MOTOR CONTROL DEVICE AND DRIVE SYSTEM INCLUDING THE SAME

Abstract

To provide a motor control device capable of implementing control improved in responsiveness by estimating a rotor position at high speed, and a drive system provided with the same.SOLUTION: A motor control device controls an AC synchronous motor through sensorless control in which a rotor position sensor is not used. The motor control device includes: a first position estimator 151 which estimates a position of a rotor of a motor in a stationary coordinate system in accordance with a first estimation method; and a second position estimator 152 which estimates the position of the rotor of the motor in the stationary coordinate system in accordance with an estimation method different from the first estimation method, and drives the AC synchronous motor on the basis of estimation results thereof. In one embodiment, the first estimation method estimates the position of the rotor by capturing a change of inductance on the basis of a current ripple which is generated when a position detection voltage vector is applied to the AC synchronous motor. A second estimation method estimates the position of the rotor on the basis of an extended induced voltage estimate.SELECTED DRAWING: Figure 2A

Description

[Technical field]

[0001] The present invention relates to a motor control device that controls an AC synchronous motor by sensorless control, and a drive system including the same. [Background technology]

[0002] An AC synchronous motor is an electric motor that has a permanent magnet built into the rotor and is configured to operate when supplied with AC current, and includes brushless DC motors, stepping motors, etc. In short, electric motors other than those that are supplied with DC current and use a commutator to change the direction of the winding current are included in the category of AC motors, while electric motors with permanent magnets built into the rotor are included in the category of synchronous motors.

[0003] A typical motor controller for an AC synchronous motor controls an inverter that converts DC to AC and supplies AC current to an electric motor. To properly control the inverter, information on the rotor position is required. Therefore, the inverter is controlled using the output of a rotor position detector that detects the rotational position of the rotor. There is a known method for driving an AC motor by estimating the rotor position and controlling an inverter based on the estimated rotor position instead of using a rotor position detector. This type of control method is called "position sensorless control" or simply "sensorless control." By eliminating the rotor position detector, there is no need to consider the mounting position accuracy of the rotor position detector and the wiring related to the rotor position detector. In addition, sensorless control has the advantage that it can be applied to motors where it is physically impossible to install a rotor position detector or motors for which the rotor position detector cannot withstand the operating environment.

[0004] In a typical sensorless control, the rotor position is estimated by the induced voltage method. The induced voltage method uses a voltage command and a current detection value, calculates an induced voltage by a motor model, and estimates the rotor position using the induced voltage. More specifically, a γδ rotating coordinate system is assumed that has an axis error of Δθ with respect to the dq axes of a dq rotating coordinate system that rotates synchronously with the rotor. The induced voltage is estimated on the γδ axes of the γδ rotating coordinate system, and PLL (phase-locked loop) control is performed to output an estimated speed so that Δθ becomes zero (Patent Document 1). In addition, a method is also known in which an adaptive observer is used to estimate the rotor magnetic flux position on the dq axes, and speed estimation is performed so that the d-axis component of the rotor magnetic flux becomes zero (Patent Document 2). These methods are applicable in the medium to high speed range where a relatively large induced voltage occurs.

[0005] In the low-speed range including zero speed, the induced voltage is small, so it is difficult to estimate the induced voltage, and therefore another method is used. Specifically, a high-frequency voltage command is superimposed on the voltage command on the dq axis, and the current response to the high-frequency voltage command is detected to calculate the rotor position information contained in the motor inductance. This gives the axis error Δθ, and the speed is estimated so that the axis error Δθ becomes zero (Patent Document 3).

[0006] A combination of the two aforementioned methods has been proposed as a method for performing sensorless control over the entire speed range. Specifically, the latter method of superimposing a high-frequency voltage command is used in the low-speed range. At medium to high speeds, the former method of reducing the high-frequency superimposed voltage and obtaining an estimated position using an adaptive magnetic flux observer is used (Patent Document 4). A method has also been disclosed that weights the estimated positions obtained by both estimation methods, enabling smooth switching of control even under load (Patent Document 5). [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP2008-011616A, paragraphs 0007-0016 [Patent Document 2] Retable No. 02-091558, formula (18) [Patent Document 3] JP 2002-058294 A, paragraph 0076, number 8 [Patent Document 4] Restated Publication No. 2010-109528, Claims 5 and 6 [Patent Document 5] Republished Publication No. 2014-128887, paragraph 0063, Figure 6 [Patent Document 6] JP 2006-158046 A, paragraph 0005 [Patent Document 7] Japanese Patent Application Publication No. 8-256496 [Patent Document 8] JP-A-10-94298, paragraph 0001,0042 [Non-Patent Document 1] Z. Chen et al., “An Extended Electromotive Force Model for Sensorless Control of Interior Permanent-Magnet Synchronous Motors,” IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 50, NO. 2, APRIL 2003, p.288~295 [Non-Patent Document 2] T. Aihara and 4 others, "Sensorless Torque Control of Salient-Pole Synchronous Motor at Zero-Speed ​​Operation", IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 14, NO. 1, JANUARY 1999, p.202~208 Summary of the Invention [Problem to be solved by the invention]

[0008] It is difficult to realize position sensorless control that drives stably and with high response over the entire speed range from zero speed to high speed. This is due to the problem of control switching when transitioning from a control method for the low speed range to a control method for the medium speed range, and the poor response of position estimation in the low speed range. As pointed out in Patent Document 6, the method of superimposing a high-frequency voltage has a problem of vibration caused by high-frequency current. In addition, since the frequency of the high-frequency voltage is less than several hundred Hz at most, the period of the position calculation is long. Moreover, the response of the position calculation is further degraded by demodulating the high-frequency voltage command, which has a not-so-high frequency, into the response of the current on the dq axis. Therefore, as long as the high-frequency superimposing method is used, the responsiveness of the position calculation in the low-speed region is limited.

[0009] Furthermore, when estimating on the dq axes of the dq rotating coordinate system, the quantity that can be calculated is not the rotor position itself but the position error Δθ. In order to obtain the rotor position, it is necessary to output the speed using the above-mentioned PLL control and calculate the position θ on the αβ fixed coordinate system fixed to the stator by integrating the speed. However, there is a limit to the responsiveness of the PLL controller. In addition, since the calculation of Δθ uses the current value in the estimated rotating coordinate system (γδ rotating coordinate system), the estimation accuracy deteriorates when Δθ becomes large due to sudden acceleration / deceleration or sudden load. Therefore, a deviation occurs in the Δθ calculated in the low speed range and the medium / high speed range. Therefore, the estimated position becomes unstable or chattering occurs in the switching region between the estimation for the low speed range (low speed estimation) and the estimation for the medium / high speed range (medium / high speed estimation).

[0010] Examples of practical problems caused by these factors are as follows. For example, when a motor is operating in the high-speed range and is stopped by a sudden load, the motor switches from medium-high speed estimation to low-speed estimation. At this time, the motor may lose synchronism or run out of control in the opposite direction to the commanded direction. Similar problems occur with sudden acceleration and deceleration. In this manner, when controlling an AC synchronous motor using sensorless control, problems have arisen due to poor response.

[0011] Therefore, one embodiment of the present invention provides a motor control device that can estimate a rotor position at high speed, thereby achieving control with excellent responsiveness, and a drive system including the same. [Means for solving the problem]

[0012] One embodiment of the present invention provides a motor control device that controls an AC synchronous motor by sensorless control without using a rotor position sensor. The motor control device includes a first position estimator that estimates a position of a rotor of the AC synchronous motor on a fixed coordinate system according to a first estimation method, a second position estimator that estimates the position of the rotor of the AC synchronous motor on the fixed coordinate system according to a second estimation method different from the first estimation method, and a drive control means that drives the AC synchronous motor based on the estimation results of the first position estimator and the second position estimator.

[0013] According to this configuration, since both the first estimation method and the second estimation method estimate the rotor position on a fixed coordinate system, both the first position estimator and the second position estimator can perform position estimation at high speed. Therefore, the control of the AC synchronous motor based on the estimation results of the first position estimator and the second position estimator has excellent responsiveness. In one embodiment of the present invention, the first estimation method and the second estimation method both estimate the rotor position without estimating an error in the rotor position. With this configuration, the rotor position can be estimated quickly without estimating an error in the rotor position, making it possible to control the AC synchronous motor with excellent responsiveness.

[0014] In one embodiment of the present invention, the first estimation method and the second estimation method both estimate the rotor position without using a PLL (phase-locked loop) control that outputs an estimated rotor speed so that an error in the rotor position becomes zero. This configuration enables high-speed position estimation, and therefore allows the AC synchronous motor to be controlled with excellent responsiveness. In one embodiment of the present invention, the first position estimator outputs an estimated position signal having two periods of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, and the second position estimator outputs an estimated position signal having one period of fluctuation per rotor rotation corresponding to one electrical angle period of the stator.

[0015] In one embodiment of the present invention, the first position estimator outputs an estimated position signal having two periods of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, the second position estimator outputs an estimated position signal having one period of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, and the motor control device further includes a periodic converter that converts the estimated position signal of the first position estimator into an estimated position signal of a periodic signal having one period of fluctuation per rotor rotation corresponding to one electrical angle period of the stator.

[0016] This configuration obtains two types of estimated position signals, each consisting of a periodic signal that has one period of fluctuation per rotor rotation equivalent to one electrical angle period of the stator, so that these can be easily combined to obtain a valid estimated position signal. In one embodiment of the present invention, the first position estimator estimates the rotor position by detecting a change in inductance of the AC synchronous motor based on a current ripple generated in a winding current of the AC synchronous motor when a position detection voltage vector is applied to the AC synchronous motor. Such a first position estimator is suitable for estimating the rotor position in a low speed range including zero speed. In one embodiment of the present invention, the drive control means applies to the AC synchronous motor a voltage vector corresponding to a voltage command for driving the AC synchronous motor, and applies to the AC synchronous motor a position detection voltage vector for detecting a rotor position of the AC synchronous motor during a period in which the voltage vector corresponding to the voltage command is a zero voltage vector. The first position estimator estimates the rotor position by capturing a change in inductance of the AC synchronous motor based on a current ripple generated in a winding current of the AC synchronous motor when the position detection voltage vector is applied to the AC synchronous motor.

[0017] In one embodiment of the present invention, the second position estimator estimates the rotor position based on the extended induced voltage estimated value. Such a second position estimator is suitable for estimating the rotor position in a medium to high speed range where a significant induced voltage is generated. In one embodiment of the present invention, the motor drive system further includes an estimated position combiner that generates a combined estimated position by switching or weighting a first estimated position signal output by the first position estimator and a second estimated position signal output by the second position estimator in accordance with a rotational speed of the rotor or an extended induced voltage vector length, and the drive control means drives the AC synchronous motor in accordance with the combined estimated position generated by the estimated position combiner.

[0018] According to this configuration, the first estimated position signal and the second estimated position signal No. By switching or weighting and combining the first and second estimated position signals, an appropriate combined estimated position can be obtained. Since the switching or weighting and combining of the first and second estimated position signals is performed according to the rotor rotation speed or the extended induced voltage vector length, a combined estimated position that accurately represents the rotor position over a wide range of rotation speeds can be generated. This allows the AC synchronous motor to be appropriately controlled.

[0019] In addition, when the first position estimator outputs an estimated position signal having two cycles of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, and the second position estimator outputs an estimated position signal having one cycle of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, it is preferable to provide a period converter that converts the estimated position signal of the first position estimator into an estimated position signal having one cycle of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, and to use the estimated position signal generated by this period converter as the first estimated position signal. In this case, since the first and second estimated position signals are both periodic signals having one cycle of fluctuation per rotor rotation corresponding to one electrical angle period of the stator, it is possible to easily combine them to obtain a suitable combined estimated position. In this case, the combined estimated position is also a periodic signal having one cycle of fluctuation per rotor rotation corresponding to one electrical angle period of the stator.

[0020] In one embodiment of the present invention, the motor control device further includes a first compensator and a second compensator that respectively compensate the first estimated position signal and the second estimated position signal in accordance with a motor current and a rotor rotation speed. In another embodiment, the motor control device further includes a composite estimated position compensator that compensates the composite estimated position in accordance with a motor current and a rotor rotation speed. With these configurations, the rotor position can be estimated more accurately, thereby enabling more appropriate control of the AC synchronous motor.

