Method for encoderless rotor position determination of a rotating field machine and device for encoderless control of a three-phase motor

Abstract

Method for encoderless rotor position determination of a rotating field machine and device for encoderless control of a three-phase motor, the rotating field machine is powered by a pulse-width modulated converter and wherein the converter has model sizes for the rotor angle and the current pointer of the rotating field machine and wherein the converter has means by which at least two values ​​are measured in controlled operation which represent a measure of the local inductances of the machine, wherein the error of the model rotor angle is determined by determining at least two weighting factors depending on the model rotor angle and the model flow vector and that a weighted sum is formed from the at least two measured values ​​and the at least two weighting factors. and that a further offset value is subtracted from the sum, which is also determined depending on the model rotor angle and the model current phasor.

Description

[0001] The invention relates to a method for rotor position identification of an electric rotating field machine, comprising a rotor and a stator, wherein the rotating field machine is controlled via pulsed terminal voltages according to the method of pulse width modulation.

[0002] The invention further relates to a device for controlling and / or regulating a rotating field machine, wherein a controller is set up and configured for carrying out the described method.

[0003] The invention further relates to a rotating field machine with a stator, a rotor, and a device for its control and / or regulation. In general, the invention relates to the technical field of sensorless control and / or regulation of a rotating field machine, wherein the rotor position or rotor angle is derived from the rotor position dependence of the differential or local inductances.

[0004] The inventive evaluation of the identified local inductances with respect to the rotor position solves the stability problem of known methods in the best possible way. Determining the local inductances is not the subject of the invention. For this purpose, reference is made to known methods that are suitable for continuously determining the local inductances during operation of the rotating field machine by means of suitable injection voltages. Due to the increasing use of rotating field motors instead of DC motors, there is strong interest in high-quality control devices for three-phase motors. Dynamic and efficient control of rotating field machines can be achieved using rotor position sensors such as rotary encoders or resolvers in conjunction with pulse inverters. A disadvantage is that the use of rotor position sensors increases costs, wiring effort, and the probability of failure.

[0005] So-called encoderless, encoderless, or sensorless methods determine the rotor position during operation. The rotor position is continuously determined based on the applied voltages and measured motor currents. A key component of fundamental frequency methods is the integration of the induced voltages caused by the rotor's rotation. A disadvantage of these methods is that they fail at low speeds due to vanishing voltages. Injection methods utilize the machine's rotor position-dependent inductances by, for example, evaluating the rotor position-dependent current response to a high-frequency voltage excitation. In this process, the high-frequency voltage excitation is typically superimposed additively on the fundamental frequency voltage phasor.

[0006] A disadvantage of all known injection methods is that they do not utilize the full information content of the rotor position-dependent local inductance matrix, for example, by using only the anisotropic component or even just the direction of the anisotropic component of the entire local inductance matrix. Furthermore, these known methods make simplifying assumptions about the anisotropy. This easily leads to unstable behavior as soon as the machine exhibits a more complex anisotropic behavior that deviates from the assumptions.

[0007] In the publication "Operating Point Dependent Anisotropies and Assessment for Position-Sensorless Control," presented at the European Conference on Power Electronics and Applications, Karlsruhe, September 5-9, 2016, W. Hammel et al. demonstrate that the parameters of the local inductance matrix depend not only on the rotor position and the instantaneous torque, but are also directly influenced by the direction of the fundamental current phasor. For known injection-based encoderless control methods with simplified assumptions, this results in the condition formulated therein for stable operation: the properties of the utilized anisotropy of the motors to be operated. If state-of-the-art methods are applied to motors that do not meet this condition, unstable operation ensues.

[0008] Known injection methods can be distinguished according to the type of injection. For example, injection methods are known that additively superimpose an alternating one-dimensional injection voltage onto the fundamental voltage phasor. A simple evaluation based on this assumes that, in a permanent magnet synchronous machine, the orientation of the anisotropy coincides with the direction of the rotor's d-axis. The alternating injection voltage is chosen to be parallel to the assumed model d-axis. With correct alignment, the injection-induced AC components are parallel to the injection voltage, and thus the injection is torque-free even at large injection amplitudes. If, however, the assumed model d-axis deviates from the actual d-axis, injection-induced AC components result with an additional component perpendicular to the direction of the injection voltage.The model angle can then be adjusted according to the sign of this perpendicular current component. This provides a particularly simple way to adjust the model angle, as long as anisotropy with the assumed properties exists. However, in real machines, the anisotropy will also depend on the magnitude and direction of the fundamental current phasor and lead to the aforementioned stability problems.

