 Y αα − Y ββ Y αβ + Y βα .

[0017] This vector y Δ s This shows a clear rotor position dependency in many machines. Similarly, two other linear combinations would be possible. Y Σ = Y αα + Y ββ 2 Y e = Y αβ − Y βα 2 , which, however, do not exhibit sufficient location dependency from the perspective of previous methods.

[0018] As the rotor rotates, the anisotropy vector typically follows trajectories similar to those in Fig. 1 are.

[0019] The small circle marks the point of each trajectory with the rotor position at 0°. The vector y Δ s Each point on the trajectory is positioned with approximately 10° rotor angle, and each trajectory is traversed twice per electric rotor rotation.

[0020] On the left is a case with no load and zero current. The anisotropy is exactly aligned with the rotor, which is why the 0° point is on the Y Δ α The axis lies. There are no anisotropy harmonics, therefore the trajectory is circular.

[0021] The diagram shows a case with a high load, i.e., a high current magnitude, where the anisotropy orientation is shifted relative to the rotor position (the 0° point is tilted) and anisotropy harmonics have formed (the trajectory is not circular). Approaches exist for compensating

[18]

[19] or exploiting [6]

[20] the shift and the harmonics, each aiming to match the measured anisotropy vector. y Δ s to assign the corresponding rotor position.

[0022] The figure on the right shows a critical anisotropy trajectory under load, as is increasingly common in modern machines with concentrated windings. The trajectory is critical because points of multiple rotor layers coincide at its peaks (ambiguity), and because the trajectory almost stagnates and reverses direction at these points (low signal component). For many methods, the fact that the angle of y Δ s The trajectory runs backwards (i.e., not monotonically) in places.

[0023] In summary, anisotropy-based methods can be divided into two stages, which are Fig. 2 The first stage is anisotropy identification, which calculates the components of the anisotropy vector from the applied voltage and the measured current response. This anisotropy vector, calculated based on the current measurement, is then referred to as the measured anisotropy vector. y ˜ Δ s denotes and contains, in contrast to the physical machine property y Δ s For example, typical measurement errors such as noise, etc. The second stage is the rotor position assignment, in which the measured anisotropy vector is assigned. y ˜ Δ s a rotor position estimate θ̂ is assigned.

[0024] The second stage of anisotropy-based methods, i.e., rotor position assignment, can lead to ambiguity and thus instability of the methods under load. This is because conventional rotor position assignment methods rely directly or indirectly on the anisotropy trajectories in Fig. 1 based on each of which is always for a specific current in rotor coordinates i s r This applies. However, in sensorless operation, the rotor coordinates are estimated, which only allows for the determination of the current magnitude. i s r = i s s known by measurement and the current angle in rotor coordinates ∠ i s r Strictly speaking, it is not known.

[0025] The anisotropy vector y Δ s However, it does not only depend on the rotor angle θ and the current magnitude. i s r off, but also in fact from the current angle ∠ i s r y Δ s = f ∠ θ , i s r , ∠ i s r

[0026] Will the angle ∠ i s r Assuming in (8) that it is uncertain / faulty, the following results for a fixed rotor position θ and a fixed current magnitude i s r the in Fig. 3 Possible locations of the anisotropy vector shown in dotted lines y Δ s In other words, the dotted trajectories result when one considers the two previously described operating points (0° point and vector). y Δ s ) the current angle in rotor coordinates ∠ i s r varies.

[0027] At zero current (left anisotropy trajectory), the current angle is irrelevant, making the dotted lines appear as circles without a radius, i.e., point-like at the 0° point or at the tip of y Δ s The rotor position can be determined unambiguously even without knowledge of the current angle.

[0028] However, when current is applied, the dotted lines run similarly to those in Fig. 3 The lines are shown in the center and on the right. As the current increases, these lines increasingly align with the anisotropic trajectory, meaning that an incorrectly assigned current (due to an estimation error) results in a progressively larger estimation error. This makes the rotor position assignment ambiguous and unstable above a certain load threshold.

[0029] A method for assigning the inductance or admittance to the rotor position of a synchronous machine, comprising a stator and a rotor with or without permanent magnets, is presented. The synchronous machine is driven by pulsed terminal voltages, and the inductance or admittance is calculated from these voltages and the measured current response. In this method, the behavior of the inductance or admittance as a function of rotor rotation, under the constraint of a current vector that remains unchanged in stator coordinates and is at least two-dimensional, is used as key information for position assignment.

[0030] The relationship (8) is not inherently ambiguous. The problem so far lies in the fact that the modeling f ∠ (·) takes place in a coordinate system (rotor) whose orientation is not known in sensorless operation (e.g., constant or directly measurable) and is therefore replaced by an assumption.

[0031] The solution to the ambiguity and stability problems lies in choosing a complete anisotropy model (i.e., dependent on one angular and at least two current dimensions) whose input variables are all measurable. For this purpose, the current vector is expressed in stator coordinates. i s s used as input y Δ s = f s θ i s s .

[0032] The anisotropy model fs(·) contains the same information as f∠(·) and both can be transformed into each other. However, this time, with the exception of θ (the later calculation result), all quantities contained in (9) are directly measurable. i s s through current sensing plus fixed Clarke transformation and y Δ s through anisotropy identification (i.e., current measurement and applied voltage) as y ˜ Δ s .

[0033] The dashed trajectories in Fig. 4 illustrate the modeling with fs (·).

[0034] While the conventional anisotropy trajectory describes the anisotropy profile for rotor-frame fixed current (RFC), the dashed lines describe the profile for stator-frame fixed current (SFC). The RFC anisotropy trajectories each show a full revolution, whereas the SFC trajectories each depict only a section of approximately ±20° around the maximum troque per ampere (MTPA) operating point.

