JP2016052232A - Control device for ac motor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a control device for an AC motor which reduces phase delay of a filter for cancelling a harmonic pulsation component and a switching noise.SOLUTION: In a control device for an AC motor, a moving average filter calculates a moving average for a predetermined integration term T for at least one of a d-axis current Id and a q-axis current which are obtained by performing dq transformation on a phase current detection value. On a frequency characteristic chart in which a lateral axis is adjusted in such a manner that gains in a low frequency domain are equalized between the moving average filter and a primary delay filter, in a high frequency domain in which an angular frequency ω is equal to about 2 to 10×f, the gain of the moving average filter is smaller than that of the primary delay filter. In particular, if the angular frequency ω is equal to 5 or 10×f, the gain becomes 0. In a nearby frequency, the gain is also sufficiently small, such that an amplitude can be significantly reduced in comparison with the primary delay filter. Namely, by appropriately setting the integration term T, a filter having a stronger amplitude reduction capability than the primary delay filter can be provided and disturbance can be significantly cancelled.SELECTED DRAWING: Figure 4

Description

  The present invention relates to a control device for an AC motor that controls energization of the AC motor by feedback control.

  Conventionally, a control device that controls energization of an AC motor (motor) is known. In motor control, a sinusoidal phase voltage including a primary rotation component is applied to each phase from an inverter. However, since the inverter cannot output a strict rotation primary component, the phase voltage includes a rotation primary harmonic component. This harmonic component voltage also appears in the phase current. Therefore, the phase current includes a primary rotation component that is originally intended to be input and a harmonic component that is a disturbance.

Normal motor control is performed based on a theory constructed on the assumption that these harmonic components do not exist. Therefore, the presence of harmonic components leads to deterioration in control accuracy of torque and rotation speed, which are the purpose of motor control.
In general, vector control, which is a theory generally used for motor control, develops and controls various electromagnetic states on a dq axis that rotates in synchronization with a rotation angle. Here, if there is no harmonic component and the motor operation state is constant, the dq-axis current becomes a constant value. However, since there is a harmonic component, it has a pulsating element having a frequency that is an integral multiple of the rotational first order.

Therefore, conventionally, a technique for suppressing a pulsation component included in a current or torque detection signal in an AC motor control device is known.
For example, the electric motor control device disclosed in Patent Document 1 adds or subtracts the correction current command generated by the cycle Id generator (10) and the cycle Iq generator (11) with respect to the d-axis and q-axis current command values. The pulsation of each frequency component is canceled by adding and subtracting with the device (19). The periodic Id generator and the periodic Iq generator include a first-order lag filter (252) that extracts a pulsation component from a current fluctuation component, and an infinite integration filter (= integration) that integrates a deviation between the pulsation component and a zero signal and outputs a corrected current value. Controller) (253) and the like.
Moreover, the control apparatus disclosed by patent document 2 extracts the primary component which carried out the Fourier series expansion | deployment of the phase current detection value. The devices of Patent Documents 1 and 2 both calculate a dq-axis current detection value from a phase current detection value, and apply a filter to the dq-axis current detection value to suppress harmonic components.

Japanese Patent No. 4958431 JP 2014-132815 A

  In the first-order lag filter, the amplitude of the pulsation component cannot be reduced to zero, and a phase lag occurs before and after the filter. If the high-frequency rejection characteristic is increased with priority given to the removal of the pulsation component, the phase delay increases. Generally, feedback is provided when the phase delay of the dq axis current is 60 deg or more (in FIG. 1 of Patent Document 1, it is constituted by a period Id generator (10), a period Iq generator (11), and an adder / subtractor (19)). The effect cannot be reduced. Divergence may occur when the phase delay is 180 degrees or more. As described above, when the phase delay is relatively increased in order to reduce the amplitude of the harmonic component, there is a problem that the control becomes unstable.

  The integration filter exhibits excellent characteristics for removing high-frequency components, but the output tends to largely inherit the past influence and the phase delay tends to increase. The phase of an integral filter that inherits an infinite past, such as the integral controller (253) used in the apparatus of Patent Document 1, is delayed by 90 degrees. Further, even with an integral filter that inherits an infinite past, the amplitude of the pulsating component cannot be reduced to zero.

