Method for determining the position of a rotor of an electric machine

Abstract

The invention relates to a method for determining the position of a rotor of an electric machine without a rotor position sensor, wherein an anisotropy-based open-loop and / or closed-loop position control algorithm with high-frequency excitation of the voltage modulation is applied for the purposes of estimating a rotor angle, wherein, furthermore, a signal containing measurement noise and having an interference-minimized tangent is calculated using a least squares algorithm (6) that is capable of identifying high-frequency current derivatives, wherein, furthermore, an angle control loop (2) is used to minimize occurring interference and step changes in the rotor angle, wherein an amplitude-optimized high-frequency injection is furthermore used.

Description

[0001] Method for determining the position of a rotor of an electrical machine

[0002] The invention relates to a method for determining the position of a rotor of an electrical machine according to the preamble of claim 1.

[0003] The operation of an electrical machine, for example a permanent magnet synchronous machine in a radial or axial arrangement or a current-excited synchronous machine, is achieved by regulating a current in the electrical machine by setting a voltage. A well-known current control concept is field-oriented control, which transforms the current and voltage (Clarke-Park transformation) into a rotor-fixed coordinate system based on the position of the permanent magnet. The position of the permanent magnet (also rotor angle) is usually determined using a rotor angle sensor (e.g. resolver). To save costs and space, the sensorless control concept makes it possible to dispense with the sensor and determine the rotor angle solely based on knowledge of other sensor and machine parameters.

[0004] A well-known method for determining the rotor position is to use methods that utilize the anisotropy of the electric machine. The degree of anisotropy R L is determined via the inductances in the rotor-fixed coordinate system L d and L q determined and changes depending on the operating point due to effects such as magnetic saturation:

[0005] For determining the rotor angle using anisotropy, a variety of methods are known from the literature that involve injecting an additional signal. The signal can be embodied as a sine or square wave and injected as a voltage injection into the rotor- or stator-fixed coordinate system. Furthermore, methods that propose current injection into the rotor-fixed coordinate system are known. What all of these methods have in common is that the injection introduces additional high-frequency losses and additional torque ripples into the electric machine, which negatively impact efficiency and the noise, vibration, and harshness (NVH) behavior. The goal is therefore to keep the injection as low as possible while maintaining a good and robust estimate of the rotor angle.

[0006] In addition to traditional high-frequency injection, it is known from the state of the art that, in addition to a dedicated high-frequency injection, the modulation of the voltage with the switching frequency can also be used as high-frequency injection. The challenge with this methodology lies in the precise identification of the current derivatives within specific phasors of a switching period. In academic literature, this identification is primarily ensured by additional prototype sensors, which are not available in the automotive environment. If the derivatives can be determined with a certain degree of accuracy, various options exist for extracting the rotor angle-dependent term. One well-known method is so-called arbitrary injection [D. Paulus, P. Landsmann and R. Kennel, “Saliency-based sensorless field-oriented control for permanent magnet synchronous machines in the whole speed range”, 3 rdIEEE International Symposium on Sensorless Control for Electric Drives, 2012], This divides the total current discharge into the parts of anisotropic high-frequency component e" ß , isotropic high-frequency component and low-frequency component ^i“p. All following considerations are made in the stator-fixed coordinate system (aß):

[0007] Where R s the stator resistance, v" ß the counter-induced voltage, Y z the isotropic (average) admittance, Y A the anisotropic admittance, Y s the admittance matrix, v" ß the tension, i" ß the current in the respective pointer and S(0 R) a transformation matrix depending on the rotor angle. Using the stator-fixed voltage and the extracted anisotropic high-frequency current derivatives, the sinusoidal signal is then calculated, which, when harmonic anisotropies are neglected and cross-coupling effects are correctly compensated, has the same phase position as the rotor angle:

[0008] With the subsequent use of an angle control loop, also known as Angle-Tracking Observer (ATO), [J. Friedmann, R. Hoffmann and R. Kennel, “A new approach for a complete and ultrafast analysis of PMSMs using the arbitrary injection scheme”, IEEE Symposium on Sensorless Control for Electrical Drives, 2016], occurring vibrations in the

