JP3244744B2 - Induction motor vector control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an induction motor vector control apparatus for performing variable speed drive control of an induction motor by vector control.

[0002]

2. Description of the Related Art Conventionally, induction motors have often been used as constant speed motors. However, with the development of power semiconductor devices, power converters capable of outputting a variable voltage and variable frequency have emerged and have been used as variable speed motors. It became so.

[0003] As a variable speed driving system for an induction motor, a vector control system is often employed. The vector control method is a method designed to treat an induction motor in the same manner as a DC motor, and separates the secondary side into a torque axis component and a magnetic flux axis component and controls each axis component.

In the vector control method, the first method is a magnetic field orientation type that uses a secondary magnetic flux as a vector quantity to control a primary current, and the second method is that a magnetic flux vector is calculated based on induction motor parameters. There is a slip frequency form to control.

FIG. 13 is a block diagram showing an example of a conventional slip frequency type vector control device, which is provided with a vector control calculator described below when operating the induction motor 2. That is, the power converter 1, the angle detector (rotational speed detector) 3, and the current detectors 4a, 4b, 4c
, Vector rotators 6a, 6b, integrator 8, proportional integrators 9a, 9b, 9c, and adder / subtracters 10a, 10b, 1
0c, an adder 11, a magnetic flux command generator 13, a magnetizing current calculator 14, and a slip frequency calculator 15.

In this case, a proportional integrator (speed control system) 9c
Is treated as a torque current command iqs ^* . The magnetic flux command generator 13 receives the current speed ωr detected by the angle detector 3 and calculates a secondary magnetic flux command λr ^* using this as a function. The magnetizing current calculator 14 receives the magnetic flux command λr ^* and calculates a magnetizing current command (magnetic flux current command) ids ^* .
Slip frequency calculator 15 inputs the torque current command iqs ^* and flux command .lambda.r ^*, calculates the slip frequency command .omega.s ^*. The integrator 8 calculates the slip frequency command ωs ^* and the current speed ω
The sum of r is integrated to calculate the rotation angle θ.

On the other hand, the vector rotator 6b converts the primary currents iu, iv, iw detected by the current detectors 4a to 4c and the rotation angle θ into a two-phase rotating magnetic flux coordinate, and outputs the magnetization current ids and torque. Output the current iqs. These ids and iqs are respectively applied to the instructions ids ^* and iqs ^{* by the} adders / subtractors 10a and 10b ^.
And the difference is amplified by the proportional integrators 9a and 9b.
The signals are converted into three phases by the vector rotator 6a and the voltage commands eu and ev
, Ew. The voltage commands eu to ew are input to the power converter 1, and the induction motor 2 is operated based on the output voltages vu, vv, and vw.

The magnetizing current calculator 14 calculates a magnetizing current command ids ^* based on equation (1).

Ids ^* = (1 / M) × λr ^* + (Lr / MRr) × dλr ^* / dt (1) The slip frequency calculator 15 calculates the slip frequency command ωs ^* based on the equation (2). Calculate.

Ωs ^* = (MRr / Lr) × (iqs ^* / λr ^* ) (2) In the equations (1) and (2), M is an exciting inductance (mutual inductance of the induction motor 2), and Lr is an induction. The mutual inductance of the electric motor 2 is shown, and Rr is a secondary resistance.

[0011]

In the conventional slip frequency type vector control device shown in FIG. 13, since the secondary resistance Rr changes according to the temperature change of the secondary winding, the slip frequency command ωs of the equation (2) is obtained. ^An error occurs in the calculation of ^* . Due to this error, the predetermined torque required by the motor load cannot be obtained. Further, minute fluctuations in the excitation inductance M and the secondary inductance lead to vibration of the magnetic flux.

Accordingly, an object of the present invention is to provide a vector control device for an induction motor that can obtain a predetermined torque without being insensitive to a variation in the mutual inductance M or a variation in the secondary resistance Rr due to temperature or the like. .