[0035] Because it is irrelevant whether zero current is applied to the rotor or stator, the left part of Fig. 4 The SFC trajectories are identical to the RFC trajectory. The greater the current magnitude, the more the two trajectories diverge. Both trajectories always intersect at the target current point (e.g., MTPA), with the SFC trajectory becoming increasingly distorted as the current magnitude increases. The actual available (not apparent), unambiguous, and stable sensorless information content lies in this distorted direction.

[0036] Furthermore, in the right part of Fig. 4 It is evident that the ambiguous areas of the critical RFC trajectory, when considered with a stator-fixed current, assume monotonic and thus clearly assignable curves.

[0037] To obtain such a model fs (·), magnetic simulation (e.g., finite element methods) or a test bench measurement with a position sensor and load machine are suitable, for example. In the test bench measurement, the load machine can, for example, rotate the machine to be modeled, while an inverter transmits a specific SFC (System Function Control) to the machine. i 1 s imprints and an anisotropy identification procedure continuously determines the current anisotropy y ˜ Δ s calculated / measured. The course of the measured anisotropy is then determined. y ˜ Δ s together with the measured rotor position θ recorded and used for the SFC i 1 s A valid anisotropy-position relationship is stored. This process is used for a large number of other SFCs. i 2 s , i 3 s … i n s repeated until the area of ​​the current plane required for operation is captured with sufficient resolution in the stator-fixed coordinate system.

[0038] In some embodiments, the rotor position is calculated using only the measured inductance or admittance and the measured current vector in stator coordinates and the stored position assignment parameters, and the estimated rotor position is neither directly nor indirectly fed back to the rotor position assignment.

[0039] For sensorless control, the model fs (·) of a machine is inverted so that in operation with injection the measured variables y ˜ Δ s and i s s a rotor position value θ is assigned θ ^ = f s ′ y ˜ Δ s i s s .

[0040] Fig. 5 Figure 1 shows an embodiment in which (10) results in a linear signal path, which, in contrast to Fig. 2 It only uses measured quantities as input signals and does not provide feedback on the estimated rotor position.

[0041] In some embodiments, several inductance or admittance curves over rotor rotation for different stator-fixed current values ​​(a set of curves) are stored in or fed into a model, from which a currently valid curve or parameter set is selected or calculated during operation according to the current measurement, and by means of which a rotor position value is then assigned to the currently measured inductance or admittance.

[0042] One such embodiment of the rotor position assignment (10) in operation is shown in Fig. 6 To illustrate this with an example: First, the currently measured current is used. i 1 s The currently valid SFC trajectory of the stored model is clearly assigned (all other trajectories are no longer displayed). Each point on this currently valid SFC trajectory corresponds to a rotor position value, which is shown in Fig. 6 The example shows a 10° grid. The current measured value. y ˜ Δ s The result of the anisotropy identification will lie next to the SFC trajectory due to measurement errors (noise, etc.) and will therefore be assigned to a point on the SFC trajectory. The dotted lines in Fig. 6 An example assignment to the nearest point is indicated. The corresponding angle value of the assigned point is then used as a position estimate. θ used.

[0043] However, the processors commonly used in drive technology offer only limited resources in terms of computing power and storage space, of which the sensorless methods should only utilize a small portion to avoid unduly impacting other processes. Against this background, a complete, 3-dimensional machine model is needed. f s (·) (over all combinations of αProviding the required level of detail / resolution for the beta current and rotor position is not always feasible due to memory limitations, and the aforementioned search for the nearest point is not always possible due to computation time constraints. Therefore, one goal of this description is to provide a significantly simplified / reduced model or assignment rule without significantly impairing the estimation quality or stability.

[0044] In some embodiments, for example, four steps are possible to reduce resource requirements, which are described separately in the following four sections. The first step concerns storage and computing requirements, the remaining three steps only storage requirements. In total, this can achieve storage and computing requirements comparable to conventional rotor position assignment.

[0045] In some embodiments, a linear rule is used to assign the measured inductance or admittance to the rotor position, which corresponds to a projection of the measured value onto the actual rotor position dependence of the inductance or admittance linearized at the target current operating point.

[0046] In some embodiments, this corresponds, for example, to a projection of the measured value onto the tangent of the rotor position dependence of the inductance or admittance at the target current operating point.

[0047] A complete SFC trajectory y Δ s θ In contrast to RFC trajectories, this one is significantly larger, more complex in shape, and only repeats after a full electrical revolution (RFC after half a revolution). Its complete description would only be possible with many harmonics or many support points. Furthermore, as previously described, finding the nearest point in a parameter-intensive model would be computationally very intensive.

[0048] In contrast, during operation at low speeds (where anisotropy must be evaluated), the current will only be on or in the area close to the target current trajectory (e.g. MTPA or q-axis), and the modeling of significantly deviating current angles will be almost unused.

[0049] Therefore, to reduce data and computing time, for example, every complex SFC trajectory at the target operating point is linearized and thus interpreted as a straight line on which y Δ s moves linearly as the rotor rotates. y ΔL s θ ≈ m α m β θ + Y 0 α Y 0 β . 11 = m Δ s i s s θ + y Δ 0 s i s s . 12

[0050] At the target operating point, equation (12) accurately describes the magnetic behavior and how the magnetic behavior changes when leaving the target operating point. However, with increasing distance from the target operating point, equation (12) can become inaccurate and eventually false, although this has no direct impact on the estimation accuracy or stability.