In the control of an AC motor, there is a demand for removing switching noise and the like in addition to harmonic pulsation components.
The present invention has been made in view of the above-described problems, and an object of the present invention is to provide an AC motor control device that reduces the phase delay of a filter that removes harmonic pulsation components and switching noise.

The present invention detects a phase current flowing in a multi-phase AC motor having three or more phases and converts it into a d-axis current and a q-axis current, and energizes the AC motor by feedback control using the d-axis current and the q-axis current. The invention relates to a control device for an AC motor to be controlled.
Here, the “AC motor” includes an AC drive motor, a generator, and a motor generator. For example, a motor that is used as a main machine of a hybrid vehicle or an electric vehicle and generates torque for driving drive wheels. Applicable to generators. In addition, for example, an electric motor control device that drives a motor generator corresponds to an “AC electric motor control device”.

This AC motor control device includes a “moving average filter” that calculates and outputs a moving average in a “finite predetermined section” for at least one of the d-axis current and the q-axis current, or at least one phase current. It is characterized by that.
The moving average filter according to the present invention appropriately sets a predetermined section and calculates a moving average in the section, thereby suppressing the amplitude of a component to be removed such as a harmonic pulsation component and switching noise, and a conventional first-order lag. The phase delay can be reduced with respect to a filter or an infinite time integration filter.

The moving average filter according to the first configuration of the present invention calculates a moving average so as to suppress pulsation due to a harmonic component of a specific order for at least one of the d-axis current and the q-axis current.
By this moving average process, it is possible to reduce the phase delay of the dq-axis current, which is the main information of control, while largely removing the harmonic component of the target frequency.

In this case, it is preferable that the moving average filter sets the section length of a predetermined section for calculating the moving average according to the rotation speed of the AC motor.
Specifically, the section length is set according to the rotation-synchronized harmonic fluctuation component generated by the inverter control. For example, by setting a section length including a multiple of “1 / n of the rotation primary period”, the n-order harmonic component can be removed. In particular, when the section length is set to “1 / n of the rotation primary period”, the amplitude of the n-order harmonic component is set to 0, and the influence can be completely removed. Here, in the multi-pole pair AC motor, the term “rotary primary” in the present invention is interpreted in the meaning of “electric primary”.
On the other hand, if the section length is set shorter in accordance with the order component having a large amplitude, the target order component can be removed while the phase delay is made smaller.

  Furthermore, it is preferable that the moving average filter sets the section length to a length of 1/6 of the primary rotation period of the AC motor. As a result, the phase lag of the dq-axis current can be reduced while effectively suppressing the rotational sixth-order component, which is the dominant component of the disturbance, and the multiple order components (such as the twelfth-order component).

  In addition, the moving average filter has a convergence value estimation unit that estimates a convergence value of the d-axis current or the q-axis current, and the difference between the input d-axis current or the q-axis current and the convergence value is a predetermined interval. A moving average may be calculated.

  The moving average filter of the 2nd structure of this invention calculates a moving average so that the switching noise of the inverter which supplies electric power to an alternating current motor may be removed about a phase current of at least one phase. In this configuration, before dq conversion, by calculating the moving average of the phase current detection value in an integration period corresponding to several sampling cycles, the increase / decrease in the phase current due to the switching noise is canceled, and the switching noise is preferably reduced. Can be removed.

1 is a block diagram of an AC motor control device according to a first embodiment of the present invention. FIG. It is a detailed block diagram of the moving average filter of FIG. It is a figure explaining the example of application of a moving average process. It is a frequency characteristic diagram of (a) gain and (b) phase for comparing a moving average filter, a first-order lag filter, and an infinite integral filter. It is a figure explaining the moving average process which suppresses the (a) 6th-order component and (b) 12th-order component superimposed on the q-axis current. It is a block diagram of the control apparatus of the alternating current motor by 2nd Embodiment of this invention. It is a figure explaining switching noise. It is a figure explaining the removal of the switching noise by a moving average process, (a) Schematic diagram of phase current before filter processing, (b) VIIIb part enlarged view, (c) Schematic diagram of phase current after filter processing. .

Embodiments of an AC motor control device according to the present invention will be described below with reference to the drawings.
The AC motor control device of this embodiment is applied as control for controlling energization of a motor generator that is a driving force source of a hybrid vehicle, for example. Hereinafter, in a plurality of embodiments, the same numerals are given to the substantially same composition, and explanation is omitted.