[0013]

Means for Solving the Problems The present invention has the following construction to achieve the above object. That is, the invention corresponding to claim 1 drives the induction motor.
Parameter detection means for detecting an input voltage and an output current of the power converter, and an input voltage and an output power of the power converter.
Calculating means for calculating a secondary flux calculation value and the torque current calculation value of the induction motor on the basis of the flow, the induction motor
The secondary magnetic flux command value and the torque current command value,
Magnetic flux calculation value and torque current calculation calculated by calculation means
The secondary magnetic flux command value and the torque current
Or a set of command values or the calculated secondary magnetic flux and torque current
Select the set of calculated values and select the secondary magnetic flux command value or the
Is obtained by multiplying the next magnetic flux calculation value by the first magnetizing current weight coefficient.
And outputs the secondary magnetic flux command value.
A second magnetizing current weighting factor
The torque current command value or the torque
Multiply the calculated current value by the first slip frequency weighting factor
Output the slip frequency command value obtained by multiplying
Of the secondary magnetic flux command value and the torque current command value.
The magnetic flux current command value and the slip when a set is selected
Frequency command value, secondary magnetic flux calculation value and torque current calculation
The magnetic flux current command value and the
The first is set so that the difference from the slip frequency command value becomes zero.
Current weighting factor and the first slip frequency weighting factor
Neural network means for changing the actual magnetic flux current value and the actual torque current value of the induction motor based on the slip frequency command value output from the neural network means and the output current of the power converter. Detecting and controlling the induction motor in accordance with a deviation between the detected actual magnetic flux current value and the magnetic flux current command value, and a deviation between the detected actual torque current value and the detected torque current command value. This is a vector control device for an induction motor, comprising a vector control means.

According to a second aspect of the present invention, an induction motor is provided.
Parameter detecting means for detecting an input voltage and an output current of a power converter to be driven; and an input voltage of the power converter.
Said calculating means for calculating a secondary flux calculation value and the torque current calculation value of the induction motor, the induction based on the output current
Secondary magnetic flux command value and torque current command value for motor
And the calculated secondary magnetic flux value and torque calculated by the calculating means.
The calculated torque current value and the actual torque current value
A set of the secondary magnetic flux command value and the actual torque current value
Or a set of the secondary magnetic flux calculation value and the torque current calculation value
Or the secondary magnetic flux command value and the torque current command value
Select a set and select the secondary magnetic flux command value or the secondary magnetic flux
A value obtained by multiplying the calculated value by a first magnetizing current weight coefficient and
To the secondary magnetic flux command value or the derivative of the secondary magnetic flux calculation value.
And a value obtained by multiplying the value obtained by multiplying the magnetizing current weight coefficient by 2
And outputs the actual magnetic flux current command value.
Or the torque current calculation value or the torque current finger
And a value obtained by multiplying the threshold value by the first slip frequency weighting coefficient.
Slip frequency command obtained by dividing the secondary magnetic flux command value
Outputs the secondary magnetic flux command value and the actual torque.
The magnetic flux current command value when a set of lux currents is selected and
The slip frequency command value, the secondary magnetic flux calculation value, and the torque
The magnetic flux current when a set of calculated current values is selected.
The difference between the flow command value and the slip frequency command value becomes zero.
The first magnetizing current weight coefficient and the second magnetizing current
The flow weight coefficient and the first slip frequency weight coefficient are changed.
A neural network means for, detecting the actual actual torque current value and the flux current value of the induction motor on the basis of said slip frequency command value output from the neural network means and the output current of the power converter A vector control for controlling the induction motor in accordance with a deviation between the detected actual magnetic flux current value and the magnetic flux current command value, and a deviation between the detected actual torque current value and the torque current command value. And a vector control device for the induction motor.