[0051] For example, with this linear approximation of the SFC trajectory (9) a very simple rotor position assignment rule can now be established: The position value θ The nearest point can be determined, for example, by vector projection of the measured value. y ˜ Δ s onto the straight y ΔL s θ ^ = m Δ s T y ˜ Δ s − y Δ 0 s m Δ s T m Δ s .

[0052] The multiplication of a transpose corresponds to ( x T< ) column vector with a non-transpose, practically a scalar product.

[0053] Because in (13) the measured value y ˜ Δ s = Y ˜ Δα Y ˜ Δβ T Since the only variable is , most of the multiplications do not need to be calculated online. θ ^ = m Δ s T m Δ s T m Δ s y ˜ Δ s − m Δ s T y Δ 0 s m Δ s T m Δ s 14 = m α m α 2 + m β 2 Y ˜ Δα + m β m α 2 + m β 2 Y ˜ Δβ − m α Y 0 α + m β Y 0 β m α 2 + m β 2 . 15 = k α Y ˜ Δα + k β Y ˜ Δβ + k 0 . 16

[0054] Therefore, only 3 coefficients are sufficient for assigning the rotor position to an SFC. k α , k β and k to save 0 and use the measurement online y ˜ Δ s to be invoiced.

[0055] In some embodiments, the position assignment parameters are stored as a function of the current magnitude and the current angle. k × ( i ∥ , θ i ).

[0056] In some embodiments, the two anisotropic values ​​of inductance or admittance are transformed from the stator-fixed coordinate system into a coordinate system aligned with twice the current angle before any further use.

[0057] In machines with slight to moderate anisotropy harmonics, the SFC trajectories for different current directions exhibit a rotation with twice the electrical frequency. For example, in the left part of Fig. 7 The SFC trajectories (at the target operating point) for current vectors of the same magnitude are plotted with a 10° current angle grid, which consequently repeat after 180°.

[0058] This fact can be used for data reduction, for example by using the anisotropy vector. y Δ s It is transformed into a double-rotating coordinate system, in which it will exhibit significantly less variation over the rotor rotation. In this transformation, a clearly known / measurable angle is used instead of the rotor position.

[0059] Because the modeling fs (·) depends on the current vector anyway i s s If this occurs and is in a fixed relation to the rotor at the target operating point, then, for example, twice the current angle or double current angle is suitable as a basis for the transformation. θ i = atan 2 i β i α y Δ ii = Y Δx Y Δy = cos 2 θ i sin 2 θ i − sin 2 θ i cos 2 θ i y Δ s .

[0060] If the anisotropy vector and the SFC trajectories are represented in these double-stream coordinates according to the exemplary embodiment, the following results are obtained: Fig. 7 The curves shown on the right are now relatively similar to each other (despite the moderate harmonic components of the left curve) and can therefore be modeled with less data. Depending on the accuracy requirements, this transformation allows for a positional assignment for all current directions using only one set of coefficients. k x , k y and k 0 possible.

[0061] The procedure for calculating these transformed constants is analogous to that for the stator-fixed constants. For example, the SFC trajectory is first linearized into double-current coordinates, which determines the slope. m Δ ii and the offset y Δ 0 ii and thus the following constants for position assignment result k x = m x m x 2 + m y 2 k y = m y m x 2 + m y 2 k 0 = − m x Y 0 x + m y Y 0 y m x 2 + m y 2 − θ i θ ^ = k x Y ˜ Δx + k y Y ˜ Δy + k 0 + θ i .

[0062] In some embodiments, only position assignment parameters for one sector of the stator-fixed current coordinate system are stored, and the position assignment in the other sectors is carried out by repeating the one sector.

[0063] For example, if it is assumed that the magnetic behavior of the machine is phase-symmetric, which applies very well to almost all rotary-misaligned machines with the exception of stator-fixed anisotropy harmonics, the amount of data to be stored can be further reduced by a factor of 6.

[0064] All example images in this description were phase-symmetric, which, among other things, Fig. 7 This results in the SFC trajectories in the right part aligning again after one-sixth of an electrical revolution, meaning that only 6 SFC trajectories are visible there. Consequently, only one-sixth of an electrical revolution needs to be stored, after which the pattern repeats.

[0065] The calculation and filing of all coefficients k x , k y and k 0 occurs, for example, depending on the amount of electricity. i ∥ = i s s and current angle θ i , however, only for the current angle range θ i = [0, π / 3). In operation, these truncated data are then applied to the entire current angle range, for example, using the modulo function. θ i = [ -π, π ) shown θ i% = mod θ i , π 3 θ ^ = k x i ∥ θ i% Y ˜ Δx + k y i ∥ θ i% Y ˜ Δy + k 0 i ∥ θ i% + θ i .

[0066] This deposition and evaluation (23)-(24) is completely correct (despite 1 / 6 data) if the magnetic behavior is phase-symmetric. As mentioned previously, some machines exhibit exactly one relevant asymmetric component: the zeroth harmonic or stator-fixed anisotropy. It represents the mean value of the anisotropy vector. y Δ s above the rotor position for a specific RFC y h 0 s i s r = 1 2 π ∫ θ = 0 2 π y Δ s θ i s r dθ i s r = const . This is the center of gravity of the RFC anisotropy trajectory. In previous illustrations, this was always zero, but in practice, this is sometimes different. The result is then a two-periodic oscillation with an estimation error. e θ = θ̂ - θ , if y h 0 s is not adequately compensated.