(First embodiment)
A first embodiment of the present invention will be described with reference to FIGS.
As shown in FIG. 1, the motor control device 101 operates the switching element of the inverter 12 by outputting three-phase voltage commands Vu ^* , Vv ^* , and Vw ^* to the inverter (indicated as “INV” in the figure) 12. And an apparatus for controlling energization of the AC motor (denoted as “MG” in the figure) 2. The AC motor 2 of the present embodiment is a permanent magnet type synchronous three-phase AC motor, and in a hybrid vehicle, an electric motor that performs power running by consuming battery power, and a generator that regenerates the generated power to the battery. It is used as a motor generator having the above functions.

  A rotation angle sensor 14 such as a resolver provided near the rotor of the AC motor 2 detects the electrical angle θ of the AC motor 2 and communicates with the motor control device 101. The electrical angle θ is converted into an angular frequency ω1 by the differentiator 37. The angular frequency ω1 is converted to “the number of rotations of the AC motor” described in the claims by multiplying by a proportionality constant.

  The motor control device 101 is configured by a microcomputer or the like, and includes a CPU, a ROM, an I / O (not shown), a bus line that connects these configurations, and the like. The motor control device 101 controls the operation of the AC motor 2 by software processing by executing a program stored in advance by the CPU or hardware processing by a dedicated electronic circuit.

The motor control device 101 of the first embodiment includes a torque subtractor 32, a PI controller 33, a voltage command conversion unit 34, an inverse dq conversion unit 35, a dq conversion unit 36, a differentiator 37, a current / torque conversion unit 38, and The moving average filter 40 is provided.
The torque subtractor 32 calculates a torque deviation Δtrq that is a difference between the torque calculation value trq fed back from the current / torque conversion unit 38 and the torque command value trq ^* .
The PI controller 33 calculates the voltage phase command VΨ by PI calculation so that the torque deviation Δtrq converges to 0 so that the torque calculation value trq follows the torque command value tr ^* .

The voltage command conversion unit 34 acquires the voltage amplitude command Va ^* and the voltage phase command VΨ and converts them into dq axis voltage commands Vd ^* and Vq ^* .
The inverse dq converter 35 converts the dq axis voltage commands Vd ^* and Vq ^* into three-phase voltage commands Vu ^* , Vv ^* and Vw ^* based on the electrical angle θ. On / off of the switching element of the inverter 12 is operated by a drive signal generated based on the three-phase voltage commands Vu ^* , Vv ^* , Vw ^* , and desired three-phase AC voltages Vu, Vv, Vw are generated.
When the three-phase AC voltages Vu, Vv, and Vw are applied to the AC motor 2, the energization of the AC motor 2 is controlled so as to output a torque corresponding to the torque command value trq ^* .

The dq converter 36 receives the phase current detection value from the current sensors 17 and 18 provided in the power path from the inverter 12 to the AC motor 2. In this embodiment, detected values of the V-phase current Iv and the W-phase current Iw are input from the current sensors 17 and 18 provided in the V-phase and the W-phase, and the remaining U-phase current Iu is estimated based on Kirchhoff's law. doing. In other embodiments, any two-phase current may be detected, and a three-phase current may be detected. Or you may employ | adopt the technique which estimates the other two-phase electric current based on the electric current detection value of one phase.
The dq converter 36 dq converts the phase currents Iv and Iw into dq axis currents Id and Iq based on the electrical angle θ, and outputs the result to the moving average filter 40.

  The differentiator 37 calculates the angular frequency (angular velocity) ω1 by time differentiation of the electrical angle θ. Here, ω1 represents the primary angular frequency of rotation of the three-phase current fundamental wave. In the present specification, in order to refer to a plurality of types of angular frequencies (ω) such as harmonic pulsation components, the rotational primary angular frequency is particularly distinguished as “ω1”. The angular frequency [rad / s] is synonymous with the angular velocity, but the term “angular frequency” is used in relation to the frequency characteristics. Furthermore, in the multi-pole pair AC motor, the term “rotary primary” in this specification is interpreted in the meaning of “electric primary”.