[0009] Rotor angle is minimized. The influence of cross-coupling is controlled by a

[0010] Current dependent table based on the inductances in d- and q-direction and the cross-coupled inductance L dq (i d ,i q ) obtained from characterization data. The compensation term is calculated as follows:

[0011] 2Lqq(iq,iq)

[0012] ^COMP — 0.5 * atan Lq(id,iq) _ Lq(iq,iq)

[0013] Furthermore, it is known from the literature that high-frequency current derivatives can be identified using a least-square (LSQ) algorithm [A. Liske, S. Decker and M. Braun, “Resource optimal FPGA implementation of a least-squares estimator for fast and robust online measurement of current slope and absolute value”, 21st European Conference on Power Electronics and Applications, 2019]. The basic idea is that the current curve can be approximated as a linear equation f(t):

[0014] By solving the linear regression equation and transforming it using the arithmetic mean of an incremental index and the square pyramid number, the equation can be simplified so that it can be easily implemented on a freely programmable logic gate array (FPGA): where f(t) is the approximation function, i k the incoming measured values, T A the sampling interval and N the number of sampled values ​​within the desired pointer T o , T2 or T7. Furthermore, it is known from the literature that at low modulation levels, a reconstruction of the active switching states can be applied [P. Landsmann, J. Jung, M. Kramkowski, P. Stolze, D. Paulus and R. Kennel, “Lowering injection amplitude in sensorless control by means of current oversampling”, 3 rdIEEE International Symposium on Sensorless Control for Electric Drives, 2012], For this purpose, a starting point of the first active phasor and an end point of the second active phasor are extrapolated (backward) from the surrounding zero phasors. The resulting mean active phasor exhibits the same behavior regarding current dissipation and applied voltage as the individual phasors considered separately.

[0015] DE 102019208497 A1 describes a method for determining the rotor position of an electrical rotating machine powered by a PWM-controlled inverter. Injection voltages determined by a controller, which consist of predetermined voltages and high-frequency voltages, are converted into corresponding PWM duty cycles, and the inverter is controlled accordingly with these PWM duty cycles. Current waveforms of phase currents are then determined by measuring at least a first phase current and a second phase current, with the measurement taking place within a PWM period in the last third of a passive switching state. The rotor position is then determined based on the determined current waveforms and the injected high-frequency voltages.

[0016] The invention is based on the object of providing a novel method for determining the position of a rotor of an electrical machine.

[0017] The object is achieved according to the invention by a method for determining the position of a rotor of an electrical machine having the features of claim 1.

[0018] Advantageous embodiments of the invention are the subject of the subclaims.

[0019] In a method according to the invention for determining the position of a rotor of an electric machine without a rotor position sensor, an anisotropy-based position control and / or regulation algorithm with high-frequency excitation of the voltage modulation is used to estimate a rotor angle. According to the invention, a signal subject to measurement noise with a minimally disrupted slope tangent is calculated using a least-squares algorithm capable of identifying high-frequency current derivatives. Furthermore, an angle control loop is used to minimize occurring disturbances and jumps in the rotor angle, and amplitude-optimized high-frequency injection is used.

[0020] The present invention provides a method for operating and monitoring a permanent magnet synchronous machine without a rotor position sensor. This enables an anisotropy-based estimation of the rotor position optimized with regard to NVH (noise, vibration, harshness) and efficiency without a rotor position sensor.

[0021] The present invention combines methods for determining the rotor position without a sensor in electrical machines. In particular, high-frequency excitations of the voltage modulation are utilized by any injection, calculated from a signal subject to measurement noise using an LSQ algorithm on a fast processing unit, and / or disturbances and jumps in the rotor angle occurring in an angle control loop are minimized. These aspects are combined, and an amplitude-optimized high-frequency injection is additionally integrated in such a way that an anisotropy-based determination of the rotor angle is possible with appropriate accuracy and robustness, even at standstill.