[0009] Among the known injection methods are those which superimpose injection voltages in different directions onto the fundamental frequency phasor for each evaluation interval.

[0010] For example, German patent DE102015217986A1 presents an injection method in which the trajectory of the injection voltage phasor tip forms a square. Based on this, a calculation method is presented that determines the anisotropy as a two-component quantity with very little computational effort. However, even this method does not solve the problem of potential instability caused by the fundamental current dependence of the anisotropy, as demonstrated in the publication "Operating Point Dependent Anisotropies and Assessment for Position-Sensorless Control," European Conference on Power Electronics and Applications, Karlsruhe, September 5-9, 2016, by W. Hammel et al. In contrast to an injection voltage with an alternating voltage phasor, this injection scheme makes it possible to determine the complete differential inductance matrix.

[0011] The present invention solves the problem of providing an improved method for rotor position identification of an electric rotating field machine, which allows for a stable and noise-insensitive rotor position identification at all operating points for rotating field machines with arbitrary properties of the entire local inductance matrix.

[0012] The invention further comprises a corresponding device for regulating or controlling a rotating field machine.

[0013] The present invention is based on the understanding that optimal rotor position identification can only be achieved if all rotor position-dependent parameters of the entire local inductance matrix are used in the manner disclosed in the invention. To achieve stable operation, the dependence of the parameters of the local inductance matrix on the orientation of the fundamental current phasor is used in an inventive manner in the control system.

[0014] The invention requires that the complete differential inductance matrix be continuously determined by a preceding method. For example, the injection method described in DE102015217986A1 can be used.

[0015] The invention will now be explained in more detail with reference to an exemplary embodiment and the following drawings: Fig. Figure 1 shows the signal-technical relationship between the model quantities known to the converter, the actual rotor position unknown to the converter, and the measurable admittance parameters resulting depending on these quantities. Fig. Figure 2 shows the signal flow diagram of the described embodiment of the method according to the invention.

[0016] The differential inductance matrix describes the relationship between current changes and the associated injection voltage u. cThis matrix is ​​symmetric and therefore contains three independent parameters. uc=(LaLabLabLb)⋅ddtic

[0017] Conversely, the current increase as a result of an applied injection voltage u c determined by the inverse matrix. ddtic=(LaLabLabLb)−1⋅uc

[0018] The inverse Y of the inductance matrix L is often, and also in the following, referred to as the admittance matrix. It is equally symmetric and is defined by the three parameters Y. a , Y a , and Y ab certainly. Y=L−1=(LaLabLabLb)−1=(YaYabYabYb)

[0019] Thus, the relationship between the applied injection voltage u is... c and the associated increase in electricity as follows: ddtic=(YaYabYabYb)⋅uc=Y⋅uc

[0020] Using the substitutions (5a) - (5c), the admittance matrix can be decomposed as shown in (6). YΣ=Ya+Yb2 YΔa=Ya−Yb2 YΔb=Yab Y=(YaYabYabYb)=YΣ⋅(1001)+YΔa⋅(100−1)+YΔb⋅(0110)

[0021] In this decomposition, Y represents Σ the isotropic part of the admittance matrix. The anisotropic part, on the other hand, is a quantity with magnitude and direction, which in (6) is represented by its Cartesian components Y. Δa and Y Δb is represented.

[0022] The embodiment is intended to apply to a permanent magnet synchronous machine. For such a machine, the local inductance matrix or the local admittance matrix can be determined, for example, using square injection, in its decomposition according to (6). The isotropic component Y Σ can be determined from the first component of equation (31) in DE102015217986A1 as follows: YΣ=Δi∑xuc⋅Δt=14⋅uc⋅Δt⋅(Δiuα0+Δiuβ1−Δiuα2−Δiuβ3)

[0023] This can be simplified to (8), whereby the isotropic component Y Σ can be determined directly from the measured current increases. YΣ=14⋅uc⋅Δt⋅(Δiα0+Δiβ1−Δiα2−Δiβ3)

[0024] The second component Δi Σy The result of equation (31) in DE102015217986A1 is zero, apart from measurement errors.