[0021] In one embodiment of the present invention, the first position estimator estimates the rotor position by capturing a change in inductance of the AC synchronous motor based on a current ripple generated in a winding current of the AC synchronous motor when a position detection voltage vector is applied to the AC synchronous motor. The estimated position synthesizer generates the synthesized estimated position without using the first estimated position signal in a high-speed region where the rotor rotation speed is equal to or higher than a predetermined value. Furthermore, application of the position detection voltage vector is stopped in the high-speed region.

[0022] According to this configuration, since accurate position estimation is possible in the high speed range by the induced voltage, a composite estimated position is generated without using the first estimated position signal. This allows the AC synchronous motor to be appropriately controlled based on the accurate estimated position. In addition, in the high speed range, the application of the position detection voltage vector is stopped, thereby suppressing the influence on the drive of the AC synchronous motor and suppressing vibrations, etc.

[0023] According to an embodiment of the present invention, there is provided a drive system including an AC synchronous motor, an inverter that supplies AC current to the AC synchronous motor, and the motor control device that controls the inverter. With this configuration, the AC synchronous motor can be driven with excellent responsiveness. Effect of the Invention

[0024] According to the present invention, it is possible to provide a motor control device that can estimate the rotor position at high speed, thereby achieving control with excellent responsiveness, and a drive system including the same. [Brief description of the drawings]

[0025] [Figure 1A] FIG. 1A is a block diagram illustrating the configuration of a drive system including a motor control device according to an embodiment of the present invention. [Figure 1B] FIG. 1B is a block diagram for explaining the functional configuration of a controller included in the motor control device. [Figure 2A] FIG. 2A is a block diagram showing an example of the configuration of a weighted position estimator of the controller. [Figure 2B] FIG. 2B is a block diagram showing another example of the configuration of the weighted position estimator of the controller. [Diagram 3] FIG. 3 is an electric circuit diagram for explaining a configuration example of an inverter included in the motor control device. [Figure 4A-4B] 4A and 4B show voltage vectors corresponding to the eight states of the inverter. [Diagram 5] FIG. 5 is an electric circuit diagram showing a model of an AC motor, and shows a Δ-connected three-phase motor model. [Figure 6A] FIG. 6A shows example waveforms of the voltage, current, and current differential when AC motor M is rotating at a low speed (including a stopped state). [Figure 6B] FIG. 6B shows other example waveform diagrams of the voltage, current, and current differential when AC motor M is rotating at a low speed (including a stopped state). [Figure 7] FIG. 7 is an electric circuit diagram showing a model of an AC motor, and shows a Y-connected three-phase motor model. [Figure 8] FIG. 8 shows an example of the inductance of an ideal sinusoid on a fixed UVW coordinate system. [Figure 9A-9C]Figures 9A, 9B, and 9C show examples of calculated plots of inductances Lα, Lβ, and Mαβ on the αβ fixed coordinate system, inductances Ld, Lq, and Mdq on the dq rotating coordinate system, and the m, n, and s components of the inductance for an ideal sinusoidal wave inductance. [Figure 10A-10C] 10A, 10B, and 10C show examples of current differential values ​​calculated when three types of voltage vectors are applied from an ideal sinusoidal inductance. [Figures 11A-11C] Figures 11A, 11B, and 11C show current differential values ​​obtained by magnetic analysis when the three types of voltage vectors described above are input to a three-phase surface permanent magnet motor with the motor's q-axis current being zero, and the rotor position is rotated through one electrical angle cycle. [Fig. 11D-11E] Fig. 11D shows three-phase signals Us, Vs, and Ws for position estimation, and Fig. 11E shows two-phase signals αs and βs for position estimation on an αβ-fixed coordinate system, and an estimated position obtained based on them. [Figure 12A-12B] 12A and 12B respectively show the results of magnetic analysis of motor inductances Lu, Lv, Lw, Muv, Mvw, and Mwu and the interlinkage magnetic flux of each phase coil when the states shown in FIGS. 11A to 11E are obtained. [Figure 13A-13D] Figures 13A and 13B respectively show the motor inductances Lu, Lv, Lw, Muv, Mvw, and Mwu and the flux linkages of the coils of each phase when the q-axis current is positive, and Figures 13C and 13D respectively show the motor inductances Lu, Lv, Lw, Muv, Mvw, and Mwu and the flux linkages of the coils of each phase when the q-axis current is negative, as determined by magnetic analysis. [Figure 14A-14D] Fig. 14A shows the three-phase signals Us, Vs, and Ws for position estimation when the q-axis current is positive. Fig. 14B shows the three-phase signals Us, Vs, and Ws for position estimation when the q-axis current is negative. Fig. 14C shows the estimated position calculated for both positive and negative q-axis currents. Fig. 14D shows the error of the estimated position relative to the ideal estimated angle for both positive and negative q-axis currents. [Figure 15A-15B]Figure 15A shows the estimated position after translation correction, and Figure 15B shows the error of the estimated position after translation correction. [Figure 16A-16B] Figure 16A shows the estimated position after harmonic correction, and Figure 16B shows the error of the estimated position after harmonic correction. [Figures 17A-17C] Figures 17A and 17B show the results of converting the inductance on the UVW fixed coordinate system shown in Figure 12A into the inductance in the αβ fixed coordinate system and the dq rotating coordinate system, respectively. Figure 17C shows the corresponding components m, n, and s. [Figures 18A-18C] 18A, 18B and 18C respectively show the current derivatives obtained using three voltage vectors with the motor current at zero. [Fig. 18D-18E] Fig. 18D shows three-phase signals Us, Vs, and Ws for position estimation formed by the difference between in-phase current differential values, and Fig. 18E shows an estimated position calculated from the three-phase signals Us, Vs, and Ws for position estimation. [Figures 19A-19C] 19A, 19B, and 19C show the results of current differential values ​​obtained when a current of zero U phase, positive V phase, and negative W phase is applied to the motor lines with fixed phase excitation and the motor is forcibly rotated from the outside. [Figure 20A-20B] Fig. 20A shows three-phase signals Us, Vs, Ws for position estimation constructed using only two types of voltage vectors from the results of the current differential values ​​in Fig. 18A, Fig. 18B, and Fig. 18C. Fig. 20B shows the results of calculating an estimated position using the three-phase signals Us, Vs, Ws for position estimation. [Figure 21A-21B] Fig. 21A shows an example in which the three-phase signals Us, Vs, and Ws for position estimation are generated by in-phase differences based on the results of the current differential values ​​in Fig. 18A, Fig. 18B, and Fig. 18C. Fig. 21B shows the results of calculating an estimated position using the three-phase signals Us, Vs, and Ws for position estimation. [Fig. 22A-22B] Fig. 22A shows three-phase signals Us, Vs, and Ws for position estimation obtained by doubling the signals Vs and Ws in Fig. 21A and recalculating them. Fig. 22B shows the results of calculating an estimated position using them. [Figure 23]FIG. 23 shows a two-period signal generated by the first estimation method using pulse application and a one-period signal generated by the second estimation method using the extended induced voltage. [Figure 24] FIG. 24 is a flowchart illustrating the process performed by a period converter to convert a two-period signal into a one-period signal. [Diagram 25] FIG. 25 shows an example of the period conversion result obtained by the process of FIG. [Figure 26] FIG. 26 is a flow chart showing an example of weighted combining or switched combining of the first and second estimated position signals by the estimated position combiner. [Figure 27] FIG. 27 shows an example of the processing result of weighting combining or switching combining by the processing of FIG. [Figure 28] FIG. 28 shows the experimental results for the rapid acceleration operation. [Figure 29] FIG. 29 shows the experimental results for the sudden stop operation. [Diagram 30] FIG. 30 shows the results of a sudden stop experiment with compensation (correction) applied to the estimated position. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0026] In the following, an embodiment will be described that solves the problem that in the conventional technology, since both estimation calculations in the low speed range and the medium and high speed range output a position error Δθ, a separate PLL control is required to obtain the estimated position θ, and realizes highly responsive sensorless control in all speed ranges. This embodiment uses a first estimation method and a second estimation method that calculate the rotor estimated position in an αβ fixed coordinate system. Both the first and second estimation methods estimate the rotor position in a fixed coordinate system without estimating the position error Δθ in a rotating coordinate system. Furthermore, both the first and second estimation methods estimate the rotor position in a fixed coordinate system without using PLL control that outputs an estimated rotor speed so that the rotor position error Δθ in the rotating coordinate system becomes zero. Therefore, high-speed position estimation is possible, and motor control with excellent responsiveness can be realized.

[0027] Specifically, the first estimation method is a method for obtaining an estimated position in an αβ fixed coordinate system based on a current ripple caused by a voltage applied every PWM control period, and is an estimation method suitable for estimating the rotor position in a low speed range. The second estimation method is a method for obtaining an estimated position in an αβ fixed coordinate system by applying a minimum order observer to the calculation of an extended induced voltage, and is an estimation method suitable for estimating the rotor position in a medium to high speed range. For example, the methods described in Non-Patent Document 1 and Patent Document 7 can be used as the second estimation method.

[0028] In this embodiment, the estimation results of both the first and second estimation methods are switched or weighted based on the estimated speed or the extended induced voltage vector length and combined to obtain a final estimated position (composite estimated position). This makes it possible to realize highly responsive and smooth sensorless control over the entire speed range, including when the motor is stopped, without using PLL control. The problem of the difference between the estimated positions obtained by the first estimation method and the second estimation method is solved by applying compensation so that each estimated position approaches the true position.

[0029] For example, the estimated position (first estimated position) by the first estimation method using current ripple detection generates a large error due to the motor current. Therefore, the first estimated position is compensated depending on the motor current. In the first estimation method, depending on the S / N ratio (signal-to-noise ratio) of the current ripple detection, a digital filter may have to be used, and a delay in the estimated position calculation may occur depending on the speed. Therefore, the estimation results by the first and second estimation methods are compensated for by the current and speed so as to approach the true estimated value, and then weighted to obtain the final estimated position (composite estimated position). As a result, even if a sudden load is applied while the motor is rotating at high speed and the motor is stopped, the estimated position error due to the load and speed is reduced, and a stable transition of control (transition of estimation method) can be achieved.

[0030] Thus, in sensorless drive over the entire speed range, including from stopped to high speeds, a stable and highly responsive motor control device can be obtained even when a sudden load that causes the motor to stop instantly or when sudden acceleration or deceleration occurs. Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings. FIG. 1A is a block diagram for explaining the configuration of a drive system including a motor control device according to an embodiment of the present invention. The motor control device 100 is a device (AC motor control device) for driving an AC motor M (AC synchronous motor). More specifically, the motor control device 100 drives the AC motor M by so-called sensorless control, which controls the AC motor M without using a rotor position detector (rotor position sensor) that detects the position of the rotor of the AC motor M. The AC motor M is a synchronous motor with a permanent magnet built into the rotor, and more specifically, may be a surface permanent magnet synchronous motor (SPMSM). In this embodiment, the AC motor M is a three-phase permanent magnet synchronous motor, and has a U-phase winding 5u, a V-phase winding 5v, and a W-phase winding 5w. Hereinafter, these windings are collectively referred to as "windings 5uvw". FIG. 1A shows an example in which the windings 5uvw are Y-connected, but as described later, the windings 5uvw may be Δ-connected.

[0031] In this example, the motor control device 100 has a feedback system with a position control loop, a speed control loop, and a current control loop, and is configured to perform position servo control that controls the rotor position of the AC motor M in response to a position command. Vector control is used for the current control. The command from the outside is not limited to a position command, but may be a speed command or a torque command (current command). When a speed command is given, the position control loop is not used. When a torque command is given, only the current control loop is used, and the position control loop and the speed control loop are not used.

[0032] The rotor position is detected by current differential detection without using a rotor position detector. In a vesselThe rotor position is estimated using the signal obtained by this. More specifically, a position estimation signal that indicates the fluctuation in inductance of each phase winding of the AC motor M is created based on the current differential value, and the rotor position is estimated based on this position estimation signal. Surface permanent magnet synchronous motors, in principle, do not have salient poles, so it is said that magnetic pole detection using inductance changes is not possible, but when using magnets with strong magnetic force such as neodymium magnets, the inductance changes slightly due to magnetic saturation of the iron core.