According to a third aspect of the present invention, an induction motor is provided.
Parameter detecting means for detecting an input voltage and an output current of a power converter to be driven; and an input voltage of the power converter.
Said calculating means for calculating a secondary flux calculation value and the torque current calculation value of the induction motor, the induction based on the output current
Secondary magnetic flux command value and torque current command value for motor
And the calculated secondary magnetic flux value and torque calculated by the calculating means.
And the secondary magnetic flux command value and the torque
The set of torque current command values or the secondary magnetic flux calculated value and torque
Select a set of calculated current values and set the secondary magnetic flux command value
Or the second magnetic flux calculation value is multiplied by a first magnetizing current weighting coefficient.
Calculated value and the secondary magnetic flux command value or the secondary magnetic flux
A value obtained by multiplying a differential value of the calculated value by a second magnetizing current weighting coefficient;
And outputs a magnetic flux current command value obtained by adding
The first current value or the torque current command value.
The value obtained by multiplying the slip frequency weight coefficient and the secondary magnetic flux command value
And output the slip frequency command value obtained by dividing
In both cases, the set of the secondary magnetic flux command value and the torque current command value is
The magnetic flux current command value and the slip frequency when selected
Number command value and the secondary magnetic flux calculation value and the torque current calculation value.
The magnetic flux current command value and the slip when a set is selected
The first magnetic field is set so that the difference from the frequency command value becomes zero.
Current weighting factor, the second magnetizing current weighting factor and the second
Neural network means for changing the slip frequency weighting coefficient of 1, the slip frequency command value output from the neural network means, and the power converter
Based on the output current of the induction motor, the actual magnetic flux current value and the actual torque current value of the induction motor are detected, the deviation between the detected actual magnetic flux current value and the magnetic flux current command value, and the detected And a vector control means for controlling the induction motor in accordance with a deviation between the actual torque current value and the torque current command value.

[0016]

According to the first aspect of the present invention, even if an appropriate internal parameter is given and the operation state is established, the motor-generated torque and the secondary magnetic flux are corrected so that no error occurs with respect to the command. You.

According to the second aspect of the invention, not only the same operation and effect as those of the first aspect of the invention can be obtained, but also the torque current detection value is input, so that the torque can be reduced as in the transient response. Even if a delay occurs between the current command value and the feedback value, it is possible to cope with the delay.

According to the third aspect of the present invention, even if an appropriate internal parameter is given and the operation state is set, the motor-generated torque and the secondary magnetic flux are corrected so that no error occurs with respect to the command. You.

[0019]

FIG. 1 is a block diagram showing a first embodiment of the present invention. The same reference numerals in the prior art of FIG. 13 denote the same circuit elements or the same signals. The vector control device for an induction motor according to the present embodiment is based on a conventional slip frequency type vector control device and mainly includes a magnetic flux detector 7 and a neural network means 12.

The magnetic flux detector 7 includes voltage detectors 5a, 5b,
The input voltage of the power converter 1 detected by 5c and the output current of the power converter 1 detected by the current detectors 4a, 4b and 4c, that is, the input current of the induction motor 2, are input, and the secondary magnetic flux calculation value cλr And the torque current calculation value ciqs is estimated.

FIG. 2 shows the neural network means 12 of FIG.
a, 121b and output side switches 121c, 121d
And a neural network circuit composed of coefficient units 122, 123, and 124, and a backpropagation circuit 12
7

The switch 121a has a secondary magnetic flux command value λr
^* And the secondary magnetic flux calculation value cλr from the magnetic flux detector 7 in FIG. 1 are selectively switched and input to the coefficient units 122 and 123, and the switch 121b is connected to the torque current command output from the proportional integrator 9c in FIG. The value iqs ^* and the torque current calculation value ciqs output from the magnetic flux detector 7 are selectively used as a coefficient unit 124.
Enter

The input signal X from the switch 121a
The reciprocal calculator 125 which receives the secondary magnetic flux command value λr ^* of 1 and outputs a calculated value X3 obtained by calculating the reciprocal, and the coefficient which receives the input signal X1 from the switch 121a and multiplies the magnetized current weighting coefficient W1. Unit 122, a coefficient unit 123 for inputting the operation value X3 from the reciprocal operation unit 125 and multiplying by the magnetizing current weighting coefficient W3, and an input signal X
5 and a coefficient multiplier 124 for multiplying the slip frequency weighting coefficient W5 by inputting a value of "5".