The moving average filter 40 acquires the dq-axis currents Id and Iq and the angular frequency ω1 of the primary rotation and performs a moving average process to be described later, thereby removing pulsating components and noise superimposed on the dq-axis currents Id and Iq. Suppressed and output as dq axis current filter values Idf and Iqf.
The current / torque converter 38 calculates a torque calculation value trq based on the dq-axis current filter values Idf and Iqf input from the moving average filter 40 and feeds back to the torque subtractor 32.

Next, a detailed configuration of the moving average filter 40 will be described with reference to FIG.
A moving average filter 401 having a simple configuration shown in FIG. 2A includes a section length setting unit 43 and a finite section integration unit 44.
The section length setting unit 43 sets a finite integration section length (integration period) T based on the primary rotation angular frequency ω1 of the AC motor 2. For example, when aiming at suppression of the rotational 6th order component as described later, the integration period T is set so that T = (1/6) × (2π / ω1) = π / (3ω1).
The finite interval integrating unit 44 integrates the d-axis current Id and the q-axis current Iq, which are input signals, in the integration period T, respectively. Then, an average value obtained by dividing the integral value (= current × time) by the integration period T is output as a d-axis current filter value Idf and a q-axis current filter value Iqf.

Incidentally, the d-axis current Id and the q-axis current Iq maintain substantially constant values when the operation state of the AC motor 2 is stable. Therefore, when the convergence value can be estimated for the d-axis current Id and the q-axis current Iq, it is conceivable to extract only the variation with the convergence value as a reference.
Therefore, the moving average filter 402 having the configuration illustrated in FIG. 2B includes a convergence value estimation unit 41 and adders / subtracters 421 and 422 in addition to the section length setting unit 43 and the finite section integration unit 44.

The convergence value estimation unit 41 estimates the d-axis current convergence value Id_c and the q-axis current convergence value Iq_c based on predictive control or the like. The subtractor 421 calculates current differences Id_δ and Iq_δ between the input d-axis current Id and q-axis current Iq and the convergence values Id_c and Iq_c.
The finite interval integration unit 44 calculates an average value in the integration period T for the current differences Id_δ and Iq_δ, and outputs them as filter values Idf_δ and Iqf_δ of the current difference. The adder 422 adds the current difference filter values Idf_δ and Iqf_δ to the convergence values Id_c and Iq_c, and outputs the d-axis current filter value Idf and the q-axis current filter value Iqf.

As a result, the finite interval integration unit 44 integrates only the fluctuating current differences Id_δ and Iq_δ, so that the calculation load can be reduced.
When the convergence value cannot be estimated due to a sudden change in the driving state, for example, the convergence values Id_c and Iq_c of the convergence value estimation unit 41 may be set to 0 to have the same configuration as in FIG. .

Next, filter calculation in the finite interval integration unit 44 will be described. In the present invention, the general form of “finite interval integration” is defined by equation (1). “Sig” means an arbitrary signal that is a function of the variable τ, and “s” is a filter value obtained by the processing.

The calculation content of Expression (1) is to integrate the target signal sig with respect to the finite interval T and divide the integrated value by the interval T. Therefore, it can be called a “finite interval integration filter” in comparison with the “infinite interval integration filter”. Further, this is a process in which the integration interval T moves in order as the variable τ changes, and corresponds to a so-called “moving average filter”. Hereinafter, in this specification, “finite interval integral filter” and “moving average filter” are used in the same meaning.
In the following embodiment, the variable τ is basically described as “time”. Therefore, the “section” may be interpreted as a “period” that is a section of the time axis.

Further, when the target signal sig is sine wave vibration, the “moving average filter” is expressed by Expression (2). ω is an angular frequency [rad / s].

When formula (2) is calculated, formula (3) is obtained.

In addition, Formula (4) is supplemented about the expansion | deployment of the 6th line to the 7th line of Formula (3).

式（５）にて、ωＴ＝２ｎπ（ｎは整数）のときｃｏｓωＴ＝１であるため、振幅Ｓａは０となる。したがって、角周波数ωが既知のとき、積分期間Ｔを「Ｔ＝２ｎπ／ω」に設定することで、フィルタ値ｓの振幅Ｓａを０とすることができる。 According to Expression (3), the amplitude Sa of the filter value s is expressed by Expression (5).

In equation (5), when ωT = 2nπ (n is an integer), cos ωT = 1, so the amplitude Sa is zero. Therefore, when the angular frequency ω is known, the amplitude Sa of the filter value s can be set to 0 by setting the integration period T to “T = 2nπ / ω”.