[0033] To explain the specific configuration, the motor control device 100 includes a controller 1, current detectors 3u, 3v, 3w, and current differential detectors 4u, 4v, 4w, and is configured to control the inverter 2. The inverter 2 converts a DC current supplied from a DC power source 7 into an AC current and supplies it to windings 5uvw of the AC motor M. The motor control device 100, the inverter 2, and the AC motor M form a drive system.

[0034] The inverter 2 and the AC motor M are connected by three current lines 9u, 9v, 9w (hereinafter collectively referred to as "current lines 9uvw") corresponding to the U phase, V phase, and W phase. Current detectors 3u, 3v, 3w and current differential detectors 4u, 4v, 4w are disposed on each of these current lines 9uvw. The current detectors 3u, 3v, 3w (hereinafter collectively referred to as "current detectors 3uvw") detect the line currents flowing through the current lines 9uvw of the corresponding phases, i.e., the U-phase line current Iu, the V-phase line current Iv, and the W-phase line current Iw (hereinafter collectively referred to as "line current Iuvw"). The current differential detectors 4u, 4v, 4w (hereinafter collectively referred to as the "current differential detector 4uvw") are current differential detection means that detect the change over time in the line current flowing through the current line 9uvw of the corresponding phase, i.e., the current differential values ​​dIu, dIv, dIw of the U phase, V phase and W phase (hereinafter collectively referred to as the "current differential value dIuvw").

[0035] When the windings 5uvw of the AC motor M are Y-connected, the line current Iuvw is equal to the phase currents iu, iv, iw (hereinafter collectively referred to as "phase current iuvw") flowing through the windings 5uvw of each phase. When the windings 5uvw of the AC motor M are Δ-connected, the relationship between the line current Iuvw and the phase current iuvw is as shown in equation (3) described below. The line current and the phase current have values ​​corresponding to the winding currents flowing through the windings 5uvw of the AC motor M.

[0036] The controller 1 controls the inverter 2 based on the position command θcmd. The controller 1 has the form of a computer, and includes a processor (CPU) 1a and a memory 1b as a recording medium on which a program executed by the processor 1a is recorded. 1B is a block diagram for explaining the functional configuration of the controller 1. The controller 1 is configured so that the processor 1a executes a program to realize the functions of a plurality of function processing units. The plurality of function processing units include a position controller 11, a speed controller 12, a current controller 13, a PWM generator 14, a weighted position estimator 15, and a speed estimator 16. The current controller 13 includes a dq current controller 131, an inverse dq converter 132, a two-phase to three-phase converter 133, a three-phase to two-phase converter 134, and a dq converter 135.

[0037] The weighted position estimator 15 calculates the signal output by the current differential detector 4uvw, i.e., the current differential value dIuvw, and the detected current value I α ,I β (The drawing shows I α,β It is written as: Using these, a calculation is performed to estimate the position of the rotor of the AC motor M, and the estimated position θ (=θnew) is fed back to the position controller 11. The position controller 11 generates a speed command ωcmd for making the rotor position coincide with the position command θcmd based on the estimated position θ, and supplies the speed command ωcmd to the speed controller 12. In this manner, a position control loop is configured.

[0038] The estimated rotor position θ is also supplied to a speed estimator 16. The speed estimator 16 performs a calculation to estimate the rotor speed by obtaining a time change of the estimated position θ, and supplies the estimated speed ω (=ωnew) to the speed controller 12. The speed controller 12 calculates current commands Idcmd, Iqcmd for making the rotor speed coincide with the speed command ωcmd based on the estimated speed ω. (In the drawings, this is written as Id,q cmd.) is generated and supplied to the current controller 13. In this manner, a speed control loop is formed.

[0039] The line current Iuvw detected by the current detector 3uvw (more precisely, the detected value of the line current Iuvw) is supplied to the current controller 13. The current controller 13 generates a U-phase voltage command Vu, a V-phase voltage command Vv, and a W-phase voltage command Vw (hereinafter collectively referred to as "voltage commands Vuvw") for matching the line current Iuvw with the current commands Idcmd, Iqcmd, and supplies them to the PWM generator 14. In this manner, a current control loop is configured.

[0040] The PWM generator 14 is a pulse width modulation signal generating means that generates a PWM control signal (pulse width modulation signal) corresponding to the voltage command Vuvw and supplies the PWM control signal to the inverter 2. The PWM generator 14 applies a voltage corresponding to the voltage command Vuvw between the windings 5uvw of the AC motor M via the current line 9uvw. The speed controller 12 generates a d-axis current command Idcmd and a q-axis current command Iqcmd according to a dq rotating coordinate system and supplies them to the current controller 13. The dq rotating coordinate system is defined as a rotating coordinate system in which the magnetic flux direction of the rotor of the AC motor M is the d-axis and a direction perpendicular to the d-axis is the q-axis, and rotates according to the rotation angle (electrical angle) of the rotor. The three-phase to two-phase converter 134 converts the three-phase line current Iuvw detected by the current detector 3uvw into two-phase current values ​​I α ,I β The dq converter 135 converts the two-phase current value I α ,I β is converted to the d-axis current value Id and the q-axis current value Iq in the dq rotating coordinate system. (In the drawings, these are indicated as Id,q.)is supplied to the dq current controller 131. The dq current controller 131 generates a d-axis voltage command Vdcmd and a q-axis voltage command Vqcmd, which are voltage commands in the dq rotating coordinate system, so as to make the d-axis current value Id and the q-axis current value Iq coincide with the d-axis current command Idcmd and the q-axis current command Iqcmd, respectively. (In the drawings, this is written as Vd,q cmd.) In the inverse dq converter 132, the voltage commands Vαcmd and Vβcmd in the αβ coordinate system are (In the drawings, this is indicated as Vα,β cmd.) Furthermore, the voltage commands Vαcmd and Vβcmd in the αβ coordinate system are converted into Change The three-phase voltage commands Vuvw are then supplied to the PWM generator 14.

[0041] The weighted position estimator 15 supplies the estimated position θ (= θnew) to the inverse dq converter 132 and the dq converter 135. The estimated position θ is used for coordinate conversion calculation between the dq rotating coordinate system and the αβ coordinate system, and is also used for speed estimation calculation in the speed estimator 16. The speed estimator 16 generates the difference θnew - θold between the estimated position θ (= θnew) in the current calculation cycle generated by the weighted position estimator 15 and the estimated position θ (= θold) in the previous calculation cycle as the rotor rotational speed ω (= ωnew).

[0042] The current controller 13 is a drive control means that controls the PWM generator 14 to drive the AC motor M in accordance with the estimated position θ supplied from the weighted position estimator 15 . FIG. 2A is a block diagram for explaining a configuration example of the weighting position estimator 15. The weighting position estimator 15 includes a first position estimator 151 and a second position estimator 152. The weighting position estimator 15 further includes an estimated position synthesizer 153. The first position estimator 151 estimates the rotor position (angle) on the αβ coordinate system, which is a fixed coordinate system, according to a first estimation method. The second position estimator 152 estimates the rotor position (angle) on the αβ coordinate system according to a second estimation method. The first estimation method and the second estimation method are position estimation methods different from each other. In this embodiment, the first position estimator 151 calculates an estimated position of the rotor of the AC motor M based on a current differential value dIuvw detected by a current differential detector 4uvw. In this embodiment, the second position estimator 152 calculates an estimated position of the rotor of the AC motor M based on αβ voltage command values ​​Vαcmd, Vβcmd and αβ current detection value I α ,I β An estimated position θ of the rotor of the AC motor M is calculated based on the above.

[0043] In this embodiment, the first position estimator 151 outputs an estimated position signal having two periods of fluctuation for each rotation of the rotor corresponding to one electrical angle period of the stator of the AC motor M. In this embodiment, the second position estimator 152 outputs an estimated position signal having one period of fluctuation for each rotation of the rotor corresponding to one electrical angle period of the stator of the AC motor M. Therefore, the weighting position estimator 15 calculates the estimated position signal θ generated by the first position estimator 151. 1pre into a first estimated position signal θ1 having a fluctuation of one period per rotation of the rotor corresponding to one electrical angle period of the stator.

[0044] The estimated position combiner 153 generates a combined estimated position signal θ(=θnew) by combining the first estimated position signal θ1 and the second estimated position signal θ2, which is an estimated position signal generated by the second position estimator 152. More specifically, the estimated position combiner 153 switches between the first estimated position signal θ1 and the second estimated position signal θ2 or weights and combines them to generate the combined estimated position signal θ(=θnew).

[0045] The weighting position estimator 15 further includes a first compensator 161 and a second compensator 162. The first compensator 161 compensates (corrects) the first estimated position signal θ1 in accordance with the motor currents Id, Iq and the rotor rotation speed ω (=ωold (the rotor rotation speed calculated in the previous calculation cycle)) in the dq coordinate system. The second compensator 162 compensates (corrects) the second estimated position signal θ2 in accordance with the motor currents Id, Iq and the rotor rotation speed ω (=ωold) in the dq coordinate system. In this embodiment, the estimated position combiner 153 combines the first estimated position signal θ compensated by the first compensator 161. 1C and a second estimated position signal θ compensated by a second compensator 162. 2C to generate a composite estimated position signal θ(=θnew). This composite estimated position signal θ(=θnew) becomes the output of weighted position estimator 15, i.e., the estimated position θ.

[0046] 2B, instead of compensating the first estimated position signal θ1 and the second estimated position signal θ2 by the first compensator 161 and the second compensator 162, respectively, the first estimated position signal θ1 and the second estimated position signal θ2 may be combined by the estimated position combiner 153 without compensation, and the combined combined estimated position signal may be compensated by the combined estimated position compensator 163 (combined estimated position compensator) according to the motor currents Id, Iq and the rotor rotation speed ω (=ωold). The compensated combined estimated position signal θnew becomes the output of the weighting position estimator 15, i.e., the estimated position θ.

[0047] 3 is an electric circuit diagram for explaining a configuration example of the inverter 2. Three-phase bridge circuits 20u, 20v, and 20w are connected in parallel between a pair of power supply lines 8A and 8B connected to a DC power source 7. A smoothing capacitor 26 is further connected between the pair of power supply lines 8A and 8B. Each bridge circuit 20u, 20v, 20w (hereinafter collectively referred to as "bridge circuit 20uvw") is configured with a series circuit of upper arm switching elements 21u, 21v, 21w (hereinafter collectively referred to as "upper arm switching elements 21uvw") and lower arm switching elements 22u, 22v, 22w (hereinafter collectively referred to as "lower arm switching elements 22uvw") In each bridge circuit 20uvw, a current line 9uvw is connected to a midpoint 23u, 23v, 23w between the upper arm switching elements 21uvw and the lower arm switching elements 22uvw for connection to a corresponding winding 5uvw of the AC motor M.

[0048] The switching elements 21uvw, 22uvw are typically power MOS transistors, and include built-in parasitic diodes 24u, 24v, 24w; 25u, 25v, 25w connected in the reverse direction to the DC power supply 7. The current differential detector 4uvw is configured to detect a current differential value dIuvw which is a time differential value of the line current Iuvw flowing through the current line 9uvw of each phase.

[0049] The PWM control signal supplied from the controller 1 is input to the gates of the switching elements 21uvw, 22uvw, thereby turning the switching elements 21uvw, 22uvw on / off. The pair of the upper arm switching element 21uvw and the lower arm switching element 22uvw in each bridge circuit 20uvw is controlled so that when one is on, the other is off. The PWM control signal value that controls the upper arm switching element 21uvw to be on and the lower arm switching element 22uvw to be off is defined as "1", and the PWM control signal value that controls the upper arm switching element 21uvw to be off and the lower arm switching element 22uvw to be on is defined as "0". Then, the PWM control signal can take eight patterns (states) that can be expressed by a three-dimensional vector. These eight patterns (states) can be expressed as components as follows: (1,0,0), (1,1,0), (0,1,0), (0,1,1), (0,0,1), (1,0,1), (0,0,0), (1,1,1). Of these, the first six patterns (1,0,0), (1,1,0), (0,1,0), (0,1,1), (0,0,1), (1,0,1) correspond to a state in which a voltage is applied between the windings 5uv and 5uv and w of the AC motor M. The remaining two patterns (0,0,0) and (1,1,1) correspond to a state in which no voltage is applied between the windings 5uv and 5uv and w.