The output of the coefficient unit 122 is the output side switch 1
The output of the coefficient unit 124 is output via the switch 121d, and is output from the adder / subtractor 10a and the back propagation circuit 127 shown in FIG.
6 is input.

The back propagation circuit 127 uses signals U1 and U2 when the switches 121c and 121d are switched to the "A" side, and signals V1 and U2 when the switches 121c and 121d are switched to the "B" side. Capture as V2.

The multiplier 126 receives the output of the reciprocal calculator 125, that is, the reciprocal of the secondary magnetic flux command value λr *.
When the switch 121d is switched to the "A" side, a value obtained by multiplying X5 by a value obtained by multiplying by the slip frequency weighting coefficient is output as the slip frequency command value ωs ^* .

When the switches 121c and 121d are switched to the "A" side, the outputs U1 and U2 of the coefficient units 122 and 124 become the back propagation circuit 1 respectively.
At the same time as being input to the adder 27, it is output to the adder / subtractor 10a in FIG. 1 as the magnetic flux current command value ids ^* which is the output U1.

When the output-side switches 121c and 121d are switched to the "B" side, the outputs V1 and V2 of the coefficient units 122 and 124 become the back propagation 127, respectively.
Is input to

In the embodiment configured as described above, the following control is performed. First, at the first timing, the switches 121a to 121d are connected to the "A" side, and the signals X1, X3, and X5 are input to the coefficient units 122, 123, and 124. Then, the coefficient unit 122 outputs the product of X1 and the magnetizing current weight coefficient W1, that is, ids ^* and U1, and
The product of X3, which is the reciprocal of 1, and the magnetizing current weight coefficient W3 is output. The product of X5 and the slip frequency weighting coefficient W5 is input to the multiplier 126 and also to the back propagation 127 as U2.

The magnetizing current command value ids ^* output from the neural network means 12 at the first timing ^.
And the slip motor 8 using the slip frequency command value ωs ^*.
Is controlled.

Next, at the second timing, the switch 121
a to 121d are connected to the “B” side, and the calculation values cλr and ciqs from the magnetic flux calculator 7 are used as inputs X1, X3 and X5, and the command values V1 and V2 are calculated. Learning of Wl, W3, and W5 is performed using the following equation. The evaluation function J is set as shown in equation (3), and J = 1/2 × {(V1−U1) 2+ (V2−U2) 2} (3) Assuming that the learning gains are Kd1 and Kd2, The change amount ΔWi (i = 1, 5) is expressed as in equations (4) and (5). ΔW1 = (LLJ / LLW1) · Kd1 = (U1-V1) · X1 · Kd1 (4) ΔW5 = (LLJ / LLW5) · Kd2 = (U2-V2) · X2 · Kd3 (5) In equations (4) and (5), LL represents Laplacian.

Then, the new internal parameters are (6),
As shown in the equation (7), W1 = W1 + ΔW1 (6) W5 = W5 + ΔW5 (7) is learned.

Here, the magnetic flux current command value ids ^* and the slip frequency command value ωs ^* are calculated using the new internal parameters.

Ids ^* = W1 · X1 = W1 · dλr ^* / dt (8) ωs ^* = (W5 · X5) · (1 / λr ^* ) (9) In particular, calculation of the slip frequency command value ωs ^* In this case, the result is the output of the neural network means 12 (W
Since the result of the product of (3.X3) and (W5.X5) is not used as data of the neural network means 12 having learning ability, the product is multiplied by the multiplier 126 outside the neural network means 12.

As described above, the magnetic flux current command value ids ^* and the slip frequency command value ωs ^* output from the neural network means 12 are input to the vector control unit, and the vector control of the induction motor 2 is performed.

The voltage and current during operation in this state are input to the magnetic flux detector 7 to calculate the secondary magnetic flux calculation value cλr and the torque current calculation value ciqs of the induction motor 2 as described above. Also, the back propagation circuit 127
At this time, the magnetic flux current command value ids ^* is stored as U1, and the slip frequency command value ωs ^* is stored as U2.