[0025] The anisotropic components Y Δa and Y Δb The components of the vector equation (43) in DE102015217986A1 result as follows: YΔa=14⋅uc⋅Δt⋅(Δiα0−Δiβ1−Δiα2+Δiβ3) YΔb=14⋅uc⋅Δt⋅(Δiβ0+Δiα1−Δiβ2−Δiα3)

[0026] Determination of the admittance components Y Σ , Y Δa and Y ΔbThis is also possible using other injection voltage profiles, such as a rotating injection as described, for example, in the publication “A Comparative Analysis of Pulsating vs. Rotating Vector Carrier Signal Injection-Based Sensorless Control”, Applied Power Electronics Conference and Exposition, Austin, 24-28 Feb 2008, pp. 879-885 by D. Raca et al.

[0027] Determining the three parameters of the admittance matrix forms the basis for the rotor position identification method according to the invention. Of course, the implementation of the method according to the invention is not bound to the chosen form or decomposition in (6). Rather, any other representation of the information contained in the admittance matrix can be used as a basis.

[0028] In particular, it is possible to carry out the method according to the invention using any three linear combinations of the admittance components Y. a, Y b , and Y ab or Y Σ , Y Δa and Y Δb to capture, provided they are linearly independent of each other.

[0029] A key component of the invention is the appropriate use of the knowledge that the three parameters Y Σ , Y Δa and Y Δb the local admittance matrix depends not only on the rotor position θ r depend on, but also depend on the operating point of, the components i d and i q depend on the current fundamental wave current vector, i.e., are influenced by its magnitude and direction. YΣ=YΣ(θr,id,iq) YΔa=YΔa(θr,id,iq) YΔb=YΔb(θr,id,iq) where θ r the electrical angle of the rotor position and i d or i q represent the components of the fundamental wave current phasor.

[0030] The following considerations are limited to the basic speed range. Here, the machine is typically operated with a torque-generating current on the q-axis, i.e., i.e., d = 0 or along an MTPA (Maximum Torque Per Ampere) trajectory, which specifies a fixed assignment of the d-current as a function of the q-current. This is described, for example, by D. Schröder in "Electric Drives - Control of Drive Systems", 3rd ed., Berlin, Springer 2009. The machine is therefore operated according to (14) or (15). id=0 or id=id,MTPA(iq)

[0031] This limits the specification of the operating points by the converter to two remaining degrees of freedom, namely the electrical rotor angle θ. r and the q-current i q , while the associated d-current i dfrom a fixed assignment, for example according to (14) or (15) from the q-current. For operating points according to this specification, the dependence of the admittance parameter Y is also reduced. Σ , Y Δa and Y Δb reduced to only two independent quantities θ r and i q : YΣ=YΣ(θr,id(iq),iq)=YΣ(θr,iq) YΔa=YΔa(θr,id(iq),iq)=YΔa(θr,iq) YΔb=YΔb(θr,id(iq),iq)=YΔb(θr,iq)

[0032] When operating the motor without a rotary encoder on an inverter, deviations between the actual electrical rotor angle θ inevitably occur. r and the associated model rotor angle θ r,modin the inverter. Even if these are ideally very small, as described in “Operating Point Dependent Anisotropies and Assessment for Position-Sensorless Control”, European Conference on Power Electronics and Applications, Karlsruhe, 05-09 Sept. 2016 by W. Hammel et al., this can lead to unstable operation in state-of-the-art encoderless methods.

[0033] The inverter is only set to the model rotor angle θ for the orientation of the fundamental wave current vector to be impressed. r,mod can fall back on this. If this does not match the actual electrical rotor angle θ r This leads to the actual d- and q-current components no longer matching the corresponding model values. The actual d- and q-current components flowing in the motor i d and i q depend on the model sizes in the inverter i d,mod and i q,mod as well as the error of the model rotor angle θ̃r as follows: id=id,mod⋅cos(θ˜r)−iq,mod⋅sin(θ˜r) iq=iq,mod⋅cos(θ˜r)−id,mod⋅sin(θ˜r) θ˜r=θr,mod−θr

[0034] In the presence of an error in the model rotor angle θ̃ r The allocation according to (14) or (15) does not take place between the actual q- and d-current components, but rather the model d-current i is used. d,mod depending on the model q-current i q,mod educated: id,mod=0 or id,mod=id,MTPA(iq,mod)