[0050] Figure 4A shows voltage vectors V0 to V7 corresponding to the above eight patterns (states). The voltage vectors V1 (1,0,0), V2 (1,1,0), V3 (0,1,0), V4 (0,1,1), V5 (0,0,1), and V6 (1,0,1) corresponding to the six patterns in which voltage is applied between the windings can be expressed by six voltage vectors that equally divide an interval of 360 electrical degrees into six, as shown in Figure 4B. The voltage vectors V0 (0,0,0) and V7 (1,1,1) are zero-voltage vectors in which no voltage is applied between the windings 5u, 5u, and 5w.

[0051] In the following, to simplify the description, the commas separating vector components may be omitted. In addition, in the following description, expressions such as "applying a voltage vector" mean that the inverter 2 is controlled to a state represented by the voltage vector, and a voltage corresponding to that is applied to the AC motor M. The first position estimator 151 estimates the rotor position in a low-speed range (including a stopped state) where the rotor rotation speed is low, according to a first estimation method. The second position estimator 152 estimates the rotor position in a medium- to high-speed range where the rotor rotation speed is relatively high, according to a second estimation method. In the following, first, the first estimation method used to estimate the rotor position in the low-speed range will be described, and then the second estimation method used to estimate the rotor position in the medium- to high-speed range will be described.

[0052] 5 is an electric circuit diagram showing a model of AC motor M, and shows a delta-connected three-phase motor model. The voltage equation of this model is given by the following equation (1).

[0053]

number

[0054] Here, we assume that position estimation is performed in the low-speed range, and that when the motor rotation speed is sufficiently low, the induced voltage term can be ignored, and that the time-varying component of inductance is sufficiently small compared to the time-varying component of current, so the time-differential term of inductance can be ignored. As shown below, the inductance matrix on the UVW coordinate system is set as Muvw, and its inverse matrix M -1 By calculating uvw and using it to describe the phase current differential value, the following equation (2) is obtained.

[0055]

number

[0056] As mentioned above, what can be detected in a delta-connected motor is the line current Iuvw. The relationship between the line current Iuvw and the phase current iuvw of each phase winding, and the relationship between their time derivatives, are given by the following equation (3).

[0057]

number

[0058] Using this, the above formula (2) can be modified to describe the differential value with respect to time t of the line current Iuvw when the voltage vectors V1 (100), V3 (010), and V5 (001) are applied, as shown in the following formula (4). However, the term (the second term in the above formula (2)) representing the voltage drop due to the electrical resistance R (phase resistance) of each phase winding 5uvw is ignored here because an almost equivalent value can be obtained by detecting the line current differential value during the period when the voltage vector (000) or (111) is applied, and it can be substantially cancelled by subtracting this value. More specifically, when the three-phase signals Us, Vs, and Ws for position estimation described next are configured based on the line current differential value, the term for the voltage drop of the winding resistance R can be omitted, so for simplicity of explanation, the line current differential value from which the voltage drop term has been omitted is shown here.

[0059]

number

[0060] The three-phase signals Us, Vs, and Ws for position estimation when using the current differential values ​​when applying three types of voltage vectors V1 (100), V3 (010), and V5 (001) are defined as in the following equation (5). In the following equation (5), gu, gv, and gw are the current differential detection gains of each line current. The following equation (5) defines the three-phase signals Us, Vs, and Ws for position estimation so that the difference in current differentials of the same phase is factored out as the gains gu, gv, and gw of each phase.

[0061]

number

[0062] Since the PWM voltages Vu, Vv, Vw (terminal voltages applied to each phase when the upper arm switching element is on) are substantially equal among the three phases, substituting equation (4) into equation (5) with V=Vu=Vv=Vw, the three-phase signals for position estimation become a cyclic symmetrical equation as shown in the following equation (6). Such three-phase signals for position estimation are defined so that even if the motor becomes magnetically saturated and the inductance fluctuates when high torque is generated, the effect appears equivalently in the three phases, suppressing position detection errors.

[0063]

number

[0064] The three voltage vectors for position detection are not limited to V1 (100), V3 (010), and V5 (001). For example, three voltage vectors V2 (011), V6 (101), and V4 (110) can also be used to derive a three-phase signal for position estimation in a similar manner. It is also possible to create three-phase signals for position estimation using two types of voltage vectors. Specifically, when using current differential values ​​when applying two types of voltage vectors V1 (001) and V3 (010), the three-phase signals for position estimation can be defined, for example, as in the following equation (7). When the current differential detection gains of each phase are different, the three-phase signals for position estimation are as shown in equation (8). When the current differential detection gains are equal (g=gu=gv=gw), the equation becomes equation (9), which is equivalent to the equation (6) when all detection gains are the same.

[0065]

number

[0066] When using the current differential value when two types of voltage vectors are applied, it is necessary to generate the three-phase signals for position estimation by subtracting signals of different phases for two phases, so the gain of the current differential detector 4uvw cannot be bundled. Therefore, it is difficult to apply this to the case of using a current differential detector 4uvw (current transformer, etc.) configured such that the magnetic material is saturated at high current and the gain decreases. However, it is effective when the gain of the current differential detector 4uvw is equal for all phases and does not vary with the current value, and there is an advantage in that the responsiveness of the position detection is improved by reducing the number of types of voltage vectors for position detection. As is self-evident from the above formula (4), there is the relationship of the following formula (10), so by swapping the terms, the three-phase signals for position estimation can be similarly defined with the current differential value when two other types of voltage vectors are applied.

[0067]

number

[0068] In any case, when two types of voltage vectors are used for position detection, one of the phases of the three-phase signals for position estimation needs to be generated by the difference between different phases, and therefore is affected by the gain of current differential detector 4uvw. The current differential may be detected for only two phases, and the current differential for the remaining one phase may be calculated using the relationship in which the sum of the currents for all phases is zero.

[0069] The three-phase signals for position estimation thus obtained are converted from three-phase to two-phase, and an arctangent is taken, whereby an estimated rotor position can be obtained as shown in the following equation (11).

[0070]

number

[0071] The normalized self-inductances Lu, Lv, and Lw of each phase are expressed as the following equation (12) using the motor electrical angle θ and normalized inductance amplitude α. Here, it is assumed that the mutual inductance is small, targeting surface permanent magnet motors and the like. The normalized self-inductances Lu, Lv, and Lw are normalized by the inductance offset L0. The offset L0 is expressed by L0 = (Ld + Lq) / 2 using the inductances Ld and Lq in the dq rotating coordinate system, and the inductance amplitude L1 of each phase is expressed by L1 = (Ld - Lq) / 2. The normalized inductance amplitude α can be expressed as α = L1 / L0, which is the ratio of the inductance amplitude L1 to the offset L0.

[0072]

number

[0073] When three types of voltage vectors are used, the three-phase signal for position estimation is calculated using the above formula (6) to obtain the following formula (13). If α<<1 and terms equal to or greater than the square of α are ignored, it can be approximated to formula (14), and a three-phase sinusoidal signal can be obtained. In this way, an estimated position with two cycles of fluctuation for one cycle of electrical angle is obtained. The same is true when two types of voltage vectors are used.

[0074]

number

[0075] 6A and 6B show example waveform diagrams of the voltage, current, and current differential when AC motor M is rotating at low speed (including a stopped state). (a) of FIG. 6A and FIG. 6B shows the waveform of the U-phase wire voltage applied to U-phase current line 9u. (b) of FIG. 6A and FIG. 6B shows the waveform of the V-phase wire voltage applied to V-phase current line 9v. (c) of FIG. 6A and FIG. 6B shows the waveform of the W-phase wire voltage applied to W-phase current line 9w. Furthermore, (d), (e), and (f) of FIG. 6A and FIG. 6B show changes in U-phase wire current Iu, V-phase wire current Iv, and W-phase wire current Iw output by current detector 3uvw, respectively. (g), (h), and (i) in Figures 6A and 6B respectively show the changes in the time derivatives of the U-phase, V-phase, and W-phase line currents, i.e., the U-phase current derivative dIu, the V-phase current derivative dIv, and the W-phase current derivative dIw, and correspond to the output of the current derivative detector 4uvw.

[0076] As shown in FIG. 3, the inverter 2 is a three-phase inverter composed of six switching elements 21uvw, 22uvw, and three terminals connected to the U-phase, V-phase, and W-phase windings 5uvw of the AC motor M are connected to either the power supply voltage Vdc (PWM voltage) or ground potential (0V). As described above, the state connected to the power supply voltage Vdc (the upper arm switching element 21uvw is on) is expressed as "1", and the state connected to 0V (the upper arm switching element 21uvw is off) is expressed as "0". Then, as shown in FIG. 4A, there are eight types of voltage vectors generated, V0 (0,0,0) to V7 (1,1,1). Of these, V0 (0,0,0) and V7 (1,1,1) are zero voltage vectors in which all winding terminals are at the same potential and the voltage applied between the windings 5uvw is zero. The remaining six voltage vectors V1 to V6 are non-zero voltage vectors in which a voltage is applied between the windings 5uvw.

[0077] The PWM generator 14 generates a PWM control signal for turning on / off the switching elements 21uvw, 22uvw of the inverter 2 by comparing each phase voltage command Vuvw output from the current controller 13 with the triangular wave carrier signal. For example, the PWM frequency (the frequency of the triangular wave carrier signal) is 14 kHz, which corresponds to a period of about 70 μs. During low-speed rotation, the phase voltage command Vuvw is low, so the period of the zero voltage vectors V0, V7 in which no voltage is applied between the windings 5uvw becomes long. Figures 6A and 6B show waveforms in a state in which the AC motor M is stopped, with the period T0 of the zero voltage vector V0 and the period T7 of the zero voltage vector V7 each being approximately half the PWM period.

[0078] The PWM generator 14 has a function of applying voltage vectors V1, V3, V5 (position detection voltage vectors) for rotor position detection during the period of the zero voltage vector V0 or V7, in addition to generating the PWM control signal. The time for applying the position detection voltage vectors is sufficiently shorter than the PWM period (e.g., about 70 μsec), and is also sufficiently shorter than half the PWM period. More specifically, the time for applying the position detection voltage vectors is preferably 10% or less of the PWM period, more preferably 5% or less.

[0079] In order to minimize the influence of application of the position detection voltage vectors V1, V3, and V5, it is preferable to apply inverted voltage vectors V4 (011), V6 (101), and V2 (110), which are inverted versions of the position detection voltage vectors, for the same period as the position detection voltage vectors immediately after application of each position detection voltage vector, to cancel out the currents due to the position detection voltage vectors. Fig. 6A shows an example in which such inverted voltage vectors are applied, and Fig. 6B shows an example in which the inverted voltage vectors are not applied.

[0080] For each PWM period, position detection voltage vectors V1, V3, and V5 and the inverted voltage vectors V4, V6, and V2 that cancel them are applied to the U phase, V phase, and W phase, respectively. This ensures that the effect of applying the voltage vectors for position detection is equalized among the three phases. As shown in (d), (e), and (f) of FIG. 6A and FIG. 6B and (g), (h), and (i) of FIG. 6A and FIG. 6B, the U-phase, V-phase, and W-phase currents change in response to the application of the position detection voltage vector (and in response to the application of the inverted voltage vector in the case of FIG. 6A), and the U-phase, V-phase, and W-phase current differential detection voltages change. When the position detection voltage vector is applied, the current differential detection voltage of each phase changes instantaneously by directly detecting the current differential value using a current differential detector such as a current transformer. Therefore, the current differential value can be detected substantially within the application time of the position detection voltage vector (for example, 3 μs). The timing corresponding to the application of the position detection voltage vector is the current differential value acquisition timing (indicated by the symbol "★") at which the current differential value should be sampled. As for the current value of each phase, the output of the current detector 3uvw is sampled at the current value acquisition timing (indicated by the symbol "●") during the period in which the voltage vector for driving the motor is applied.

[0081] The three-phase signals Us, Vs, and Ws for position estimation can be obtained by substituting the current differential value detected in this manner into equation (5). Furthermore, the motor electrical angle θ can be obtained by performing the calculation of equation (11). Such calculation is performed by the first position estimator 151 (see FIGS. 2A and 2B). When two types of voltage vectors are used, the three-phase signals Us, Vs, and Ws for position estimation can be obtained by performing the calculation of equation (7) instead of equation (5).