Next, an application example of the moving average filter will be described with reference to FIG.
3A shows the original signal (A = sin ωt) of the harmonic component to be suppressed. The amplitude of the original signal A is 1, and one period of the original signal A is “Ta”. With respect to the original signal A, according to the equation (3), B is 0.25 period (= Ta / 4), C is 0.5 period (= Ta / 2), and D is 0.75 period (= (3 / 4) Ta) and E indicate moving average signals calculated in an integration interval T of one period (= Ta). As calculated from Equation (3), the amplitude and phase delay of the processed signals B, C, D, and E with respect to the original signal A are as shown in Table 1.
As can be seen from FIG. 3 and Table 1, the amplitude decreases as the integration interval T becomes longer in the range up to Ta, and when the integration interval T is Ta, the amplitude becomes zero.

Further, the signal F represents an infinite time integration with respect to the original signal A as a comparison with the moving average. However, for the convenience of creation in FIG. 3, the signal F is calculated by Expression (6) with the absolute reference time being a time that is several cycles of the original signal A from the display start time t <b> 1 shown at the left end of FIG. The average (= integral value / time) of the original signal A from the absolute reference time (τ = 0) to the present (τ = t) ”is shown.

The signal F in FIG. 3 becomes 0 at the start point and end point (ωt = 0, 360 deg) of one cycle of the original signal A because the point at which the original signal A increases from 0 (ωt = 0 deg) is the absolute reference time. , Appear on the positive side in the meantime. If the point at which the original signal A decreases negatively from 0 (ωt = 180 deg) is the absolute reference time, the signal F appears from 0 to the negative side.
The signal F ′ is an enlargement of the vertical axis of the signal F. As shown by the signal F ′, the infinite integral amplitude attenuates as the time from the absolute reference time becomes longer. However, the amplitude does not become zero.

Next, referring to FIG. 4, the frequency characteristics of (a) gain and (b) phase of the moving average filter (solid line), the first-order lag filter (dashed line), and the infinite integral filter (dashed line) are compared. The phase is expressed as “the phase is delayed” and “the phase delay is increased” as it increases in the negative direction from 0 deg.
As shown in Expression (3), the gain of the moving average filter is expressed by Sa in Expression (5), and the phase delay is expressed by (−ωT / 2) using the integration period T.
The gain and phase delay of the infinite integration filter (infinite time integration filter) are as shown in Equation (6). That is, the gain of the infinite integral filter is inversely proportional to ωt, and the phase delay is uniformly −90 deg.

The gain | G | and the phase ∠G of the first-order lag filter G (s) according to the general expression (7.1) of the transfer function are expressed by the expressions (7.2) and (7.3).

The value on the horizontal axis in FIG. 4 is a value obtained by multiplying the scale value by the adjustment frequency f _adj for each filter in order to make it easy to compare the filter characteristics. Specifically, the adjustment frequency f _adj is set so that the gains of the moving average filter and the first-order lag filter in the low frequency region (ω <1 × f _adj ) are approximately equal to 1.0. For example, the adjustment frequency f _adj for the moving average filter corresponds to f _adj = 0.4π / T [rad / s].

Hereinafter, the three types of filter characteristics will be compared with reference to the numerical example of FIG. Here, the angular frequency ω will be described from the viewpoint of sequentially increasing from the low frequency side toward the high frequency side.
The gain of the moving average filter starts to decrease from 1.0 before ω = 1 × _fadj , and rapidly decreases when it exceeds the vicinity of ω = 2 × _fadj . _{Then, ω = 5 × f adj (} ωT = 2π), and, when _{ω = 10 × f adj (ωT} = 4π), the gain becomes 0.

The gain of the first-order lag filter starts to decrease from 1.0 in the same manner as the moving average filter, just before ω = 1 × _fadj . Thereafter, the gain decreasing gradient at ω = about 2 to 10 × _fadj is almost constant. For example, the gain at ω = 5 × _fadj is about 0.4, and the gain at ω = 10 × _fadj is about 0.2. While the gain of the moving average filter at the same frequency is zero, the gain of the first-order lag filter does not become zero. Also in the frequency regions before and after that, the gain of the first-order lag filter is higher than the gain of the moving average filter.
The gain of the infinite integral filter antagonizes the gain of the moving average filter in the frequency range of ω = 1 to 10 × _fadj , and is almost equal on average.
Table 2 shows the phase delay of each filter at ω = 1 to 10 × _fadj .