[0035] Since the admittance parameters according to (11) - (13) depend on the actual d- and q-currents i d and i q These depend on an additional dependency on the error of the model rotor angle θ compared to (16) - (18). r θ̃ r or the model rotor angle θ r,mod exhibit: YΣ=YΣ(θr,iq,mod,θ˜r) YΔa=YΔa(θr,iq,mod,θ˜r) YΔb=YΔb(θr,iq,mod,θ˜r) or YΣ=YΣ(θr,iq,mod,θr,mod) YΔa=YΔa(θr,iq,mod,θr,mod) YΔb=YΔb(θr,iq,mod,θr,mod)

[0036] Thus, the measurable admittance parameters Y depend Σ , Y Δa and Y Δb both of the model sizes known in the inverter i q,mod and θ r,mod as well as from another quantity unknown to the converter, namely the actual rotor angle θ r away.

[0037] In Fig. Figure 1 shows the internal interaction of the essential equations, which together lead to the dependencies (27) - (29). The numbers of the equations, which represent the respective relationships, are given in parentheses. The model d-stream i d,mod arises from the model q-stream i q,modaccording to the selected MTPA characteristic curve 103 according to equations (22) and (23). The actual motor current components i d and i q result from the model current components i d,mod and i q,mod by transforming model rotor coordinates into actual rotor coordinates 102 according to equations (19) and (20) using the error angle θ̃ r The error of the model rotor angle θ̃ r is according to (21) the difference between the model rotor angle θ r,mod and the actual rotor angle θ r Finally, within the rotating field machine, the measurable admittance parameters Y are determined. Σ , Y Δa and Y Δb depending on the actual current components i d and i q as well as the actual rotor position θ r according to equations (11) - (13) by the Fig. educated.

[0038] Overall, this results in a Fig. the model size of the q-stream i q,mod and the model rotor angle θ r,mod as well as the actual rotor angle θ r on the measurable admittance parameters Y Σ , Y Δa and Y Δb according to equations (27) - (29).

[0039] The crucial finding is that while an error in the model rotor angle may affect the admittance parameters, these can still be measured accurately using an injection method.

[0040] Based on this finding, it is fundamentally possible to use the measured admittance parameters for rotor position identification. This would be very easy to implement if one of the relationships (27) - (29) were uniquely reversible with respect to the rotor angle θ. r could be resolved. However, this is generally not the case for any of the three quantities.

[0041] In any case, it is necessary to know the dependencies of the admittance parameters on the operating point according to (11) - (13). These can be determined, for example, by a prior offline measurement, in which case measuring instruments for rotor position determination can also be used. However, to implement the method according to the invention, it is not necessary to determine these parameters over the entire dq current plane. If the machine is operated on a current trajectory according to (22) or (23) and it is further assumed that the error angles occurring during operation remain small, it is sufficient to determine the admittance parameters on the current trajectory and in its vicinity.

[0042] According to the invention, the stability problem described by W. Hammel et al. in “Operating Point Dependent Anisotropies and Assessment for Position-Sensorless Control”, European Conference on Power Electronics and Applications, Karlsruhe, September 5-9, 2016, is solved by first generating an internal inverter error signal δ F from the quantities available to the inverter θ r,mod and i q,mod , as well as the measured admittance parameters Y Σ , Y Δa and Y Δb is formed, meaning that this error signal itself is only known to the inverter from the model q-current i. q,mod and the model rotor angle θ r,mod as well as the unknown rotor angle θ r depends: δF=δF(θr,iq,mod,θr,mod)

[0043] According to the invention, this signal is generated such that it provides a measure of the deviation of the model rotor angle θ. r,mod from the actual rotor angle θ ris, and this signal is fed to a controller which controls the model angle θ r,mod The rotor angle is adjusted accordingly. This can be achieved, for example, using a simple PLL control loop. Alternatively, the error signal δ can be used. F It can be used as a corrective intervention in a fundamental wave model, which thereby also becomes applicable in the low speed range and at standstill.

[0044] According to the invention, the error signal δ F according to Fig. 2 is generated by first forming a sum signal F, which is a weighted sum of the three measured admittance parameters: F=GΣ⋅YΣ+GΔa⋅YΔa+GΔb⋅YΔb

[0045] A quantity F0 is subtracted from this sum signal F, resulting in the error signal δ. F results in: δF=F−F0

[0046] The weights are G Σ , G Δa and G ΔbFurthermore, the quantity F0 is generally not a constant but rather an operating point-dependent value. It is noteworthy that it is not necessary to determine the actual operating point θ to determine these quantities. r , i d and i q to use. Rather, the use of the - possibly erroneous - model operating point θ leads to r,mod and i q,mod Nevertheless, the result is stable operation, even with a non-negligible error angle.