[0082] To cancel the term due to the voltage drop in the winding resistance (the second term in equation (2)), also obtain the current differential value when the voltage vector is V7 (111) or V0 (000), and subtract it from the current differential value obtained when the position detection voltage vectors V1 (100), V3 (010), and V5 (001) are applied. The expansion of the above formula was considered for a delta connection model as shown in Figure 5, but the same applies to the Y connection as shown below. Considering the model as shown in Figure 7, the voltage equation is set as formula (15) using the midpoint potential Vn.

[0083]

number

[0084] Similarly to equation (2), the current derivative can be expressed using the inverse matrix of the inductance matrix as equation (16).

[0085]

number

[0086] Similarly to the derivation of equation (4), the differential value of the line current (equal to the phase current in a Y connection) when the position detection voltage vectors V1 (100), V3 (010), and V5 (001) are applied is expressed as equation (17).

[0087]

number

[0088] Here, when equation (16) is solved for the midpoint potential Vn using the following equation (18), equation (19) is obtained.

[0089]

number

[0090] When the three-phase signals for position estimation are defined as in-phase differences as in equation (5), equation (20) is obtained, which is a cyclic symmetrical equation in which the gain of the current derivative is bracketed, as in equation (6). Therefore, like the Δ connection, even if the motor becomes magnetically saturated and the inductance fluctuates when high torque is generated, the effect of this appears equivalently in the three phases, so position detection errors are suppressed. The same is true when V4 (011), V6 (101), and V2 (110) are used as the position detection voltage vectors. When two types of position detection voltage vectors are used, the effect of the gain of the current differential detector 4uvw is also felt, as in the case of the Δ connection.

[0091]

number

[0092] The conversion of the inductance matrix when moving from the UVW fixed coordinate system to the αβ fixed coordinate and dq rotating coordinate system is described below. Transformation matrix T from UVW fixed coordinate system to αβ fixed coordinate system αβ and the general inverse matrix T + αβ is defined as follows:

[0093]

number

[0094] Furthermore, the transformation matrix Tdq from the αβ fixed coordinate system to the dq rotating coordinate system and the inverse matrix T -1 dq is defined as follows:

[0095]

number

[0096] The product of each transformation matrix and its inverse matrix is ​​given by the following equation (23).

[0097]

number

[0098] The voltage equation for the phase in the UVW fixed coordinate system is expressed as the following equation (24) using the motor induced voltage e. From the left, the αβ transformation matrix T αβ By multiplying this by 1 and inserting a unit matrix between the inductance matrix and the current, we can define the voltage equation in the αβ fixed coordinate system as shown in equation (25).

[0099]

number

[0100] Here, since iu+iv+iw=0, the following equation (26) holds. Therefore, only the first term of the first equation of (23) remains. Therefore, the product of the αβ transformation matrices T + αβ T αβ is a unit matrix.

[0101]

number

[0102] Similarly, the dq transformation matrix T dq (equation (22)) and inserting a unit matrix (see the second equation of equation (23)), a voltage equation in the dq rotating coordinate system is obtained as shown in the following equation (27).

[0103]

number

[0104] From the derivation of equations (25) and (27), the inductance matrix M αβ ,M dq can be defined as follows:

[0105]

number

[0106] Here, we define the quantities m, n, and s in the following equation (29), which are composed of inductance components in the UVW coordinate system.

[0107]

number

[0108] Using equation (28), the inductance matrix in each coordinate system is calculated, and the inductance in the αβ fixed coordinate system or the dq rotating coordinate system is expressed using m, n, and s from equation (29), resulting in the following equations (30) and (31).

[0109]

number

[0110] Moreover, the inductance transformation from the αβ fixed coordinate system to the dq rotating coordinate system is expressed by the following equation (32).

[0111]

number

[0112] Figure 8 shows an example of the inductance of an ideal sinusoidal wave on the UVW fixed coordinate system. In this example, the amplitudes of the self-inductance and mutual inductance are 0.1 and 0.02, respectively, and the offsets are 1.3 and -0.11, respectively, and sinusoidal waves with a phase shift of 120° between phases are assumed. Regarding such an ideal sinusoidal inductance, the inductance L in the αβ fixed coordinate system is calculated using equations (29), (30), and (31). α ,L β ,M αβ When the inductances Ld, Lq, Mdq, and the m, n, and s components on the dq rotating coordinate system are calculated and plotted, they are as shown in Figures 9A, 9B, and 9C. As is generally known, the inductances Ld and Lq on the dq rotating coordinate system are not dependent on the rotor position, and in this example, Lq = 1.34 and Ld = 1.48. The salient pole ratio Lq / Ld = 1.10.

[0113] 10A, 10B, and 10C show the results of calculating the current differential value using equation (4) from the ideal sine wave inductance shown in FIG. 8. FIG. 10A shows the current differential value when a voltage vector V1 (100) is applied, FIG. 10B shows the current differential value when a voltage vector V3 (010) is applied, and FIG. 10C shows the current differential value when a voltage vector V5 (001) is applied. All of them show the changes in the current differential value of the U-phase, V-phase, and W-phase with respect to the rotor electrical angle. The detection gain and voltage were set to 1.

[0114] As an example, in a magnetic analysis, the above three types of voltage vectors are input to a three-phase surface permanent magnet motor with the q-axis current of the motor being zero, and the rotor position is rotated for one electrical angle period. The current differential values ​​and the position estimation results calculated using equations (5) and (11) are shown in Figs. 11A to 11E. Fig. 11A shows the current differential values ​​when a voltage vector V1 (100) is applied, Fig. 11B shows the current differential values ​​when a voltage vector V3 (010) is applied, and Fig. 11C shows the current differential values ​​when a voltage vector V5 (001) is applied. Fig. 11D shows the three-phase signals Us, Vs, and Ws for position estimation calculated using equation (5). Furthermore, Fig. 11E shows the two-phase signals αs and βs for position estimation on an αβ fixed coordinate system calculated using equation (11), and the estimated position θ obtained based on them. Although the three-phase signals Us, Vs, and Ws for position estimation have waveforms with superimposed harmonics, they can be roughly regarded as sine waves, and it is clear that the estimated position θ can be calculated.

[0115] 12A and 12B respectively show the results of magnetic analysis of the motor inductances Lu, Lv, Lw, Muv, Mvw, and Mwu and the interlinkage magnetic flux of each phase coil (winding) when Figures 11A to 11E are obtained. Comparing with Figure 8, it can be seen that the reason why the three-phase signals Us, Vs, and Ws for position estimation in Figure 11D are not ideal sine waves is that the inductance deviates from an ideal sine wave and that there exists a higher-order term in the ratio α (normalized inductance amplitude) between the inductance offset amount and its amplitude as shown in Equation (13).

[0116] Next, the results of a similar analysis performed when the q-axis current of the motor is positive or negative are shown in Figures 13A, 13B, 13C, and 13D. Figures 13A and 13B respectively show the results of the motor inductance Lu, Lv, Lw, Muv, Mvw, and Mwu and the interlinkage magnetic flux of each phase coil (winding) obtained by magnetic analysis when the q-axis current is positive. Figures 13C and 13D respectively show the results of the motor inductance Lu, Lv, Lw, Muv, Mvw, and Mwu and the interlinkage magnetic flux of each phase coil (winding) obtained by magnetic analysis when the q-axis current is negative. In these figures, the positive and negative rotor electrical angles are defined as the direction in which the rotor electrical angle advances to the positive (advance angle direction) when the q-axis current is positive with no load. In other words, the direction of the torque generated when the q-axis current is positive is defined as the positive direction of the rotor electrical angle.

[0117] It can be seen that the coil's flux linkage shifts in the advance angle direction (torque generation direction) when the q-axis current is positive, and in the delay angle direction (torque generation direction) when the q-axis current is negative, compared to when the current is zero. As for inductance, the amplitude changes depending on the difference in magnetic resistance in the positive or negative direction of the d-axis, and harmonics are included due to the slot combination. However, essentially, the phases of the self-inductances Lu, Lv, Lw and the mutual inductances Muv, Mvw, Mwu can all be considered to shift in the same direction as the phase shift of the coil's flux linkage.

[0118] 14A, 14B, and 14C show analysis results of the three-phase signals for position estimation and the estimated position when the q-axis current is applied. FIG. 14A shows the three-phase signals for position estimation Us, Vs, and Ws calculated by equation (5) when the q-axis current is positive. FIG. 14B shows the three-phase signals for position estimation Us, Vs, and Ws calculated by equation (5) when the q-axis current is negative. FIG. 14C shows the estimated position θ calculated by equation (11) when the q-axis current is positive and negative. FIG. 14C also shows the rotor electrical angle used in the analysis (analytical true value) as the ideal estimated angle. Furthermore, FIG. 14D shows the error of the estimated position with respect to the ideal estimated angle when the q-axis current is positive and negative.

[0119] It can be seen that, in accordance with a phase shift of the inductance, the estimated position θ shifts in the positive direction (torque generation direction) compared to the ideal estimated angle when the q-axis current is positive, and in the negative direction (torque generation direction) when the q-axis current is negative. If the deviation of the estimated position θ from the ideal estimated angle becomes large, the torque decreases, and in the worst case, the motor may lose synchronization. Therefore, a correction amount expressed as a function of the q-axis current is introduced. As an example, a translational correction (first correction) is performed by multiplying the q-axis current value Iq by a proportional constant k (k>0) to obtain a correction amount C1 (C1=k·Iq), and subtracting this correction amount C1 from the estimated position θ before correction. This translational correction is an operation to shift the estimated position θ by the absolute value of the correction amount C1 (first correction amount) in the torque generation direction. The estimated position θ after translational correction for both positive and negative q-axis currents is C1 (=θ-C1) is shown in FIG. 15A. Furthermore, the estimated position θ C1 The estimated angle error with respect to the ideal estimated angle is shown in FIG. 15B.

[0120] The estimated angle error, which was about ±50° before correction (see Figure 14D), is suppressed to within ±20°. This shows that the estimation error caused by the phase shift of the inductance accompanying an increase in the q-axis current (increase in absolute value) can be resolved by this correction. In this example, a proportional equation is used as a function of the q-axis current. However, if translational correction is performed by introducing a correction amount using a function including higher-order terms related to the q-axis current, a value closer to the ideal estimated value can be obtained in response to changes in the q-axis current.

[0121] Such an estimated position θ after translation correction C1 After calculating, a harmonic correction (second correction) is further performed to reduce the estimated angle error. For example, a correction amount C2 having a harmonic that is n times (n is a natural number equal to or greater than 2, for example, n=3) the estimated position θ is introduced as the harmonic correction amount. More specifically, a harmonic correction amount C2 (second correction amount) of the following equation (33) is introduced as an n-th harmonic with the q-axis current value Iq as an amplitude. The harmonic correction amount C2 is a function of the estimated position θ and the q-axis current value Iq, and more specifically, the estimated position θ (estimated position after translation correction θ C1It can also be said that the function is a product of a harmonic component using the phase of q-axis current and a function of the q-axis current value. In the following equation (33), the function of the q-axis current value is the q-axis current value itself, but it may be, for example, a function of the q-axis current value multiplied by a proportionality constant, or a function including higher-order terms.

[0122] C2 = Sin(nθ C1 + δ) × Iq ​​(33) This harmonic correction amount C2 is calculated by the estimated position θ C1 This further subtracts the estimated position θ after harmonic correction. C2 is expressed by the following equation (34). θ C2 =θ C1 -Iq Sin(nθ C1 +δ)=θ-C1-Iq·Sin(n(θ-C1)+δ) ···(34) When n=3, the estimated position θ after correction C2 16A and 16B, respectively, where the phase offset δ is selected so as to reduce the estimation error.

[0123] As shown in Fig. 16B, by performing harmonic correction in addition to translational correction, the estimated position error is suppressed to less than ±8°. This makes it possible to reduce torque ripple due to estimated position error. In the harmonic correction example of Fig. 16A and Fig. 16B, only the third harmonics are reduced, but higher-order harmonic correction may be performed, and the estimated position error can be further reduced by performing correction using a correction amount that includes higher-order terms for the q-axis current, as in the case of translational correction.

[0124] Also, when the inductance phase shift due to the q-axis current is small, the translation correction may be omitted and only the harmonic correction may be performed. In this case, C1=0, so the estimated position θ after correction C2 is expressed by the following equation (35). θ C2 =θ-Iq·Sin(nθ+δ) ···(35) It is also possible to omit the harmonic correction and perform only the translation correction.