From FIG. 4 and Table 2, the following can be said about the prior art first-order lag filter, infinite integral filter, and moving average filter of the present invention.
(First-order lag filter)
In first-order lag filter, also referred to as "basic filter", the gain at ω = 10 × f _adj about 0.2, the phase lag is about -80Deg. That is, if the amplitude of a disturbance pulsation component having a certain frequency is to be suppressed to 1/5, the phase will be delayed by about -80 deg.

In general, it is preferable to suppress the phase delay to about 120 deg in the entire dq axis current feedback control system. However, if 80 deg is used only with a filter that suppresses pulsation, which is a disturbance, the performance of the entire control system tends to be low.
Furthermore, as the filter is strengthened to reduce the amplitude of the harmonic pulsation, the fluctuation of the dq axis currents Id and Iq due to the primary components when the operating state changes is lost. The responsiveness of is reduced.

(Infinite integration filter)
The infinite integral filter has an integral period T of the moving average filter that is infinite, and is equivalent to the integral controller (253) described in FIGS. The infinite integration filter can make the gain smaller than that of the first-order lag filter in a high frequency region of ω = about 1 to 10 × _fadj . However, the phase delay is −90 deg regardless of the frequency, and is larger than the first-order lag filter or the moving average filter, which is not preferable from the viewpoint of responsiveness.

(Moving average filter)
In the high frequency region of ω = about 2 to 10 × _fadj , the moving average filter has a smaller gain than the first-order lag filter. In particular, when ω = 5 and 10 × _fadj , the gain is zero. Further, since the gain is sufficiently small even in the vicinity of the frequency, the amplitude can be greatly reduced as compared with the first-order lag filter. In the region of ω = about 2 to 10 × _fadj , the phase delay of the moving average filter is smaller than that of the first-order lag filter.

That is, by appropriately setting the integration period T, a filter having a stronger amplitude reduction capability than the first-order lag filter is realized, and disturbances are largely removed, while the dq-axis currents Id and Iq, which are main information of control, are reduced. The phase delay can be reduced.
Note that the longer the integration period T of the moving average filter, the more affected by past information, the greater the phase delay.

Next, effective use of the moving average filter will be described with reference to FIG. In FIG. 5, the q-axis current Iq is illustrated as the dq-axis current, but the same applies to the d-axis current Id.
In general, in an AC motor, the phase voltage includes odd-order harmonic components such as fifth, seventh, eleventh, and thirteenth. Further, in the dq conversion calculation from the fixed coordinate system to the rotating coordinate system, ± 1st order is given to the harmonic component, so the dq axis currents Id and Iq include even-order pulsation components. Therefore, a filter is often used to suppress a rotational sixth-order component that is known to be particularly strong among even-order pulsation components. In this case, the frequency to be suppressed is a frequency that is six times the primary rotation frequency.

Therefore, as shown in FIG. 5A, when the angular frequency of the primary rotation is ω1 and the primary rotation period is T1 (= 2π / ω1), the integration period T is 1/6 of the primary rotation period T1. That is, the moving average is calculated by setting “T = T1 / 6 = π / (3ω1)”. As a result, one cycle of the sixth rotation of rotation is exactly within the integration period T, and the total of one cycle of the sine wave is 0. Therefore, the amplitude of the rotation sixth-order component superimposed on the q-axis current can be reduced to zero. it can.
In FIG. 5A, for the sake of easy understanding, the moving average is calculated every sine wave cycle, but in actuality, the moving average is calculated continuously.

Further, as shown in FIG. 5B, since the 12th order component, which is a 6th order multiple, fits in two periods in the integration period T, the sum in the integration period T is 0 as in the 6th order component. Become. Therefore, the 12th order component and other 6th order multiple components can be suppressed together with the 6th order component.
Regarding the relationship with FIG. 4, when the angular frequency ω is sequentially increased from the low frequency side to the high frequency side in FIG. 4A, “ω = 5 × f _adj ” where the gain becomes zero first. This point may correspond to the angular frequency of the rotational sixth-order component. At this time, the point of “ω = 10 × _fadj ” at which the gain becomes 0 for the second time corresponds to the angular frequency of the twelfth component that is twice the angular frequency of the sixth component.