[0047] Thus, the sizes G Σ , G Δa , G Δb and F0 general functions of model sizes: GΣ=GΣ(iq,mod,θr,mod) GΔa=GΔa(iq,mod,θr,mod) GΔb=GΔb(iq,mod,θr,mod) F0=F0(iq,mod,θr,mod)

[0048] Depending on the shape of these functions G Σ , G Δa , G ΔbFor storing or calculating F0 in the inverter, a tabular or functional representation, or a combination of both, is appropriate. Furthermore, a tabularly stored dependency of the values ​​G is assumed. Σ , G Δa , G Δb and F0 is assumed to be from the model operating point.

[0049] For the error signal δ F This results in the fact that this depends on the model sizes i q,mod and θ r,mod as well as the actual rotor angle θ r depends: δF=δF(θr,iq,mod,θr,mod) Crucial to the invention is the realization that it is possible to modify the functions G Σ , G Δa , G Δb and to design F0 in such a way, solely depending on the model parameters, that the error signal δ F It acquires the following properties ∂∂θrδF(θr,iq,mod,θr,mod)|θr,mod=θr=1 ∂∂θr,modδF(θr,iq,mod,θr,mod)|θr,mod=θr=−1 δF(θr,iq,mod,θr,mod)|θr,mod=θr=0 and this, due to the aforementioned properties, enables stable operation.

[0050] The required property according to equation (38) therefore states how the error signal δ F starting from the regulated operating point θ r,mod = θ r with a fixed model angle θ r,mod on a change in the actual rotor angle θ r to react, namely with a slope of 1 when the actual rotor angle θ changes. r ·

[0051] The required property according to equation (39) further states how the error signal δ F starting from the regulated operating point θ r,mod = θ r with a fixed actual rotor angle θ r on a change in the model angle θ r,mod to react, namely with a slope of -1 when the model angle e changes. r,mod .

[0052] This will reduce the error signal δF near the regulated operating point θ r,mod = θ r proportional to the error angle θ̃ r = θ r,mod - θ r and is therefore suitable for adjusting the model angle to the actual motor angle using a control loop.

[0053] From the required properties (38) and (39) of the error signal δ F It also follows that for all regulated operating points θ r,mod = θ r independent of the rotor position θ r and the model q-current i q the value of the error signal δ F is constant, e.g. as chosen (40), constant zero.

[0054] In a further inventive step, G Σ , G Δa , G Δb and F0 as a function of the model parameters i q,mod and θ r,mod designed so that the error signal δ F according to (31) and (32) exhibits the required properties according to (38) - (40).

[0055] This is achieved by G Σ , G Δa , G Δb and F0 are executed as follows: GΣ(iq,mod,θr,mod)=DΣ(iq,mod,θr,mod)DΣ2(iq,mod,θr,mod)+DΔa2(iq,mod,θr,mod)+DΔb2(iq,mod,θr,mod) GΔa(iq,mod,θr,mod)=DΔa(iq,mod,θr,mod)DΣ2(iq,mod,θr,mod)+DΔa2(iq,mod,θr,mod)+DΔb2(iq,mod,θr,mod) GΔb(iq,mod,θr,mod)=DΔb(iq,mod,θr,mod)DΣ2(iq,mod,θr,mod)+DΔa2(iq,mod,θr,mod)+DΔb2(iq,mod,θr,mod) F0(iq,mod,θr,mod)=DΣ(iq,mod,θr,mod)⋅YΣ(iq,mod,θr,mod)+DΔa(iq,mod,θr,mod)⋅YΔa(iq,mod,θr,mod)+D Δb(iq,mod,θr,mod)⋅YΔb(iq,mod,θr,mod)DΣ2(iq,mod,θr,mod)+DΔa2(iq,mod,θr,mod)+DΔb2(iq,mod,θr,mod)