[0125] In the above example, the correction amounts C1 and C2 are determined using a function for the q-axis current value Iq and the estimated position θ before correction, but instead of using a function, the correction amounts may be tabulated in advance. Furthermore, instead of generating the correction amounts using a function or a table, the corresponding estimated position after correction itself may be tabulated. Such translational correction and / or harmonic correction is performed by the first compensator 161 (see FIG. 2A), and the corrected estimated position θ 1C That is, the estimated position θ 1C =θ C2 When only translation correction is performed, the estimated position θ 1C =θ C1 It is.

[0126] In the above example, a surface permanent magnet motor was described. However, even when an interior permanent magnet motor is used, the inductance waveform shifts due to a shift in the coil's flux linkage, and harmonics are superimposed on the estimated value, in a similar manner, although to a lesser extent. Figures 17A and 17B show the results of converting the inductance on the UVW fixed coordinate system shown in Figure 12A into the inductance in the αβ fixed coordinate system and the dq rotating coordinate system using equations (29), (30), and (31), respectively. Figure 17C shows the corresponding components m, n, and s.

[0127] Because the inductance change in Figure 12A is not a perfect sine wave, both inductances Ld and Lq in the dq rotating coordinate system exhibit rotor position dependency. In addition, it can be seen that the mutual inductance Mdq, which is an interference component of the dq axes, is not zero. From the results in Figure 17A, the average inductances with respect to the rotor electrical angle are Ld = 1.4, Lq = 1.5, and the average salient pole ratio Lq / Ld = 1.07.

[0128] This shows that position estimation is possible with sufficient accuracy even for surface permanent magnet motors with a small salient pole ratio, such as a salient pole ratio of approximately 7% on average when not excited, and even as low as 1% depending on the rotor electrical angle. A three-phase surface permanent magnet motor was prepared as an actual motor with the same conditions as in the above analysis, and the above PWM pattern was applied to this motor. A current transformer whose gain varies due to saturation of the magnetic material caused by the magnitude of the current was used as the current differential detector 4uvw to obtain the current differential value and estimate the position. The results are shown below.

[0129] Figures 18A, 18B, and 18C show current differential values ​​obtained using three voltage vectors V1 (100), V3 (010), and V5 (001) when the motor current is zero. Figure 18D shows three-phase signals Us, Vs, and Ws for position estimation constructed according to equation (5) from the difference between in-phase current differential values. And Figure 18E shows an estimated position calculated from the three-phase signals Us, Vs, and Ws for position estimation according to equation (11). It can be seen that when the current is zero, an estimated position can be calculated in the same way as the analysis result.

[0130] Figures 19A, 19B, and 19C show the results of current differential values ​​obtained when a current with zero U-phase, positive V-phase, and negative W-phase is applied to the motor wires with fixed phase excitation, and the motor is forcibly rotated from the outside. Figures 19A, 19B, and 19C show the results of current differential values ​​obtained by applying voltage vectors V1 (100), V3 (010), and V5 (001), respectively. The relationship between the rotor electrical angle on the horizontal axis and the excitation angle phase is d-axis excitation at a rotor electrical angle of 0°, q-axis excitation at 90°, and inverse d-axis excitation at 180°.

[0131] For example, when comparing the U-phase signal of the pattern of voltage vector V1 (100) (see FIG. 19A) with the V-phase signal of the pattern of voltage vector V3 (010) (see FIG. 19B) and the W-phase signal of the pattern of voltage vector V5 (001) (see FIG. 19C), which should essentially provide signal values ​​of the same level, it can be seen that the V-phase and W-phase signals are attenuated to about half their original value due to the effects of saturation of the magnetic material that constitutes the current transformer of current differential detector 4uvw.

[0132] Fig. 20A shows the result of constructing three-phase signals Us, Vs, and Ws for position estimation using equation (7) that uses only two types of voltage vectors V5 (001) and V3 (010) from the results of the current differential values ​​in Fig. 18A, Fig. 18B, and Fig. 18C. Fig. 20B shows the result of calculating an estimated position θ using the three-phase signals Us, Vs, and Ws for position estimation according to equation (11). As shown in equation (7), the signals Vs and Ws are composed of the differences of signals with different phases, and therefore have different gains. Because signals with different gains are subtracted, the three-phase signals cannot be calculated properly, and the estimated position cannot be calculated correctly.

[0133] The three-phase signals for position estimation appear to have a simple offset. Also, in this example, the absolute values ​​of the V-phase and W-phase currents are set to be equal, so the gains of the V-phase and W-phase are equal, and the behavior is similar to when an offset occurs. However, in reality, the currents of the UVW phases change over time, and the gains of each phase behave without any special constraints. Therefore, in reality, the three-phase signals for position estimation change in a complex manner according to the motor current, due to the terms that appear due to the sum and difference of the gains in equation (8). This makes it difficult to perform correction.

[0134] Therefore, when using two types of voltage vectors to detect the rotor position, it is preferable to use a current differential detector that uses an element with a structure that can avoid saturation of the magnetic material, such as a current transformer that uses an air-core coil. FIG. 21A shows an example in which the three-phase signals Us, Vs, and Ws for position estimation are formed by in-phase differences using equation (5) based on the results of the current differential values ​​in FIGS. 18A, 18B, and 18C, and FIG. 21B shows the result of calculating an estimated position using equation (11) using the three-phase signals Us, Vs, and Ws for position estimation. It can be seen that by subtracting the in-phase signals, the estimated position can be calculated while suppressing the influence of the gain of the current differential detector 4uvw. The distortion of the estimated position is caused by the gains gu, gv, and gw that bracket the entire equation and appear in equation (6). Specifically, the gains gv and gw are approximately half of the gain gu, and therefore the amplitudes of the three-phase signals Vs and Ws for position estimation are approximately half of the three-phase signal Us for position estimation, which is the cause of the distortion of the estimated position.

[0135] This can be easily corrected by simply multiplying each of the three-phase signals for position estimation by a gain according to the current. Figure 22A shows the three-phase signals Us, Vs, and Ws for position estimation obtained by doubling the signals Vs and Ws in Figure 21A and recalculating them, and Figure 22B shows the result of calculating the estimated position using these signals according to equation (11). It can be seen that the three-phase signals for position estimation have a symmetrical shape for all three phases, and there is no distortion in the estimated position.

[0136] The correction of multiplying the three-phase signals for position estimation by a gain according to the current may be replaced with a calculation of varying the current differential detection gains gu, gv, and gw (hereinafter collectively referred to as "guvw") according to the current. For example, the gain guvw of each phase may be determined according to the function of the following equation (36) based on the absolute value |Iuvw| of the line current Iuvw of each phase of the motor. According to the function of equation (36), when the absolute value of the line current Iuvw of each phase is equal to or smaller than the first constant I1 (I1>0), the gain guvw of the corresponding phase becomes a constant first gain g1 (g1>0), and when the absolute value of the line current Iuvw of each phase is greater than the second constant I2 (I2>I1), the gain guvw of the corresponding phase becomes a constant second gain g2 (g2>g1). When the absolute value of the line current Iuvw of each phase is greater than the first constant I1 and is equal to or smaller than the second constant I2, the gain guvw of that phase varies linearly between the first gain g1 and the second gain g2 in accordance with the absolute value of the line current Iuvw of that phase.

[0137]

number

[0138] The gain of the current differential detection for the motor current may be measured in advance, and the constants I1, I2, g1, and g2 may be determined by fitting with the formula (36). Also, the fitting results may be compiled into a table, and the gain guvw of each phase according to the current may be determined by referring to the table. Moreover, the gain guvw may be determined by a function to which a higher-order term is added in accordance with equation (36).

[0139] The change in signal amplitude is not due to the gain of the current differential detector 4uvw. At an electrical angle of 0°, the magnet is excited on the d-axis, strengthening the magnetic flux, and at an electrical angle of 180°, the magnet is excited on the reversed d-axis, weakening the magnetic flux. Weakening the magnetic flux of the magnet approaches a state where there is no magnet, and the position dependency of the inductance caused by saturation of the core disappears, causing a change in the signal amplitude.

[0140] By performing a correction in which a gain according to the current is multiplied by the three-phase signals for position estimation, it becomes possible to use a detection element made of a magnetic material such as a current transformer in the current differential detector 4uvw even when the current ripple is minute, making it possible to detect the current differential value with high sensitivity. Even if a current differential detection element such as a current transformer is not used, it is generally difficult to make the current differential detection gains of all UVW phases completely the same. When the current differential detection gains of the three phases are different, using this calculation process can reduce position estimation errors.

[0141] In addition, in the sensorless control using the estimated position as described above, there is uncertainty due to the fact that the estimated position appears two periods for one period of the motor electrical angle. Therefore, there is a risk that the initial excitation position will be in the opposite phase. If this is a problem, for example, an initial position estimation method using magnetic saturation (for example, see Non-Patent Document 2) may be used in combination to determine the initial excitation position. In this embodiment, since the estimated position is obtained on the αβ fixed coordinate system, if the excitation phase shift of the initial excitation due to the initial excitation position being in the opposite phase does not become a problem, the two-period signal of the estimated position can be converted into a one-period signal and directly used in the coordinate system of the dq transformation, so that the motor can be rotated synchronously without performing initial position estimation.

[0142] Next, a specific example of position estimation (second estimation method) by the second position estimator 152 will be described. In this embodiment, the second position estimator 152 estimates the rotor position on the αβ coordinate system by an extended induced voltage observer. This method is described in detail in Non-Patent Document 1 and Patent Document 7, and therefore will only be briefly described here.

[0143] The voltage equation of the motor in the αβ coordinate system can be written as the following equation (37) using the differential operator p (time differential operator).

[0144]

number

[0145] Inductance L in the αβ coordinate system α ,L β ,L αβ is expressed as equation (39) using inductances Ld, Lq in the dq coordinate system. Furthermore, by transforming equation (37) using the currents id, iq (= Id, Iq) in the dq coordinate system, equation (38) is obtained. It is assumed that there is no angle dependency of inductance in the dq coordinate system, and that the time derivatives of inductances Ld, Lq are zero. The second term of equation (38) includes terms that depend on the motor position θ, and this is defined as the extended induced voltage vector e (equation (40)). Note that the symbol "· " is a differential operator that represents the first-order time derivative of the variable in question only (only iq in equation (38)). The same applies below.

[0146]

number

[0147] By transforming equation (38) to derive the state equations, we obtain equations (41) and (42). In deriving equation (42), which is the time derivative of the extended induced voltage, the first time derivative of the d-axis current id, the second time derivative of the q-axis current iq, and the first time derivative of the angular velocity ω are all approximated to zero.

[0148]

number

[0149] The minimum-order state observer (extended back electromotive force observer) for calculating the extended back electromotive force estimated value e^ (where the symbol "^" represents an estimated value; the same applies below) is given by equations (43) and (44). The observer gain G may be determined, for example, so as to be proportional to the absolute value of the speed.

[0150]

number

[0151] Substituting equation (43) into equation (44) and rearranging it while noting that GJ = JG, we obtain equation (45).

[0152]

number

[0153] In order to eliminate the current differential detection value, a parameter ξ = e^ + Gi is introduced and transformed to obtain equation (46), from which the first-order time differential of ξ can be found using the current detection value i and voltage detection value v.

[0154]

number

[0155] Therefore, ξ can be obtained by time integration of equation (46), and the extended induced voltage estimate ê can be obtained by performing variable transformation again as in equation (47). The estimated position is given by taking the argument of the extended induced voltage as shown in equation (48).

[0156]

number

[0157] The second position estimator 152 (see FIGS. 2A and 2B) receives the current detection value I α ,I β is obtained and used as the current i in equation (46), and the current in the αβ fixed coordinate system is obtained from the inverse dq converter 132 (see FIG. 1B). Pressure The command values ​​Vαcmd and Vβcmd are obtained and used as the voltage detection value v in equation (46). As a result, the second position estimator 152 calculates the induced voltage estimation value e^(e^ α ,e^ β ) is calculated, and then the estimated rotor position θ2 is calculated using equation (48).