As described above, by using the moving average filter in which the integration interval T is set to 1/6 of the rotation primary period T1, the dominant pulsation component due to the disturbance is set to 0 and the phase delay is smaller than that of the primary delay filter. Low frequency can be passed. Specifically, the primary rotation period T1 of the AC motor 2 is about several milliseconds to several tens of milliseconds, and the integration period T corresponding to 1/6 of the primary rotation period T1 is several hundred μs to several milliseconds. Set to order.
When the torque command or the rotational speed command is changed, the dq-axis currents Id and Iq of the primary rotation component change with the low-frequency component as the main component. Therefore, the influence on the controllability is suppressed by suppressing the pulsation due to the disturbance. Can be small.

(Second Embodiment)
Next, a second embodiment of the present invention will be described with reference to FIGS.
In the first embodiment, the moving average filter 40 is provided after the dq converter 36 in the current feedback, whereas in the motor control device 102 of the second embodiment, the front of the dq converter 36 is provided as shown in FIG. The difference is that a moving average filter 45 is provided.
The moving average filter 45 acquires a phase current, calculates a moving average in a predetermined integration period T, and outputs it as phase current filter values Ivf and Iwf.
The dq conversion unit 36 dq-converts the phase current filter values Ivf and Iwf into dq-axis currents Id and Iq based on the electrical angle θ and outputs them to the current / torque conversion unit 38.

  The feature of the moving average processing of the second embodiment is that the main object to be removed by the filter is switching noise with respect to the harmonic vibration component of the first embodiment, and in the first embodiment, for example, rotation 1 The integration period T is a period of the order of several hundreds μs to several ms corresponding to one-sixth of the next period T1, whereas the integration period T is a period of the order of several μs to several tens μs. is there.

Switching noise will be described with reference to FIGS.
As is well known, the inverter 12 is configured by bridge-connecting switching elements such as IGBTs (insulated gate bipolar transistors), and the switching elements of the upper arm and the lower arm are complementarily turned on and off according to a command signal. Generate output voltage.

As shown in FIG. 7, when the switching element is turned on / off (“SW-ON / OFF” in the figure), the current flowing through the switching element changes abruptly, so that a surge voltage (−LdI) generated according to the inductance L of the circuit. / Dt) generates switching noise.
Conventionally, if the sampling period is relatively long in the configuration in which the current sensor output is acquired and AD (analog-digital) conversion is performed, it is sufficient to acquire a stable current value after noise attenuation. However, in recent years, if sampling at the order of several hundred kHz becomes possible by increasing the speed of AD conversion, the possibility of sampling switching noise before attenuation increases.

FIG. 8A is a diagram showing exaggerated switching noise of the VIIIb portion for the sampled phase current. Actually, switching noise exists over the entire period of the fundamental wave. Further, the relationship between the switching noise and the sampling timing is expressed as shown in FIG. 8B, for example.
Therefore, as shown in FIGS. 7 and 8B, a period corresponding to one period of the switching noise, for example, “a period of four sampling timings” is set as the integration period T in the case of FIG. . Here, one cycle of the switching noise is known information determined by the hardware specification of the inverter 12. The frequency of the switching noise is on the order of several MHz (period: several μs), and it is preferable that the half period is a sampling period.

  And by calculating the moving average with the moving average filter 45 of 2nd Embodiment, the increase / decrease in the phase current by switching noise can be canceled and switching noise can be removed. As a result, as shown in FIG. 8C, the phase current filter values Ivf and Iwf from which the switching noise has been removed are input to the dq converter 36.

  The rotation primary frequency of the phase current is about 2000 Hz at maximum, that is, the rotation primary period T1 is about 500 μs at the shortest, and is sufficiently low for switching noise of the order of several μs. Therefore, switching noise can be suitably removed using the moving average filter 45 without affecting the fundamental wave itself. It has been confirmed that the values of the dq-axis currents Id and Iq obtained by performing the dq conversion on the phase current filter values Ivf and Iqf after the moving average filter 45 are close to the results obtained by removing the switching noise after the dq conversion.

  As described above, the moving average filter according to the second embodiment can suitably remove switching noise on the order of several μs by setting the sampling period to the integration period T. Further, similarly to the first embodiment, the phase delay can be reduced as compared with the first-order lag filter or the like while reducing the amplitude of the signal to be removed.