[0056] In this context, D Σ , D Δa and D Δb the differentials of the local admittance parameters Y Σ , Y Δa and Y Δb after the rotor position θ rIf the dependencies of the local admittance parameters are determined according to (27) - (29) as described above, then their differentials D can also be determined. Σ , D Δa and D Δb after the rotor position θ r Specify the regulated operating point: DΣ(iq,mod,θr,mod)=∂∂θrYΣ(θr,iq,mod,θr,mod)|θr=θr,mod DΔa(iq,mod,θr,mod)=∂∂θrYΔa(θr,iq,mod,θr,mod)|θr=θr,mod DΔb(iq,mod,θr,mod)=∂∂θrYΔb(θr,iq,mod,θr,mod)|θr=θr,mod

[0057] If the alternative representation according to (24) - (26) is used to describe the admittance parameters, i.e., depending on the actual rotor angle θ r , of the model q stream i q,mod and the error angle θ̃ r , the relevant differentials are represented as follows: DΣ(iq,mod,θr,mod)=[∂∂θrYΣ(θr,iq,mod,θ˜r)−∂∂θ˜rYΣ(θr,iq,mod,θ˜r)]θr=θr,mod,θ¯r=0 DΔa(iq,mod,θr,mod)=[∂∂θrYΔa(θr,iq,mod,θ˜r)−∂∂θ˜rYΔa(θr,iq,mod,θ˜r)]θr=θr,mod,θ¯r=0 DΔa(iq,mod,θr,mod)=[∂∂θrYΔa(θr,iq,mod,θ˜r)−∂∂θ˜rYΔa(θr,iq,mod,θ˜r)]θr=θr,mod,θ¯r=0

[0058] The determination of the values ​​G chosen according to (41) - (44) in conjunction with (45) - (47) or with (48) - (50) Σ , G Δa , G Δb and F0 not only fulfills conditions (38) - (40) for the fault signal but also yields the best possible signal-to-noise ratio for the fault signal δ F , assuming that the measured values ​​of the admittance parameters Y Σ , Y Δa and Y Δb Uncorrelated, normally distributed noise with the same standard deviation.

[0059] Naturally, the invention also includes provisions that deviate from this. For example, the above provision can be deviated from in the following variants: 1. A setting of G that is not optimal for the signal-to-noise ratio Σ , G Δa , G Δb and F0, so that the properties (38) - (40) for the error signal δ F are fulfilled. 2. A setting such that the error signal δ F For different noise characteristics of the measured admittance parameters, noise optimization is achieved. For example, the measured admittance parameters may exhibit noise with different standard deviations, or the noise of the individual admittance parameters may not be uncorrelated but correlated with each other, or the admittance parameters may follow a distribution other than the normal distribution. Even under these conditions, there is a specification for the dependence of the values ​​G. Σ , G Δa , G Δb and F0 of the model quantities for which the error signal δ F exhibits an optimal signal-to-noise ratio. 3. A definition such that the derivatives in (38) - (39) are not constant + / -1 but deviate from this or even vary depending on the operating point. In this case, the control loop that tracks the modeled rotor position results in an operating point-dependent control loop gain and thus an operating point-dependent transient response. 4. A definition where a single weight, e.g. G, is used. Σ The value is chosen to be smaller in magnitude or even zero. This would be useful, for example, if the corresponding admittance parameter exhibits significant variations between individual units and different units from an ensemble of identical motors are to be operated with a single parameter set. In this case, the individual units of the ensemble differ with respect to the admittance parameter in question due to manufacturing tolerances. 5. Determining the values ​​G Σ , G Δa , G Δband F0 as a function of three model parameters instead of the two model parameters described here, such as as a function of the model rotor angle and both model current components. This would be advantageous if the machine is to be operated not only along a fixed current trajectory but in a larger area of ​​the dq current plane or in the entire dq current plane, as is applied, for example, in the field weakening regime.

[0060] In summary, the following steps are required to carry out the method according to the invention in the present embodiment. First, the following steps are performed in a preliminary offline process: 1. Determination of the operating point dependence of the local admittance parameter Y Σ , Y Δa and Y Δbalong the defined current trajectory and in its vicinity. This can advantageously be done offline for a single example of a motor type on a test bench with rotor position measuring devices. 2. Determination of the differentials of the offline measured admittance parameters with respect to rotor position. 3. Determining the table contents for the weight factors G Σ , G Δa and G Δb as well as the term F0 for all operating points.