[0158] Next, the period conversion of the estimated position signal generated by first position estimator 151 (see FIGS. 2A and 2B), that is, the operation of period converter 154, will be described. First position estimator 151 performs position estimation based on the variation in inductance accompanying the rotation of the rotor, in other words, by utilizing the dependency of inductance on the rotational position of the rotor. Therefore, the estimated position signal generated by first position estimator 151 has two cycles of variation per rotor rotation corresponding to one electrical angle period of the stator. Therefore, as described above, the estimated position signal generated by first position estimator 151 is converted by period converter 154 (see FIGS. 2A and 2B) into a first estimated position signal having one cycle of variation per rotor rotation corresponding to one electrical angle period of the stator.

[0159] In the low-speed range, a signal that fluctuates two periods per electrical angle period is obtained by a first position estimator 151 that calculates the position information contained in the motor inductance by detecting the current ripple, as shown in equations (11) and (14), etc. In the medium- to high-speed range, a signal that fluctuates one period per electrical angle period is obtained by a second position estimator 152 that calculates the estimated position according to equations (47) and (48) using an extended induced voltage observer.

[0160] As a specific example, a surface permanent magnet synchronous motor with a rated rotation speed of 3000 [r / min] was rotated under unloaded conditions at 160 [r / min], and position estimation was performed using the two estimation methods described above, with the results plotted over two electrical angle periods, as shown in Fig. 23. It can be seen that position estimation using extended induced voltage (second estimation method using second position estimator 152) provides an estimated position for one period per rotor electrical angle rotation, whereas position estimation using pulse application (first estimation method using first position estimator 151) provides an estimated position for two periods per rotor electrical angle rotation.

[0161] In position estimation by pulse application (first estimation method), a signal with two periods per electrical angle is generated, and therefore conversion to one period is required. This period conversion is performed by period converter 154 (see FIGS. 2A and 2B). FIG. 24 shows an example of processing for period conversion by the period converter 154. This processing is repeatedly executed at a predetermined operation period. In the following description, a "one-cycle signal" refers to a signal that repeats a monotonically increasing change from a lower limit value (specifically, 0) to an upper limit value msk (a value corresponding to one cycle of the electrical angle, for example, 2048) in accordance with the rotor position corresponding to one cycle of the electrical angle. Also, a "two-cycle signal" refers to a signal that repeats a monotonically increasing change from a lower limit value to an upper limit value msk / 2 (half of the value corresponding to one cycle of the electrical angle) in accordance with the rotor position corresponding to half a cycle of the electrical angle. That is, a one-cycle signal is a signal having one cycle during one cycle of the electrical angle, and a two-cycle signal is a signal having two cycles during one cycle of the electrical angle. As shown in FIG. 23, the second estimation method using the extended induced voltage generates a one-cycle signal, and the first estimation method using pulse application generates a two-cycle signal. As understood from FIG. 23, the two-cycle signal has a section that coincides with the one-cycle signal and a section that is shifted in value by half of the electrical angle with respect to the one-cycle signal.

[0162] According to the procedure described in FIG. 24, if a correct one-cycle signal is obtained even once at a certain time, a one-cycle signal can be continuously generated only by the current two-cycle signal (the two-cycle signal in the current operation cycle) in each operation cycle. Specifically, the absolute value wk1 of the difference between the one-cycle signal value 1cyc_t-1 in the previous (previous operation cycle) and the two-cycle signal value 2cyc_t in the current (current operation cycle) is obtained (step S1). It is determined whether the absolute value wk1 of the difference is less than one-fourth of the value msk (for example, 2048) of one electrical angle (step S2. First determination). If wk1 < msk / 4 (step S2: YES), it is determined that the periods match, and the two-cycle signal value 2cyc_t is temporarily used as the current one-cycle signal value 1cyc_t as it is (step S3). If wk1 ≧ msk / 4 (step S2: NO), it is determined that there is a half-cycle shift, and 2cyc_t + msk / 2 obtained by adding half of the value msk of one electrical angle to the two-cycle signal value 2cyc_t is temporarily used as the current one-cycle signal value 1cyc_t (step S4).

[0163] Based on only the processing according to this first determination, when both the previous one-cycle signal value 1cyc_t-1 and the current two-cycle signal value 2cyc_t are values near 0 which are cycle boundaries, it may not be possible to process correctly due to measurement noise. For example, when the value of one electrical angle msk = 2048 and the ideal signal values when the rotor position is 0 are 1cyc_t-1 = 2cyc_t = 0, but assume that signal values such as 1cyc_t-1 = 2038 and 2cyc_t = 10 are obtained due to noise. At this time, it would be desirable for the current one-cycle signal value 1cyc_t to be 10. However, when performing the processing based on the above first determination, wk1 = 2028 (= 2038 - 10), and since wk1 < 512 (= msk / 4) is not satisfied (step S2: NO), the one-cycle signal value 1cyc_t = 10 + 1024 = 1034 (step S4), and a value shifted by half a cycle is output.

[0164] To avoid this, the absolute value wk2 of the difference between the one-cycle signal value 1cyc_t obtained by performing the processing (steps S3, S4) based on the above first determination (step S2) and the one-cycle signal 1cyc_t-1 of the previous operation cycle is obtained (step S5), and a second determination is further performed (step S6). That is, if msk / 4 < wk2 < 3*msk / 4 (step S6: YES), it is determined that the one-cycle signal value 1cyc_t obtained based on the first determination (step S2) (steps S3, S4) is shifted by half a cycle, and the value is corrected and output as 1cyc_t = 1cyc_t - msk / 2 (step S7). If the determination in step S6 is negative, since the one-cycle signal value 1cyc_t obtained by the processing (steps S2, S3, S4) based on the first determination is the correct value, the correction in step S7 is not performed.

[0165] In the above example, the one-cycle signal value was determined to be 1cyc_t=1034 using only the process based on the first judgment (steps S2, S3, S4), but adding the process based on the second judgment (steps S6, S7) results in wk2=|2038-1034|=1004, which satisfies the condition of being greater than or equal to 512 (=msk / 4) and less than or equal to 1536 (=3*msk / 4). Therefore, the one-cycle signal value is corrected to the correct position of 1cyc_t=1034-1024=10, and the correct one-cycle signal value is obtained even if the measurement value contains noise.

[0166] The obtained one-period signal value 1cyc_t is further subjected to a remainder operation (Mod operation (operator %)) by the value msk of one electrical angle period (step S8), and a one-period signal value 1cyc_t in the range of 0 to msk (=2048) is output. n (where n is a natural number), and the Mod operation may be replaced with a bitwise AND operation.

[0167] The correct one-period signal to be applied at a certain time can be obtained by using, for example, an initial position estimation method that utilizes the saturation of the iron core (see Non-Patent Document 2). Fig. 25 shows the result of converting the two-period signal obtained by the first estimation method, which obtains an estimated position from the current ripple, into a one-period signal according to the procedure in Fig. 24 when the motor is in the same driving state as in Fig. 23. The value msk of one electrical angle is set to 2048. It can be seen that an excellent one-period signal is generated.

[0168] In this way, two types of estimated position signals having one period for one period of rotor rotation of an electrical angle on the αβ fixed coordinate system are obtained by the first estimation method and the second estimation method. Therefore, depending on the estimated speed or the extended induced voltage vector length, the estimated position signal to be used is switched, or the two types of estimated position signals are weighted and combined. This makes it possible to obtain a reliable estimated position over the entire speed range, thereby enabling sensorless control over the entire speed range.

[0169] 26 shows an example of processing by the estimated position combiner 153 (see FIGS. 2A and 2B). In the configuration of FIG. 2A, the estimated position combiner 153 receives the first and second estimated position signals θ 1C ,θ 2C In the configuration of FIG. 2B, the first and second estimated position signals θ1 and θ2 before compensation are weighted and combined. law The first estimated position signal θ 1C The rotor estimated position represented by θ1 is collectively referred to as the "first estimated position θ PWM Similarly, the second estimated position signal by the second estimation method using the extended induced voltage is θ2C The rotor estimated position represented by θ2 is collectively referred to as the “second estimated position θ EMF " Let's call it that.

[0170] The weighting of the estimated position is a sawtooth wave weighting calculation. Specifically, the first estimated position θ PWM The second estimated position θ EMF A difference dif between the values ​​msk and msk is calculated (step S11). It is then checked whether this difference dif is outside a range of half the value msk of one electrical angle period (for example, 2048). That is, it is checked whether dif<-msk / 2 (step S12) and whether dif>msk / 2 (step S14). If dif<-msk / 2 (step S12: YES), a value msk+dif obtained by adding the difference dif to the value msk of one electrical angle period is substituted for variable wk3 (step S13). If dif>msk / 2 (step S14: YES), a value msk-dif obtained by subtracting the difference dif from the value msk of one electrical angle period is substituted for variable wk3 (step S15). If the difference dif is within a range of half the value msk of one electrical angle period (-msk / 2≦dif≦msk / 2) (NO in both steps S12 and S14), the difference dif is substituted into the variable wk3 (step S16). Then, the variable wk3 is weighted by a weighting coefficient ρ (0≦ρ≦1) according to the following equation (49). 3 The weighted value ρ·wk3 is the first estimated position θ PWMThe remainder is calculated by modulating the value of one electrical angle msk (operator %) to obtain the weighted composite estimated position θ WT is calculated (step S17).

[0171] θ WT =(θ PWM +ρ×wk3)%msk …(49) If wk3=dif (step S16), then θ PWM +ρ×wk3=θ PWM +ρ(θ EMF -θ PWM )=(1-ρ)θ PWM +ρθ EMF The first estimated position θ PWM and the second estimated position θ EMF It can be seen that the signals are weighted and combined. The modulus operation Mod (operator %) is an operation for making the combined result into an estimated position signal that changes periodically in a sawtooth wave shape in the range of 0 to msk.

[0172] For example, when the true rotor position value is "5", the first estimated position θ PWM = 2038 and the second estimated position θ EMF Assume that the weighting coefficient ρ is 0.5. The weighted composite estimated position θ WT In order to obtain such a result, it is appropriate to calculate the two estimated positions θ PWM ,θ EMF A difference dif between the two estimated positions is calculated (step S11), and if the difference dif is within ±msk / 2, that value is used; if it exceeds msk / 2, a value obtained by subtracting the difference dif from one electrical angle value msk is used; if the difference dif is smaller than -msk / 2, a value obtained by adding the difference dif to one electrical angle value msk is used as a variable wk3 representing the distance between the two estimated positions (steps S12 to S16). After multiplying the variable by a weighting coefficient ρ, PWM ,θ EMF (In the example of FIG. 26, the first estimated position θ PWM ) and add it to

[0173] In the above example, the difference dif=-2018 (=20-2038) is obtained, so the variable wk3=2048-2018=30. Therefore, the composite estimated position θ WT =(2038+0.5×30)%2048=5, which shows that the correct weighting result is obtained. 26 uses branch processing, but a method for performing high-speed processing using Mod or bit shift operations is shown in equation (50). The operation result is equivalent to that in the procedure of FIG.

[0174] The variable wk in equation (50) is the calculation result equivalent to wk3 in FIG. 26, and θ WT =θ PWM , ρ=1 and θ WT =θ EMF It is possible to obtain a weighting result such that

[0175]

number

[0176] For example, θ PWM and θ EMF Each has a certain amount of noise, and θ EMF θ PWM When offset by +512, the composite estimated position θ when ρ is changed in the range of 0≦ρ≦1 WT The results of calculation using equation (50) are shown in FIG. WT When ρ=0, θ PWM and when ρ=1, θ EMF Furthermore, θ WT When ρ is between 0 and 1, the weighting is smooth with the change in ρ, including the noise level. In this example, we use θ EMF is offset, but in reality, EMF and θ PWM Of course, even in this case, it is possible to correctly weight and combine the signals by using equation (50).

[0177] By changing the weighting coefficient ρ according to the estimated speed or the vector length of the extended induced voltage vector e (see equation (40)), the first estimation method and the second estimation method can be appropriately weighted according to the rotor rotation speed, and the estimation method can be smoothly changed. An example is shown in equation (51). In equation (51), the estimated speed is ω, the weighting start speed is ω, s , the weighted end speed is ω e The weighting coefficient ρ is changed in proportion (linearly) to the estimated speed ω. s In the low speed range below ρ = 0, and only the estimation results by the first estimation method are used. Similarly, the weighted end speed ω e In the high speed range exceeding ρ=1, and only the estimation results by the second estimation method are used.