[0022] According to the invention, an anisotropy-based position control and / or regulation algorithm with sensorless modulation excitation is used, together with a least-squares algorithm capable of identifying high-frequency current leakages. Additionally, an optimized injection scheme is proposed to ensure stable and robust estimation at low engine speeds and when the engine is at a standstill. All three functionalities together enable the algorithm to be applied in a motor vehicle engine control unit.

[0023] The elimination of a rotor position sensor offers advantages in terms of cost, installation space and diagnostic effort.

[0024] Embodiments of the invention are explained in more detail below with reference to drawings.

[0025] Shown are: Fig. 1 a schematic view of a method for estimating the rotor position for feedback to the field-oriented control,

[0026] Fig. 2 is a schematic view of a method for estimating the rotor position for redundant monitoring of the rotor position,

[0027] Fig. 3 is a schematic view of characteristic maps for the operating point-dependent selection of frequency and amplitude of an injection,

[0028] Fig. 4 is a schematic diagram of voltages and currents of an inverter for operating the electrical machine, and

[0029] Fig. 5 is a schematic view of an angle control loop.

[0030] Corresponding parts are provided with the same reference numerals in all figures.

[0031] As described in the prior art, there are ways to estimate the rotor angle based on anisotropy by utilizing the high-frequency excitations of the voltage modulation. One known method for this is arbitrary injection. Furthermore, it is known that a signal subject to measurement noise can be calculated using an LSQ algorithm 6 on a fast computing unit (e.g., an FPGA) in such a way that a slope tangent with minimal disturbances can be calculated. Finally, the concept of the angle control loop 2 is known, which minimizes disturbances and jumps in the rotor angle.The core of the invention is to implement a novel combination of the individual aspects not previously included in the prior art and to additionally incorporate an amplitude-optimized high-frequency injection in such a way that an anisotropy-based determination of the rotor angle is possible with corresponding accuracy and robustness right down to standstill. For this purpose, a distinction is made between Case 1: Estimation of the rotor position for feedback to the field-oriented control (Figure 1) and Case 2: Estimation of the rotor position for redundant monitoring of the rotor position (Figure 2). Depending on the requirements of the electric drive and other framework conditions, either one of the two methods or a combination of the methods is used. Figure 1 is a schematic view of a method for estimating the rotor position for feedback to the field-oriented control.Figure 2 is a schematic view of a rotor position estimation method for redundant rotor position monitoring.

[0032] While in Case 1 a separate operating point-dependent high-frequency injection (Figure 3) is used to ensure a good and robust estimation of the rotor position at low modulation, in Case 2 no separate injection is used, as an existing high-frequency injection from another algorithm can be used. The injection in Case 1 is rectangular in shape in order to achieve an approximately constant modulation level, in contrast to other shapes (e.g. sinusoidal). Figure 3 is a schematic view of characteristic maps KF for the operating point-dependent selection of frequency f and amplitude ai of the injection in Case 1. The operating point-dependent selection of frequency f and amplitude ai of the injection in Case 1 is dependent on torque M, speed n, DC voltage VDC and the switching frequency f sand the FM driving mode using at least one KF map. By including the FM driving mode, a distinction can be made between injection options that optimize efficiency, comfort, and performance.

[0033] Figure 4 is a schematic diagram of voltages v a and streams i a an inverter 1 for operating the electrical machine.

[0034] The information about the rotor angle is located in the anisotropic high-frequency component of an active phasor Ti, T2, and is extracted by subtracting the other components, as described in the prior art. The active phasor, or T2, refers to the voltage state in which inverter 1 is not in freewheeling mode. This includes the six switching states 100, 110, 010, 011, 001, and 101, where 1 represents a closed switch and 0 represents an open switch in the upper part of a half-bridge. always denotes the first active pointer of a new switching period and T2 the second. The free-running states are defined as zero pointers (000 or 111), where T o always denotes the zero vector at the beginning of a new switching period, and T7 denotes the zero vector in the middle of a switching period. The isotropic high-frequency component is calculated by multiplying the calculated voltages in the active vector Ti, T2, and a mean admittance characteristic map as a function of d- and q-current. The low-frequency component is extracted from the long zero vector T7. The characteristic map of the isotropic (mean) admittance Y s (i d ,i q ) is calculated from the inductance characteristics, which can be obtained, for example, through an offline characterization of the electrical machine:

[0035] Y S (l d , lq)

[0036] An overview of the pointers is given in Figure 4 on the left for half a switching period. For the active pointer Ti, T2, the longer of the two active pointers Ti, T2 is always used with T A, called use, is chosen. For the null pointer, T7 is always used, with T Z use If the long active pointer Ti, T2 is shorter than a defined threshold, it is known from the prior art, as mentioned, that a reconstruction of the mean active pointer Ti, T2 by evaluating both zero pointers T o and T7. For this purpose, the points P and Q in Figure 4 on the right are reconstructed, and the mean current derivative is calculated using the lengths of the two active vectors Ti, T2. By defining a minimum vector length for the longer active vector Ti, T2, a decision is made as to whether a direct determination from the long active vector Ti, T2 or a determination by reconstruction is used.

[0037] Pointer selection 3 is the same in both cases, with only the input signal differing. Based on the incoming signal, the duty cycles are extracted and the current pointer position is determined. By incrementing and sorting, the pointers are then prepared so that the correct pointers (long active pointer Ti, T2, long zero pointer T o , T7) are selected. Furthermore, an evaluation is performed to determine whether the method for reconstructing the mean active phasor Ti, T2 is used. After selection and sorting, the LSQ algorithm 6 is triggered accordingly, and the voltage levels of the individual phasors are provided using an offline calculated table compensated for component-specific voltage drops. The triggers of the LSQ algorithm 6 start with a defined delayed start time, dependent on the power semiconductor, to minimize the influence of current overshoots.

[0038] Figure 5 is a schematic view of an angle control loop 2. Using the anisotropic high-frequency component from the long / medium active phasor Ti, T2, and the associated voltage, a transformation is performed as described in the prior art. The transformed signal, with a similar phase position, is then further processed using an angle control loop 2 extended by an additional PT1 element and compensated for the influence of cross-coupling.

[0039] The presented method is not limited to applications with a constant switching frequency and the modulation method Space Vector Modulation (SVPWM). While a variable switching frequency f sprimarily influences the angle control loop 2, the arrangement or presence of the pointers changes when discontinuous methods are used. This is taken into account in a pointer selection block 3, so that the method can be used for a variety of discontinuous methods (GDPWM, DPWMMIN, DPWMMAX, DPWMO, DPWM1, DPWM2, DPWM3).

[0040] In one embodiment, the method for operating a synchronous machine is carried out as shown in Figure 1, wherein the synchronous machine is designed to be multi-phase and is fed via an inverter or inverter 1.

[0041] In one embodiment, an offline optimized operating point-dependent (with respect to torque M, speed n, DC voltage VDC, circuit breaker temperature, driving mode FM) specification of a minimum high-frequency rectangular injection is carried out using several characteristic maps KF (see Figure 3) with application-specific optimization objectives. For example, one of the characteristic maps KF can have the optimization objective of efficiency, another the optimization objective of comfort, and another the optimization objective of dynamics.

[0042] In one embodiment, switch positions calculated in a pulse width modulation component 4 are transferred to a sensorless algorithm 5 for identifying a relevant phasor characteristic, i.e., phasor lengths, phasor start points, and phasor end points, as well as for calculating the resulting voltage levels within individual phasors. In one embodiment, the sensorless determination of the rotor position is carried out by combining the arbitrary injection method and a least-square algorithm 6, wherein the long active phasor Ti, T2 is used to determine the total current derivative, and the long mean zero phasor T o , T7and for the high-frequency isotropic part an offline calculated table is used depending on d- and q-current.

[0043] In one embodiment, depending on the power semiconductors of the inverter 1, a delayed start time of the least-square algorithm 6 is defined at the beginning of each pointer to minimize the influence of current overshoots.

[0044] In one embodiment, a connected angle control loop 2 is used to minimize high-frequency oscillations in the estimated rotor position signal.