[0067] However, compensation for stator-fixed anisotropy is also possible, for example, as an extension of the phase symmetry assumption. In some embodiments, after measuring the two anisotropic values ​​of inductance and admittance in stator-fixed coordinates, a stored vectorial displacement value is subtracted. For this purpose, y h 0 s first determined according to (25) and then subtracted both for modeling (19)-(21) and in operation (24) before transformation to double-stream coordinates y Δ ii = cos 2 θ i sin 2 θ i − sin 2 θ i cos 2 θ i y Δ s − y h 0 s .

[0068] This compensation can be carried out either constantly or depending on the amount of current, with the former usually being sufficient if y h 0 s was determined in the uncontested case.

[0069] Typically, both the rotor design and the target current trajectories (MTPA, etc.) are mirror-symmetric about the d-axis. Therefore, a load-symmetric magnetic behavior is assumed (in a specific way). This allows for an additional data reduction by a factor of 2. In some embodiments with low parasitic effects, this optional measure does not lead to a significant estimation error, so it can be implemented for the purpose of reducing the data by a factor of 2.

[0070] In this exemplary approach, for example, the coefficients are k x , k y and k0 is only intended for positive loads and is used in operation for positive loads according to (24) for position estimation. Assuming load symmetry, the following law can then be used for position estimation for negative loads. θ i% = mod − θ i , π 3 θ ^ = − k x i ∥ θ i% Y ˜ Δx + k y i ∥ θ i% Y ˜ Δy − k 0 i ∥ θ i% + θ i .

[0071] In some embodiments, the position assignment rule for negative torques is derived from the parameters for positive torques in such a way that the signs of the first anisotropic coefficient, the isotropic coefficient and the offset coefficient are each negated.

[0072] In some embodiments, the inductance or admittance is described by three values, one of which describes its isotropic component and two of which describe its stator-fixed anisotropic component.

[0073] As mentioned previously, the isotropic component shows Y Σ Under conventional considerations with rotor-fixed current (RFC), there is hardly any position dependence. Fig. 8 The continuous line shows an example of an RFC curve of Y Σ with its offset and the typically weak variation over the rotor position, which cannot be used for position estimation.

[0074] Furthermore, nine exemplary Y-Σ curves for different stator-fixed currents (SFC) are shown with dashed lines. Analogous to Fig. 7 The currents are each rotated by 10° relative to each other, resulting in the shifted waveforms. At zero current, the RFC and SFC waveforms are identical, even for the isotropic component. However, with increasing current magnitude, the slope of the SFC waveforms, i.e., their positional dependence, becomes increasingly pronounced and, at very high currents, even provides the largest information component of all signals. Furthermore, the SFC waveforms of Y Σ around the target operating point are even more reliably monotonic (i.e., for even more machines over an even wider operating range) than the SFC trajectories. y Δ ii are, which can further increase the stability and applicability of the designs described here.

[0075] The isotropic component Y Σ is, for example, in addition to y Δ ii as a third dimension of positional information and linearized at the target operating point y ΔΣ ii = Y Δx Y Δy Y Σ 29 ≈ m x m y m Σ θ + Y 0 x Y 0 y Y 0 Σ 30 = m ΔΣ ii θ + Y 0 ΔΣ ii . 31

[0076] In this exemplary embodiment, this third dimension yields the coefficients and the equation for position estimation. k x = m x m x 2 + m y 2 + m Σ 2 k y = m y m x 2 + m y 2 + m Σ 2 k Σ = m Σ m x 2 + m y 2 + m Σ 2 k 0 = − m x Y 0 x + m y Y 0 y + m Σ Y 0 Σ m x 2 + m y 2 + m Σ 2 − θ i θ ^ = k x Y ˜ Δx + k y Y ˜ Δy + k Σ Y ˜ Σ + k 0 + θ i .

[0077] Assuming load symmetry, the following equation, for example, applies for position estimation under negative loads. θ ^ = − k x Y ^ Δx + k y Y ˜ Δy − k y Y ˜ Σ − k 0 + θ i .

[0078] The assumption of phase symmetry can be based on Y Σ and thus y ΔΣ ii to be applied in the same way as previously applied y Δ ii . However, the transformation to double-stream coordinates is, as before, only applied to the first two components in y ΔΣ ii applied - k Σ remains unaffected.

[0079] By introducing the isotropic component Y While Σ significantly improves positional information, stability, and applicability, it also increases the temperature dependence of the position assignment. Specifically, the temperature of the permanent magnets (PMs) (in / on the rotor) influences the estimation results. As this temperature rises, the PM flux decreases (reversibly, by up to 15%) and reduces the saturation level of the iron. This primarily reduces the isotropic component. Y Σ, but also the magnitude of the anisotropy y Δ s The orientation of the anisotropy ∠ y Δ s The temperature is only minimally affected, which is why conventional methods for rotor position assignment exhibit only a lower temperature dependence. In contrast, the prior art methods

[23] ,

[24] , and

[25] are comparably temperature-sensitive to the embodiment described here; they fundamentally involve only a few dimensions of admittance and are methods without temperature tracking. The above description applies primarily to machines with surface-mounted PMs because their anisotropy is mainly caused by saturation. Machines with buried PMs show a lower temperature dependence, and reluctance machines (without PMs) show no relevant temperature dependence.

[0080] The following describes two exemplary approaches to temperature compensation. The first requires no additional measurement data, while the second requires measurement data from a machine operating at different temperatures. However, all exemplary approaches require multidimensional admittance information, which can only be obtained from linearly independent stress vectors in the injection pattern. This means that, in particular, the so-called alternating injection method is insufficient as a basis.