[0061] The subsequent encoderless rotor position determination, performed online, comprises the following steps, as shown in Fig. 2 shown: 1. Measurement of the local admittance parameters Y Σ , Y Δa and Y Δb using a suitable high-frequency injection voltage. 2. Determination of the current values ​​for the weight factors G Σ , G Δa and G Δb by depending on the current model parameters iq,mod and θ r,mod The previously defined tables are accessed. 3. Formation of a weighted sum of the measured admittance parameters using the current weighting factors. 4. Determining the current value for the term F0 to be subtracted, depending on the current model parameters. 5. Formation of the error signal δ F by subtracting the term F0 from the weighted sum. 6. Supplying the fault signal δ F into a control loop or a fundamental wave machine model. 7. Tracking the model rotor angle θ r,mod , by the error signal δ F is regulated to zero. 8. Cyclical repetition of online steps 1-7.

[0062] In Fig. Figure 2 shows the signal flow diagram for the encoderless control of a rotating field machine 111 using the method according to the invention. The rotating field machine 111 is powered by the power output stage 109 of an inverter. The currents flowing to the rotating field machine are measured via a two- or three-phase current sensing device 110.

[0063] Depending on the target q-current i, which depends on the desired torque q,soll The corresponding target d current i is determined according to (14) or (15). d,soll determined via the MTPA characteristic curve 103 and the resulting target current vector in model rotor coordinates isollr The target-actual comparison 104. The actual current phasor in model rotor coordinates. imodr is created by the inverse transformation 107 from the actual current phasor in stator coordinates i s using the model rotor angle θ r,mod .

[0064] The current regulator 105 generates the fundamental frequency voltage in model rotor coordinates. ufr and thus guides the actual current pointer imodr the target current pointer isollr The fundamental frequency voltage is transformed from model rotor coordinates to stator coordinates using the transformation device 106, for which the model rotor angle θ is used. r,mod is used. The fundamental frequency phasor in stator coordinates. ufs The injection voltage will be ucs by means of summation 108 additively superimposed, whereby the total motor voltage in stator coordinates about This is generated, which is amplified by the power output stage 109 and supplied to the machine 111. Alternatively, the injection voltage can also be added before the transformation 106 into model rotor coordinates.

[0065] The currents flowing in the machine are measured by the current sensing unit 110. From this, the fundamental waveform current in stator coordinates i is determined in the separation unit 112. s as well as the admittance parameters Y from the high-frequency current components Σ , Y Δa and Y Δb determined.

[0066] The weights G Σ , G Δa and G Δb The weighted sum F is calculated using tables or functional formulas. Fig. depending on the model rotor position θ r,mod and the model q stream i q,mod educated.