[0178]

number

[0179] The vector length |e| of the extended induced voltage may be used instead of the estimated speed ω, and the weighting coefficient ρ may be defined as a different function (nonlinear function) instead of proportional (linear). In equation (51), ω s ≒ω e Then, the speed ω s The weighting coefficient ρ switches from 0 to 1. This essentially corresponds to switching the estimation method at a specific speed instead of weighting, that is, switching synthesis.

[0180] The weighted synthesis in the configuration of FIG. 2A is expressed by the first line of equation (52). In the equation, Δθ1 represents the compensation amount by the first compensator 161, and θ 1C Similarly, Δθ2 represents the amount of compensation by the second compensator 162, and θ 2C=θ2+Δθ2. The first line of equation (52) can be transformed into the third line, which expresses the weighted combining using the configuration of FIG. 2B. The first term in the third line of equation (52) represents the weighted combining before compensation, and the second term in the third line of equation (52) represents the compensation after weighted combining. Therefore, substantially equivalent processing is possible with either the configuration of FIG. 2A or FIG. 2B.

[0181] θnew=(1-ρ)θ 1C +ρθ 2C =(1-ρ)(θ1+Δθ1)+ρ(θ2+Δθ2) =(1-ρ)θ1+ρθ2+{(1-ρ)Δθ1+ρΔθ2} …(52) The compensation (correction) by the first compensator 161 may be the process described above with reference to FIGS. 15 and 16 (see equations (34) and (35)).

[0182] As an example, the weighting start speed ω s = 600 [r / min], weighting end speed ω e The controller 1 (see FIG. 1A) was designed (programmed) to calculate a weighted composite estimated position θnew with ρ=1200 [r / min]. However, compensators 161, 162, and 163 (see FIG. 2A and FIG. 2B) were omitted. The speed of the AC motor M was controlled by this controller 1, and it was rapidly accelerated from 0 to 1600 [r / min] in 30 [ms (milliseconds)]. The results of plotting the transition of the weighting coefficient ρ, the time change of the estimated speed ω, and the composite estimated position θnew at this time are shown in FIG. 28.

[0183] It can be seen that the weighting coefficient ρ changes linearly depending on the speed between 600 and 1200 [r / min], and the weighted composite estimated position θnew changes smoothly without large fluctuations even during sudden acceleration. For example, Patent Document 8 discloses a control method for a synchronous motor that realizes rapid and continuous acceleration and deceleration performance between low and high speed regions without the need for a magnetic pole position detector. Patent Document 8 shows the results of an experiment in which the forward and reverse rotation of ±1300 [r / min] is performed at an acceleration rate of 1.5 [s]. The acceleration rate in the above example is about 30 times that of Patent Document 8.

[0184] Weighted speed range ω s ~ω e It is possible to narrow the range of ω s ~ω e This may cause chattering and lead to unstable control. s and the final velocity ω e may be determined based on the upper limit of the speed at which the first estimation method using pulse application can be applied and the lower limit of the speed at which the second estimation method using extended induced voltage estimation can obtain sufficient accuracy.

[0185] In another experiment, while the AC motor M was being speed-controlled at a command speed of 3000 [r / min] using the controller 1 set as in the above embodiment, a sudden load was applied and the AC motor M was suddenly stopped in a deceleration time of about 30 [ms]. The changes over time of the weighting coefficient ρ, the estimated speed ω, the weighted composite estimated position θnew, and the q-axis current iq in this case are shown in Fig. 29. During the transition from the medium-high speed range to the low speed range, that is, during the transition from the second estimation method to the first estimation method, a large error of about 30 degrees in electrical angle occurs in the composite estimated position, as indicated by reference symbol PE. This occurs when there is a large deviation between the two estimation results obtained by the first and second estimation methods, and if the error in the composite estimated position becomes even larger, step-out may occur. This is due to the effect of the estimated position being shifted by the change in the motor inductance phase caused by the application of the q-axis current iq in the first estimation method using pulse application (see Figures 14A and 14B above), and the effect of a delay (delay in the estimated value relative to the speed) caused by the filter or the like used for the second estimation method using the extended induced voltage.

[0186] This problem can be solved by the action of the first and second compensators 161, 162 (see FIG. 2A). That is, the first and second estimated position signals θ1, θ2 obtained by the first and second estimation methods are compensated in accordance with the dq-axis currents id, iq and the estimated speed ω, and the first and second estimated position signals θ 1C ,θ 2C are weighted and combined. Compensation proportional to the linear combination a·iq+b·ω (where a and b are constants) of the q-axis current iq and estimated speed ω was added to the first and second estimated position signals θ1, θ2, and an experiment was conducted under operating conditions similar to those of the experimental example in Fig. 29. The results are shown in Fig. 30. Even during the period when the first and second estimation methods transition due to a sudden stop caused by a sudden load, no substantial error occurs in the combined estimated position, and a smooth combined estimated position is obtained.

[0187] This experimental example reproduces the case of a sudden stop due to a sudden load, but the compensation described above will also have a similar effect when the position estimation method transitions when a large q-axis current iq is flowing due to an external load. The degree of compensation can be set by adjustments made in advance. For example, an encoder for position detection can be attached to the motor to detect the ideal position (true position), and the degree of compensation can be determined so that the deviation between the estimated position and the ideal position is small for any dq-axis current or speed. If such adjustments are made and the degree of compensation is expressed as a function or table, then a position detector will no longer be necessary.

[0188] The above compensation is possible because the position estimators 151, 152 output an estimated position on the motor fixed coordinate system. If a position estimator that outputs a position error Δθ on the motor rotating coordinate system is used, the estimated position cannot be corrected directly, and it is necessary to adjust the motor parameters for calculating the position error Δθ so that a true estimated position can be obtained, which is difficult.

[0189] As described above, it has been demonstrated that highly responsive and stable sensorless control can be achieved by performing two types of estimated position on a fixed coordinate system, estimation using current ripple (first estimation method) and estimation using extended induced voltage (second estimation method), and transitioning between them. In the speed range (medium-to-high speed range) where estimation using extended induced voltage (second estimation method) is performed, position estimation using current ripple (first estimation method) is essentially unnecessary, except when responding to sudden deceleration.

[0190] The voltage vector pattern applied for the first estimation method narrows the time width that can be used to apply the drive voltage for driving the motor (the time during the PWM control period during which the drive voltage is applied), and causes problems with high-frequency noise due to current oscillations with a period equal to the product of the number of voltage vector patterns and the PWM control period. To prevent these problems, a speed range (weighting end speed ω e In the high speed range exceeding 100 rpm), it is preferable to control the PWM generator 14 (see FIG. 1B) so as not to apply the position detection voltage vector patterns of FIG. 6A and FIG. 6B.

[0191] Although the embodiment of the present invention has been described above, the present invention can be embodied in other forms. 1B shows a configuration in which a command position is given from the outside to perform position control, but it is also possible to omit position controller 11 and give a command speed from the outside to speed controller 12 to perform speed control. Furthermore, it is also possible to omit speed controller 12 and give a command current from the outside to dq current controller 131 to perform torque control.

[0192] Moreover, a voltage for compensating for the speed electromotive force may be added to the output of the dq current controller 131 to perform decoupling control of the dq axes, or a field weakening control may be applied according to the speed. As another modification, the observer expressed by the equations (43) and (44) is used to estimate the extended electromotive force in the description of the embodiment above, but the extended electromotive force vector expressed by the equation (40) may be used as it is. Also, the weighted position estimator 15 may be configured using another position estimator that can calculate an estimated position on the motor fixed coordinate system without estimating the position error Δθ.

[0193] The weighted position estimator 15 may also be configured to include three or more position estimators that output an estimated position on a motor fixed coordinate system, and to combine the estimated positions obtained by these estimators. For example, three position estimators, a position estimator that uses a current ripple, a position estimator that uses an extended induced voltage, and a position estimator that uses a magnetic flux observer, may be used as position estimators that output an estimated position on a fixed coordinate system, and the estimated positions obtained by these estimators may be weighted and combined. This allows the configuration of a more accurate position estimator.

[0194] 1B, the current ripple (current differential value) is directly detected by the current differential detector 4uvw, but instead of providing the current differential detector 4uvw, the change amount (variation) of the current may be detected. For example, the current values ​​before and after the application of the position detection voltage vector may be detected, and the difference between them may be used as a value representing the current ripple. In addition, various design modifications can be made within the scope of the claims. [Explanation of symbols]

[0195] 1: Controller 2: Inverter 3u ,3v,3w : Current detector 4u ,4v,4w : Current differential detector 5u ,5v,5w : Winding 11: Position controller 12: Speed ​​controller 13: Current controller 14:PWM generator 15: Weighted position estimator 16: Speed ​​estimator 100: Motor control device 131 :dq current controller 132: Inverse dq converter 135 :dq converter 151: 1st position estimator 152:Second position estimator 153: Estimated position synthesizer 154: Period converter 161: 1st compensator 162:Second compensator 163:Synthetic position compensator

Claims

1. A motor control device that controls an AC synchronous motor by sensorless control without using a rotor position sensor, a first position estimator that estimates a position of a rotor of the AC synchronous motor on a fixed coordinate system according to a first estimation method; a second position estimator that estimates a position of the rotor of the AC synchronous motor on a fixed coordinate system in accordance with a second estimation method different from the first estimation method; a drive control means for driving the AC synchronous motor based on the estimation results of the first position estimator and the second position estimator; Including, the drive control means applies to the AC synchronous motor a voltage vector corresponding to a voltage command for driving the AC synchronous motor, and applies to the AC synchronous motor a position detection voltage vector for detecting a rotor position of the AC synchronous motor during a period in which the voltage vector corresponding to the voltage command is a zero voltage vector; the first position estimator estimates a rotor position by capturing a change in inductance of the AC synchronous motor based on a current ripple generated in a winding current of the AC synchronous motor when the position detection voltage vector is applied to the AC synchronous motor.

2. The motor control device according to claim 1 , wherein the first estimation method and the second estimation method both estimate the rotor position without estimating an error in the rotor position.

3. 2. The motor control device according to claim 1, wherein the first estimation method and the second estimation method both estimate the rotor position without using a PLL (phase-locked loop) control that outputs an estimated rotor speed so that an error in the rotor position becomes zero.

4. the first position estimator outputs an estimated position signal having a fluctuation of two periods per rotation of the rotor corresponding to one electrical angle period of the stator; the second position estimator outputs an estimated position signal having a fluctuation of one period per rotation of the rotor corresponding to one electrical angle period of the stator; The motor control device according to any one of claims 1 to 3.

5. the first position estimator outputs an estimated position signal having a fluctuation of two periods per rotation of the rotor corresponding to one electrical angle period of the stator; the second position estimator outputs an estimated position signal having a fluctuation of one period per rotation of the rotor corresponding to one electrical angle period of the stator; 4. The motor control device according to claim 1, further comprising a periodic converter that converts the estimated position signal of the first position estimator into an estimated position signal that is a periodic signal having one period of fluctuation per rotation of the rotor corresponding to one electrical angle period of the stator.

6. 6. The motor control device according to claim 1, wherein the second position estimator estimates the position of the rotor based on an extended induced voltage estimate value.

7. an estimated position combiner that generates a combined estimated position by switching or weighting a first estimated position signal output from the first position estimator and a second estimated position signal output from the second position estimator in response to a rotation speed of the rotor or an extended induced voltage vector length, 7. The motor control device according to claim 6, wherein the drive control means drives the AC synchronous motor in accordance with the composite estimated position generated by the estimated position combiner.

8. a first compensator and a second compensator for compensating the first estimated position signal and the second estimated position signal, respectively, in response to a motor current and a rotor rotation speed, or The motor control device of claim 7 , further comprising a composite estimated position compensator that compensates the composite estimated position in response to a motor current and a rotor rotation speed.

9. the estimated position synthesizer generates the synthesized estimated position without using the first estimated position signal in a high speed region where the rotational speed of the rotor is equal to or higher than a predetermined value; 9. The motor control device according to claim 7, wherein the drive control means stops application of the position detection voltage vector in the high speed range.

10. An AC synchronous motor; an inverter for supplying an AC current to the AC synchronous motor; A motor control device according to any one of claims 1 to 9, which controls the inverter; A drive system including:

Images