[0081] Temperature has the strongest influence on the isotropic fraction. Y Σ, which, due to the relationships, often has the greatest influence on the position assignment of the vector projection (32)-(34) at high loads. In particular, the temperature primarily changes its offset value at zero load. Y Σi 0 = Y Σ i s s = 0 , while from that point onward the relative course is almost the same. Thus, Fig. 8 This example qualitatively illustrates how the curves of Y Σ versus current magnitude on the target current trajectory are merely shifted relative to each other for a cold (solid curve) and a hot machine (dashed curve). A cold machine, for example, operates at approximately room temperature, while a hot machine operates at approximately the temperature that results from continuous operation at rated load.

[0082] The dotted curves show the change in Y Σ when the rotor is rotated left or right while SFC remains constant - i.e., the position information that simultaneously explains why an estimation error occurs when the machine is hot and estimation is performed with cold parameters.

[0083] Therefore, in some embodiments, the isotropic component of the admittance or inductance is used exclusively for position estimation at large current values ​​and for correcting a variable containing the value of the isotropic component that applies to zero current at small current values.

[0084] For small amounts of electricity i s s The situational information in Y Σ is weak, which means that the situational estimation relies mainly on y Δ s This is based on the principle that, in some embodiments, temperature-related changes in the magnetic properties of the synchronous machine are compensated in the position assignment even without measurement data on the temperature dependence by not including the measured isotropic component itself in the position assignment, but rather the difference between this measured value and the subsequently adjusted value of the isotropic component applicable at zero current.

[0085] For parameter-free temperature compensation, for example in this current range i s s < i thr The influence of the isotropic component Y Σ on the position estimation was artificially switched off. k Σ = 0 (estimation only with y Δ s ) and then, for example, the measurement of Y Σ , also Ỹ Σ is used to determine the offset value Y Σ i to track 0.

[0086] Y Σ i Therefore, 0 is a variable that is above i s s > i thr in the situation assessment (35) and (36), for example, the following must be taken into account k 0 = − m x Y 0 x + m y Y 0 y + m Σ Y 0 Σ − Y Σi 0 m x 2 + m y 2 + m Σ 2 − θ i θ ^ = k x Y ˜ Δx + k y Y ˜ Δy + k Σ Y ˜ Σ − Y Σi 0 + k 0 + θ i .

[0087] Unterhalb i s s < i thr becomes Y Σ i 0 gradually tracked using an exemplary filter law dY Σi 0 dt = ω Σ Y ˜ Σ − Y Σi 0 , where ω Σ is the filter corner frequency. In this simplest form of tracking (40), a rise in the curves within i s s < i thr not taken into account what happens with increased threshold values i thr to errors in tracking of Y Σ i 0 and therefore in the estimate of θ̂ This can lead to errors. These errors can be avoided, for example, if the increase of Y Σ is taken into account in the tracking equation (40) using an (offset-free) nth-order polynomial. dY Σi 0 dt = ω Σ Y ˜ Σ − Y Σi 0 − ∑ j = 1 n a j i s s j in which the coefficients a j The respective powers are weighted. Usually, within... i s s < i thr The second power alone already describes the course very well, which is why the following reduced and therefore still relatively computationally efficient tracking law is usually sufficient. dY Σi 0 dt = ω Σ Y ˜ Σ − Y Σi 0 − a 2 i s s 2 .

[0088] For machines with complex anisotropy behavior, which exhibit anisotropy trajectories as in Fig. 1 On the right, in addition to Y Σ, the temperature dependence of the SFC trajectories can also be shown. y Δ ii so severe that parameter-free compensation is insufficient, resulting in excessively large estimation errors and even instabilities.

[0089] Therefore, in some embodiments, different position assignment parameters, which apply to different machine temperatures, are derived from several measurement data obtained at different machine temperatures. These embodiments are used, for example, for machine types with complex anisotropy behavior.

[0090] In one embodiment, the position dependency at high machine temperature can also be determined and corresponding coefficient fields can be generated. k xh , k yh , k Σh and k 0 h about current magnitude i ∥ and -angle θ i can be stored. Because the position assignment rule (37) or (39) is completely linear, the cold (k × c ) and hot coefficients (k × h ) can simply be linearly interpolated. k × = k × c + v h k × h − k × c , which then occurs when the machine is cold ( v h = 0), with a hot machine ( v h = 1) and also in the intermediate range (0 < v h < 1) results in a correct location assignment.

[0091] For example, there are two sources of information regarding the weighting factor. v h to track changes in temperature online, which are described in the following two subsections and can be used either separately or in combination.

[0092] In some embodiments, temperature-related changes in the magnetic properties of the synchronous machine are compensated in the position assignment by using the tracked value of the isotropic component applicable at zero current to adjust the weighting between position assignment parameters that apply to different machine temperatures.

[0093] The cold and hot data include the respective zero-current offset values ​​of the isotropic component Y. Σi 0 c (cold) and Y Σi 0 c(hot), which should differ significantly due to the reversible decrease in PM flux. With these boundary parameters and one of the tracking laws (40) to (42) for measuring Y Σi It can be determined to what extent the current operating state can be assigned to which data. v h = Y Σi 0 − Y Σi 0 c Y Σi 0 h − Y Σi 0 c .

[0094] Because Y Σi Since 0 provides a relatively strong signal, these embodiments enable a relatively accurate assignment of the corresponding data set. Tracking Y Σi and thus the update of v h However, it only works in the lower current range. i s s < i thr Furthermore, for machines with complex behavior, the threshold value is... i thr for example, chosen to be low in order to be able to use Y Σ for situation estimation as early as possible.

[0095] In some embodiments, the temperature-related changes in the magnetic properties of the synchronous machine in the position assignment are compensated by using a shift of the measured inductance or admittance that is orthogonal to the direction of the position dependence of the inductance or admittance in order to adjust the weighting between position assignment parameters that apply to different machine temperatures.