[0067] Finally, the offset F0, which also depends on the model rotor position θ, is subtracted from the calculated weighted sum F. r,mod and the model q stream i q,mod in the table or functional Fig. This is generated. This ultimately creates the error signal δ. F, which in the present embodiment is fed to a PLL controller 119. This controller typically consists of a series connection of a PL element 117 and an I element 118. The PLL controller applies the model rotor angle θ generated at its output. r,mod the actual rotor angle θ r after, so that it finally matches the actual rotor angle in the regulated state and the error signal δ F then becomes zero. At the output of the PI element 117, there is also a model value of the electric angular velocity ω. mod available, which can be used, for example, as an actual value for a superimposed speed control loop. Reference symbol list 100 Formation of the admittance parameters depending on the model sizes and the actual rotor position 101 Formation of admittance parameters depending on the actual motor sizes 102 Transformation of model rotor coordinates into actual rotor coordinates 103 MTPA characteristic curve (Maximum Torque Per Ampere) 104 Target-actual comparison of the current control loop 105 current regulators 106 Transformation of model rotor coordinates to stator coordinates 107 Inverse transformation of stator coordinates into model rotor coordinates 108 Additive switching of the injection voltage 109 Power output stage 110 Current measurement 111 Rotating field machine 112 Means for determining the admittance parameters and the fundamental current 113 Formation of the weight factor for the isotropic admittance component 114 Formation of the weight factor for the anisotropic α-admittance component 115 Formation of the weighting factor for the anisotropic β-admittance component 116 Calculation of the offset to be deducted 117 PL controller of the PLL controller for generating the model rotational speed 118 I-controller of the PLL controller for forming the model rotor angle 119 PLL controller (phase-locked loop) List of symbols D Δa Differential of the anisotropic admittance a-component with respect to rotor position D Δb Differential of the anisotropic admittance-b component with respect to rotor position D Σ Differential of the isotropic admittance component with respect to rotor position F Weighted sum of the measured admittance components F0 Offset to be deducted G Δa Weighting factor for the α-component of the anisotropic admittance fraction G Δb Weighting factor for the b-component of the anisotropic admittance fraction G Σ Weighting factor for the isotropic admittance fraction i c Carrier current pointer i d , i qActual fundamental wave current components in rotor coordinates i d,mod , i q,mod Actual fundamental wave current components in model rotor coordinates Actual fundamental wave current vector in model rotor coordinates i s Actual fundamental wave current phasor in stator coordinates i d,soll , i q,soll Target current components of the fundamental wave current Target current vector of the fundamental wave current in model rotor coordinates Di αn , Δi βn Components of current increases in stator coordinates Di uαn , Δi uβn Components of current increases in voltage coordinates Di Σx,y Current increases due to the isotropic admittance component L Inductance matrix in stator coordinates L a a-component of the inductance matrix in stator coordinates L b b-component of the inductance matrix in stator coordinates L abCoupling inductance in stator coordinates Δt time interval u c Amplitude of the injection voltage u c Injection voltage indicator Injection voltage vector in stator coordinates Fundamental wave voltage vector in model rotor coordinates Fundamental wave voltage vector in stator coordinates Machine voltage in stator coordinates y Admittance matrix in stator coordinates Y a a-component of the admittance matrix in stator coordinates Y b b-component of the admittance matrix in stator coordinates Y ab Coupling admittance in stator coordinates Y Δa a-component of the anisotropic admittance in stator coordinates Y Δb b-component of the anisotropic admittance in stator coordinates Y Σ Isotropic admittance fraction δ F Error signal θr Rotor position θ r,mod Model rotor position θ̃ r Model rotor position error QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] DE 102015217986A1 [0010, 0014, 0022, 0024, 0025] Cited non-patent literature

[0000] Operating Point Dependent Anisotropies and Assessment for Position-Sensorless Control," European Conference on Power Electronics and Applications, Karlsruhe, September 5th - 9th, 2016 by W.Hammel [0007, 0010, 0032, 0042] A Comparative Analysis of Pulsating vs. Rotating Vector Carrier Signal Injection-Based Sensorless Control," Applied Power Electronics Conference and Exposition, Austin, February 24-28, 2008, pp. 879-885 by D. Raca

[0026] Electric Drives - Control of Drive Systems,“ 3rd ed., Berlin, Springer 2009 by D. Schröder

[0030]

Claims

[1] Method for encoderless rotor position determination of a rotating field machine, the rotating field machine is powered by a pulse-width modulated converter and wherein the converter has model sizes for the rotor angle and the current pointer of the rotating field machine and wherein the converter has means by which at least two values ​​are measured in controlled operation which represent a measure of the local inductances of the machine, characterized by , that The error of the model rotor angle is determined by determining at least two weighting factors depending on the model rotor angle and the model current vector. and that a weighted sum is formed from the at least two measured values ​​and the at least two weighting factors. and that a further offset value is subtracted from the sum, which is also determined depending on the model rotor angle and the model current phasor. [2] Method according to claim 1, characterized by , that local admittances are used as a measure of local inductances. [3] Method according to claim 1 or 2 characterized by , that in a step preceding the position determination, it is determined once how the weight factors and the offset value are assigned to the model sizes during the position determination. [4] Method according to claim 3, characterized by , that to determine the assignment, the local inductances or local admittances are determined depending on the rotor position and the current phasor. [5] Method according to claim 4, characterized by, that the assignment of the weight factors and the offset value to the model quantities depends on the rotor position-related differential of the local inductances or local admittances. [6] Method according to claim 5, characterized by , that each of the at least two weighting factors is formed as the quotient of the differential of the associated local inductance or admittance and the squared sum of all differentials. [7] Method according to any one of claims 3 to 6, characterized by , that the assignment for the offset value to be subtracted is set in such a way that it matches the weighted sum formed later in online operation when the actual rotor angle matches the model angle. [8] Method according to any of the preceding claims, characterized by, that the weighting factors are chosen in such a way that the admittance parameter which has the strongest instance variations is given a lower weight or is not taken into account in the weighted sum. [9] Device for encoderless control of a three-phase motor, which is supplied by a pulse inverter, which is designed to carry out the control according to a method of the preceding claims.

Images