[0045] In one embodiment, a pointer length-dependent switching occurs between the direct calculation from the active long pointer Ti, T2 and the reconstruction from the mean value of both active pointers Ti, T2.

[0046] In one embodiment, a variable sequence control of the pointer selection 3 and sorting is carried out to handle discontinuous modulation methods.

[0047] In one embodiment, a method for monitoring a permanent magnet synchronous machine as shown in Figure 2 is carried out, wherein the synchronous machine is designed in multiple phases and is fed via an inverter or inverter 1, and is operated at least partially via a classic anisotropy-based method of high-frequency injection.

[0048] In one embodiment, the switch positions on the gate drivers 7 of the inverter 1 are read back to identify the pointer times and resulting voltage levels within individual pointers to comply with safety-relevant requirements.

[0049] 1 inverter

[0050] 2 Angle control loop

[0051] 3 Pointer selection

[0052] 4 Pulse width modulation

[0053] 5 sensorless algorithm

[0054] 6 Least square algorithm, LSQ algorithm

[0055] 7 Gate driver ai Amplitude of injection fi Frequency of injection

[0056] FM driving mode f s Switching frequency i a Electricity

[0057] KF map

[0058] M torque n speed

[0059] P point

[0060] Q point

[0061] To null pointer

[0062] Ti active pointer

[0063] T2 active pointer

[0064] T? Null pointer

[0065] T A, use selected active pointer

[0066] Tz, use selected null pointer

[0067] VDC DC voltage v a Tension

Claims

Patent claims Method for determining the position of a rotor of an electrical machine without a rotor position sensor, wherein an anisotropy-based position control and / or regulation algorithm with high-frequency excitation of the voltage modulation is used to estimate a rotor angle, characterized in that a signal affected by measurement noise with a minimally disturbing slope tangent is calculated using a least-squares algorithm (6) that is capable of identifying high-frequency current derivatives, wherein an angle control loop (2) is also used to minimize occurring disturbances and jumps in the rotor angle, wherein an amplitude-optimized high-frequency injection is also used. Method according to claim 1, characterized in that the high-frequency excitations of the voltage modulation are generated by means of the arbitrary injection.Method according to claim 1 or 2, characterized in that the estimation of the rotor angle is used as a function of a requirement of the electric machine for feedback to the field-oriented control and / or for redundant monitoring of the rotor angle. Method according to claim 3, characterized in that an operating point-dependent, high-frequency, rectangular injection is used for feedback to the field-oriented control, wherein a. Operating point-dependent selection of frequency (fj) and amplitude (ai) of the injection depending on a torque (M), a speed (n), a DC voltage (VDC) and a switching frequency (f s) and a driving mode (FM) by means of several characteristic maps (KF). Method according to claim 4, characterized in that by including the driving mode (FM), a distinction is made between injection that is optimal in terms of efficiency, comfort, and performance. Method according to one of the preceding claims, characterized in that information about the rotor angle is obtained by forming the difference between an anisotropic high-frequency component of an active pointer (T 1; T2) is extracted, where the active pointer (T 1; T2) is a voltage state in which an inverter (1) for controlling the electrical machine is not in freewheeling mode, wherein an isotropic high-frequency component is generated via calculated voltages in the active vector (T 1;T2) and a mean admittance characteristic map as a function of d- and q-currents by means of multiplication, wherein the low-frequency component is extracted from a long zero vector (T7), wherein the admittance characteristic map is calculated from inductance characteristic maps. Method according to claim 6, characterized in that the inductance characteristic maps are obtained by an offline characterization of the electrical machine. Method according to one of claims 6 or 7, characterized in that the anisotropic high-frequency component from a long and / or mean active vector (T 1; T2) and an associated voltage, a transformation is carried out, whereby the transformed signal with a similar phase position is then further processed by means of an angle control loop (2) extended by an additional PT1 element and compensated for the influence of cross-coupling. Method according to one of the preceding claims, characterized in that the method is applied to a permanent magnet synchronous machine used as a drive machine of a vehicle.

Images