[0096] Calculating the weighting factor v h Orthogonal evaluation also works above i s s > i thr This completes the tracking range across the entire target current range. However, this exemplary approach provides a lower signal strength than the isotropic component and is therefore more likely to lead to estimation errors. It should therefore only be used, for example, when the lower current range i s s < i thr was not completed for an extended period of time.

[0097] In Fig. 10 The dotted lines on the left illustrate how the RFC trajectory and two SFC trajectories shift when the machine is measured hot instead of cold.

[0098] The arrows show the specific displacement vector. Δy Δ s at the target operating point, which has a component parallel to the rotor position dependency (dashed lines) and a component orthogonal to it. The same relationship can also be found on the right side of Fig. 10 again shown in double-stream coordinates, where the directions of the displacement vectors Δ y Δ ii as expected, they are more similar to each other.

[0099] The proportion parallel to the location dependency y ΔΣ ii This is precisely the vector component that leads to an estimation error under temperature changes. It can be calculated, for example, by projection as follows. Δy ΔΣ ∥ ii = Δy ΔΣ ii T m ΔΣ ii m ΔΣ ii T m ΔΣ ii m ΔΣ ii .

[0100] The orthogonal part Δ y ΔΣ ⊥ ii For example, it lies precisely in the direction free of positional information (in three dimensions, these possible directions span a plane) in which the highest temperature dependence occurs. It is calculated, for example, by taking the actual temperature dependence into account. Δy ΔΣ ii the inherent positional dependency Δy ΔΣ ∥ ii vectorial subtraction Δy ΔΣ ⊥ ii = Δy ΔΣ ii − Δy ΔΣ ∥ ii .

[0101] If permanent admittance values ​​are now being used in the operation y ˜ ΔΣ ii If measured in a specific direction next to the SFC trajectory, then the temperature weighting factor, for example, can be derived from the orthogonal part of the deviation. v h can be derived. The calculation is performed using projection. v h = y ˜ ΔΣ ii − Y 0 ΔΣ ii T Δy ΔΣ ⊥ ii Δy ΔΣ ⊥ ii T Δy ΔΣ ⊥ ii .

[0102] Due to the typically very slow temperature change of the rotor, (47) is additionally filtered with a low-pass filter and correspondingly very low bandwidth. For example, the bandwidth is further increased by the square of the magnitude of the current temperature information ( Δy ΔΣ ⊥ ii T Δy ΔΣ ⊥ ii ) scaled to have less of an impact on tracking at operating points with little temperature information than at points with a lot of information.

[0103] Furthermore, the actual temperature of the machine is irrelevant for all approaches. The methods merely determine that the machine transitions from one magnetic behavior to another and use a correspondingly different mapping rule between admittance / inductance (i.e., magnetic behavior) and rotor position. This very direct type of compensation thus completely avoids assumptions regarding the relationship between temperature and magnetic behavior, as well as the problems and obstacles of determining the PM temperature (PMs rotate with the rotor, phase temperature measurement is only moderately common, temperature models are inaccurate and prone to interference).

[0104] All subsequent measurement results were obtained with the same, highly critical machine, which can only be stably controlled down to less than half its rated load using conventional methods. Fig. 11 shows the locus of the anisotropy vector. y ˜ Δ s and the time histories of its components at half rated load, which stagnate at the extreme points and then return to themselves. These histories cannot be clearly assigned using conventional methods, and large estimation errors and a significantly fluctuating position estimate would occur at the extreme points. θ̂ result.

[0105] The in Fig. 11 The depicted position estimate, however, shows a linear behavior that matches the actual course of events. θ̂ , so that the q-current regulated on this basis constantly generates the target torque.

[0106] Above half the rated load, conventional methods become unstable with this machine for the reasons described in

[22] . Fig. 12 However, this shows that the method described here enables a stable position estimation for all loads.

[0107] The estimation error remains within a band of ±5° (electrical position) except for brief overshoots and becomes smaller (more accurate) with increasing load.

[0108] Fig. 13 This shows the process of temperature tracking, according to the exemplary embodiments mentioned above, using cold and hot position assignment data. The cold machine is initially controlled only by hot data ( v h = 1) regulated, whereby estimation errors of almost 20° occur and the position assignment is just before instability.

[0109] At time 0, temperature tracking is activated, which changes the weighting ratio between cold and hot location assignment data within a few seconds. v h ≈ 0 is corrected.

[0110] As a result, the estimation error drops to below an error band of ±2.5°, demonstrating that, contrary to the initial state, the actual magnetic behavior of the machine was correctly detected and the position assignment adjusted accordingly. Even this critical machine can thus be controlled stably and accurately across the entire temperature range without a position sensor.

[0111] Other aspects include: (i) A device for controlling and regulating a rotating field machine, comprising a stator and a rotor, with a device for detecting a number of phase currents and with a controller for controlling the PWM converter, which is configured and designed to carry out the method as described above; and (ii) a synchronous machine, comprising a stator and a rotor with or without permanent magnets, with a device for controlling and / or regulating as described in point (i).

[0112] Today's established, highly efficient control of electric motors and generators (both machines) requires that the rotor angle be known at all times, i.e., usually measured. Without this knowledge, only significantly less efficient control methods can be used. The measurement is performed during operation using a sensor attached to the rotor shaft – the so-called rotor position sensor, or simply sensor.

[0113] Encoders bring with them a number of disadvantages, such as increased system costs, reduced robustness, increased probability of failure and greater installation space requirements, which justify the great industrial interest in obtaining the angle signal without using an encoder and using it for efficient control.