(Other embodiments)
(A) In the first embodiment, the moving average filter 40 is provided for each of the d-axis current Id and the q-axis current Iq. However, the moving average filter 40 is provided only for one of the d-axis current Id and the q-axis current Iq. You may make it provide. For example, the harmonic fluctuation component is suppressed using the moving average filter 40 only for the q-axis current that affects the torque ripple, and a general first-order lag filter is used for the d-axis current used for field weakening in the high rotation region. You may make it use. Also, the convergence value estimation unit 41 may be applied to only one of the d-axis current Id and the q-axis current Iq.
Moreover, you may provide the moving average filter 45 about the phase current detected value of any one phase in two phases with respect to the said 2nd Embodiment. Or in the structure which detects the electric current of three phases, you may provide the moving average filter 45 about the phase current detection value of any one phase or two phases among them.

(A) The integration period T in the first embodiment is not limited to an example in which the integration period T is set to 1/6 of the primary rotation period. When the purpose of removing n-order harmonic components other than the 6th-order as the rotation-synchronized harmonic components generated by the inverter control is to set the section length of “1 / n of the rotation primary period” to the integration period T It is preferable to set to.
(C) The moving average filter of the present invention may be applied to a signal sig other than sinusoidal vibration as shown in the general formula (1). Further, the variable τ in integration is not limited to time.

(D) The configuration of the control block of the motor control devices 101 and 102 is not limited to that illustrated in FIGS. For example, instead of the configuration in which the torque trq is fed back with respect to the torque command value trq ^* , the dq-axis currents Id and Iq may be fed back with respect to the dq-axis current command value.

  (E) The AC motor may be a synchronous motor other than the permanent magnet type synchronous motor. The motor generator is not limited to a motor generator having both a function as an electric motor and a function as a generator, and may not have a function as a generator. Furthermore, the present invention is not limited to a three-phase AC motor, and can be widely applied to a multi-phase AC motor having three or more phases.

  (F) The control device for an AC motor according to the present invention is not limited to a system in which only one set of an inverter and an AC motor is provided as in the above-described embodiment, but may be applied to a system in which two or more sets of inverters and AC motors are provided. Good. Further, the present invention may be applied to a system such as a train in which a plurality of AC motors are connected in parallel to one inverter.

(G) The control device for an AC motor according to the present invention is not limited to an AC motor of a hybrid vehicle, and may be applied to an AC motor of an electric vehicle having any configuration. Moreover, you may apply to AC motors other than an electric vehicle.
As mentioned above, this invention is not limited to the said embodiment at all, In the range which does not deviate from the meaning of invention, it can implement with a various form.

101, 102 ... Electric motor control device (AC motor control device),
2 ... AC motor,
40, 45 ... Moving average filter.

Claims (6)

A phase current flowing in a multiphase AC motor (2) of three or more phases is detected and converted into a d-axis current and a q-axis current, and the AC motor is energized by feedback control using the d-axis current and the q-axis current. A control device (101, 102) of an AC motor to be controlled,
An AC motor comprising a moving average filter (40, 45) for calculating and outputting a moving average in a finite predetermined section for at least one of a d-axis current and a q-axis current, or at least one phase current. Control device. The moving average filter (40)
The AC motor control device (101) according to claim 1, wherein a moving average is calculated so as to suppress pulsation due to a harmonic component of a specific order for at least one of the d-axis current and the q-axis current. The moving average filter is:
The control apparatus for an AC motor according to claim 2, wherein a section length of the predetermined section for calculating a moving average is set according to a rotation speed of the AC motor. The moving average filter is:
4. The control apparatus for an AC motor according to claim 3, wherein the section length is set to a length of 1/6 of a primary rotation period of the AC motor. The moving average filter has a convergence value estimation unit (41) that estimates a convergence value of a d-axis current or a q-axis current,
The AC motor according to any one of claims 1 to 4, wherein a moving average in the predetermined section is calculated for a difference between the input d-axis current or q-axis current and the convergence value. Control device. The moving average filter (45)
The control device for an AC motor according to claim 1, wherein the moving average is calculated so as to remove switching noise of the inverter (12) that supplies power to the AC motor for at least one phase current. 102).

Images