[0114] Methods that enable this are called "sensorless" or "sensorless" control and are divided into 2 classes: 1. Fundamental wave methods evaluate the voltage induced by motion, deliver very good signal characteristics at medium and high speeds, but fail at lower speeds, especially when stationary. 2. Anisotropy-based methods evaluate the position dependence of the machine's inductance, for which no rotational speed is necessary, but exhibit several problems and hurdles that explain why many applications still require a position sensor (with its disadvantages).

[0115] One of the most significant challenges of anisotropy-based methods is their load limit. When a certain torque threshold is exceeded, conventional anisotropy-based methods lose the clarity of position information, leading to instability in the control loop. The value of this torque threshold depends on the machine design and can be significantly below the rated load in modern machines (with concentrated windings).

[0116] The method described here concerns one of two components of anisotropy-based methods, namely the rotor position assignment. This is derived and implemented in a more complex way, ensuring that uniqueness under load is not lost and that the control loop is and remains generally stable.

[0117] However, this unambiguous rotor position assignment is more dependent on the rotor temperature than conventional methods. To achieve stable and sufficiently accurate control across the entire temperature range, the method is extended by temperature tracking according to certain embodiments. Temperature tracking requires at least three-dimensional admittance information, which can only be measured using linearly independent injection pulses. Literaturverzeichnis

[0118] [1] M. Schrödl, "Detection of the rotor position of a permanent magnet synchronous machine at standstill," IEEE Conf. ICEM, p. 51-56, 1988. [2] P. L. Jansen und R. D. Lorenz, "Transducerless position and velocity estimation in induction and salient ac machines," IEEE Trans. on Industrial Applications, vol. 31, p. 240-247, 1995. [3] M. J. Corley und R. D. Lorenz, "Rotor position and velocity estimation for a salient-pole permanent magnet synchronous machine at standstill and high speeds," IEEE Trans. on Industrial Applications, vol. 34, p. 784-789, 1998. [4] D. Paulus, P. Landsmann und R. Kennel, "Sensorless field- oriented control for permanent magnet synchronous machines with an arbitrary injection scheme and direct angle calculation," IEEE Conf. SLED, pp. 41-46, 2011. [5] P. Landsmann.DE, EP, WO Patent DE102015217986A1, 2015. [6] D. Paulus, P. Landsmann, S. Kuehl und R. Kennel, "Arbitrary injection for permanent magnet synchronous machines with multiple saliencies," in IEEE Conf.ECCE, Denver, 2013. [7] T. Frenzke, “Impacts of cross-saturation on sensorless control of surface permanent magnet synchronous motors,” in Conf. EPE, Dresden, 2005. [8] D. Reigosa, P. Garcia, D. Raca, F. Briz and RD Lorenz, "Measurement and Adaptive Decoupling of Cross-Saturation Effects and Secondary Saliencies in Sensorless-Controlled IPM Synchronous Machines", in IEEE Industry Applications Annual Meeting, New Orleans, 2007. [9] S. Paulus and G. Götting, "Self-sensing control of permanent-magnet synchronous machines with reluctance and saturation effects using high-frequency signal injection", in APEC, Long Beach, 2013.

[10] M. Schrödl and R. Ensbacher, "Methods and circuit arrangements for determining machine-related electromagnetic and mechanical state variables in converter-fed electrodydynamic rotating field machines". US EP DE Patent WO1992019038A1, 08 04 1992.

[11] PL Jansen and RDLorenz, "Method and apparatus for transducerless flux, position and velocity estimation in drives for ac machines". US EP JA KR CA DE Patent WO1995017780A1, 16 12 1994.

[12] M. M. Wogari, K. Ide und M. Sawamura, "Electric motor pole position sensing method, pole position sensing apparatus, and electric motor control apparatus using the same". US JP CN Patent WO2003015256A1, 05 08 2002.

[13] N. Bianchi und S. Bolognani, "Influence of rotor geometry of an interior PM motor on sensorless control feasibility," in IEEE Conf. IAS Annual Meeting, 2005. .

[14] P. Guglielmi, M. Pastorelli und A. Vagati, "Cross-Saturation Effects in IPM Motors and Related Impact on Sensorless Control," IEEE Trans. on Industry Applications, pp. 1516-1522, Nov.-Dec 2006.

[15] O. C. Ferreira und R. Kennel, "Encoderless Control of Industrial Servo Drives," in EPE / PEMC, Portoroz, 2006.

[16] P. Balazovic und R. Filka, "Motor controller for determining a position of a rotor of an AC motor, AC motor system, and method of determining a position of a rotor of an AC motor". US Patent US8217605B2, 20 08 2007.

[17] J. Liu, M. M. Harbaugh, T. A. Nondahl, P. B. Schmidt und S. Royak, "Method and apparatus for automatically identifying electrical parameters in a sensor-less PMSM". US Patent US8228013B2, 10 09 2008.

[18] C. Y. Yu, J. Tamura, D. Reigosa und R. D. Lorenz, "Position self-sensing evaluation of a FI-IPMSM based on high frequency signal injection methods," in IEEE Conf. ECCE, Phoenix, 2011.

[19] L. Chen, G. Götting, S. Dietrich und I.Hahn, "Self-Sensing Control of Permanent-Magnet Synchronous Machines With Multiple Saliencies Using Pulse-Voltage-Injection," IEEE Trans. on Industry Applications, pp. 3480-3491, July-Aug. 2016.

[20] T. S. Slininger und R. D. Lorenz, "Enhancing self-sensing estimation accuracy via negative sequence current image registration, with evaluation on a low saliency ratio machine," in IEEE Conf. SLED, Catania, 2017.

[21] B. Meier, M. Oettmeier, J. O. Fiedler und T. Bertram, "Enabling sensorless control of a permanent magnet synchronous machine in the low speed region using saturation," in IEEE Conf. IECON, Florence, 2016.

[22] W. Hammel, P. Landsmann und R. M. Kennel, "Operating point dependent anisotropies and assessment for position-sensorless control," in EPE / ECCE Europe, Karlsruhe, 2016.

[23] A. K. Jebai, F. Malrait, P. Martin und P. Rouchon, "Sensorless position estimation of Permanent-Magnet Synchronous Motors using a nonlinear magnetic saturation model," in IEEE Conf.ICEM, Marseille, 2012.

[24] A. Jebai, F. Malrait, P. Martin und R. Rouchon, "Control method implemented in a power converter and intended for identifying parameters linked to the magnetic saturation of an electric motor". US EP CN FR WO Patent US9197152B2, 29 Feb 2012.

[25] M. Roetzer, U. Vollmer, L. Chen und R. Kennel, "Anisotropy-based position estimation approach for symmetrical dual three-phase permanent magnet synchronous machines," in IEEE Conf. SLED, Catania, 2017.

Claims

1. Method for assigning the inductance or admittance to the rotor position of a synchronous machine, comprising a stator and a rotor with or without permanent magnets, the synchronous machine being driven by means of clocked terminal voltages and the inductance or admittance being calculated from these and the measured current response, the course of the inductance or admittance over the rotor rotation under the boundary condition of an unchanged, at least two-dimensional current vector in stator coordinates as key information for position assignment, characterized in that the isotropic component of the admittance (YΣ) or inductance is used for position estimation exclusively at a current magnitude greater than a threshold value, and at a current magnitude smaller than the threshold value, the isotropic component of the admittance (YΣ) or inductance is used to correct a variable that contains the value of the isotropic component (YΣ) that applies to the zero current.

2. Method according to claim 1, characterized in that the temperature-related changes in the magnetic properties of the synchronous machine in the position assignment are compensated, even without measurement data for the temperature dependence, by the fact that it is not the measured isotropic component (YΣ) itself which is entered into the position assignment, but the difference between this measured value and the tracked value of the isotropic component (YΣ) that applies to the zero current.

3. Method according to claim 2, characterized in that the temperature-related changes in the magnetic properties of the synchronous machine in the position assignment are compensated by using the tracked value of the isotropic component (YΣ), that applies to zero current, in order to adapt the weighting between position assignment parameters which apply to different machine temperatures.

4. Method according to one of Claims 1 to 3, characterized in that the temperature-related changes in the magnetic properties of the synchronous machine in the position assignment are compensated by using a shift, orthogonal to the direction of the position dependency of the inductance or admittance of the measured inductance or admittance, in order to adapt the weighting between position assignment parameters which apply for different machine temperatures.

5. Method according to claim 3 or 4, characterized in that the position assignment parameters which apply for different machine temperatures are derived from a plurality of measurement data obtained at different machine temperatures.

6. Method according to one of the preceding claims, characterized in that a plurality of courses of the inductance or admittance over the rotor rotation for respectively different stator-fixed current values are stored in a model or incorporated into a model, from which a currently valid course or parameter set is selected or calculated during operation in accordance with the current measurement, by means of which a rotor position value is then assigned to the currently measured inductance or admittance is assigned to the currently measured inductance or admittance.

7. Method according to any of the preceding claims, characterized in that the inductance or admittance is described by three values, one of which describes its isotropic part (YΣ) and two of which describe its stator-fixed anisotropic part.

8. Method according to one of the preceding claims, characterized in that only the measured inductance or admittance and the measured current vector in stator coordinates and the stored position assignment parameters are used to calculate the rotor position and in that the estimated rotor position is not fed back either directly or indirectly for rotor position assignment.

9. Method according to one of the preceding claims, characterized in that, after measuring the two anisotropic values of inductance or admittance in stator-fixed coordinates, a stored vectorial displacement value is subtracted.

10. Method according to one of the preceding claims, characterized in that the two anisotropic values of the inductance or admittance are transformed, prior to each further use, from the stator-fixed coordinate system into a coordinate system which is aligned with twice the current angle.

11. Method according to one of the preceding claims, characterized in that a linear regulation is used to assign the measured inductance or admittance to the rotor position, which corresponds to a projection of the measured value onto the actual rotor position dependence of the inductance or admittance linearized at the desired current operating point.

12. Method according to claim 11, characterized in that a projection of the measured value onto the tangent of the actual rotor position dependence of the inductance or admittance corresponds to the linear regulation.

13. Method according to one of the preceding claims, characterized in that the position assignment parameters are stored as a function of the current magnitude and the current angle.

14. Method according to one of the preceding claims, characterized in that position assignment parameters are only stored for one sector of the stator-fixed current coordinate system, and in that the position assignment in the remaining sectors is carried out by repeating the one sector.

15. Method according to one of the preceding claims, characterized in that the rule for assigning the position for negative torques is derived from the parameters for positive torques in precisely such a way that the signs of the first anisotropic coefficient, of the isotropic coefficient and of the offset coefficient are reversed in each case.

16. A device for controlling and regulating a rotating-field machine, comprising a stator and a rotor, with a controllable PWM converter for outputting clocked terminal voltages, with a device for detecting a number of phase currents and with a controller for controlling the PWM converter, which is set up and designed to carry out the method according to one of the preceding claims.

17. A synchronous machine comprising a stator and a rotor with or without permanent magnets, with a control and / or regulation device according to claim 16.