Next, at the second timing, the switch 121
a to 121d are connected to the “B” side, and the secondary magnetic flux calculation value cλ
The r and the calculated torque current value ciqs are input to the neural network means 12. Then, the neural network means 12 obtains the magnetic flux current command value ids * 'and the slip frequency command value ωs ^* based on the secondary magnetic flux calculation value cλr and the torque current calculation value ciqs. The back propagation circuit 127 stores ids ^* 'at this time as V1 and ωs ^* as V2.

The back propagation circuit 12
7 is the command value ids ^* and ωs ^* and ids obtained earlier.
^{* And} ωs ^* according to their respective differences δ.
Learning is repeated so that the load is changed to 24 and finally the difference becomes zero. Thus, finally, λr ^* = cλr, iqs ^* = ciqs, ids ^* = id
s ^* , ωs * = ωs ^*, and calm down. At this time, assuming that the secondary magnetic flux calculation value cλr and the torque current calculation value ciqs obtained by the magnetic flux detector 7 are sufficiently accurate values, the magnetic flux current command value ids ^* and the slip frequency command obtained by the neural network means 12 are obtained. The value ωs ^* can be expected to be sufficiently accurate. As a result, a vector control induction motor having the same output characteristics as a DC motor can be obtained.

During operation, in this embodiment, even if the secondary resistance changes due to an increase in the temperature of the rotor, the weighting coefficients of the respective coefficient units of the neural network means 12 are updated accordingly, Vector control of the induction motor 2 is always possible in an optimal state. The same applies when the core of the induction motor 2 is saturated and the mutual inductance M is changed.

In this way, it is possible to provide a vector control device for an induction motor that can obtain a predetermined torque without being insensitive to the variation of the mutual inductance M and the variation of the secondary resistance Rr due to the temperature and the like.

Next, a second embodiment of the present invention will be described with reference to FIGS. FIG. 3 is a block diagram showing a schematic configuration of the second embodiment, and FIG. 4 is a diagram for explaining the neural network means 12 of FIG. Here, different points from the above-described first embodiment and the conventional device of FIG. 13 will be mainly described. In each drawing, the same reference numerals indicate the same circuit elements or the same signals.

That is, the neural network means 1
2 is an input-side switch 131a, 131b, an output-side switch 131c, 131d, 131e,
2, 134, 135, a differentiator or differentiator 135, a back propagation circuit 138, a divider 136,
And an adder 137.

The switch 131a has a secondary magnetic flux command value λr
^* And the secondary magnetic flux calculation value cλr from the magnetic flux detector 7 are selectively switched and input to the neural network means 12. The switch 131ba selects a torque current command value iqs ^* output from the proportional integrator 9a, a torque current calculation value ciqs output from the magnetic flux detector 7, and a torque current detection value iqs output from the vector rotator 6b. Input to the neural network means 12.

The neural network means 12 adds the magnetizing current weight coefficient W to the input signal X1 from the switch 131a.
A coefficient multiplier 132 for multiplying by 1 is provided. Similarly, a coefficient unit 134 is provided which multiplies the input signal X2 obtained from the switch 131a via a differentiator or differentiator 133 by a magnetizing current weighting coefficient W2. Further, a coefficient unit 135 for multiplying the input signal X5 from the switch 131b by the slip frequency weighting coefficient W5 is provided.

The switch 131c selectively switches the transmission destination of the signal output from the coefficient unit 132. Switch 1
31d selectively switches the transmission destination of the signal output from the coefficient unit 134. The switch 131e includes a coefficient unit 135
The destination of the signal output from is selectively switched.

The divider 136 outputs the output of the coefficient unit 135 input via the switch 131e to the secondary magnetic flux command value λ.
The value is divided by r ^* , and the value is output as the slip frequency command value ωs ^* . The adder 137 includes a coefficient unit 132 and a coefficient unit 13
4 and add the sum to the magnetic flux current command value ids ^*.
Is output.

The back propagation circuit 138 stores the signals U1, U2, U5 input from the switches 131c to 131e at the first timing, and stores the signals V1, U1 input from the switches 131c to 131e at the second timing.
V2 and V5 are stored. Then, the load of the neural network means 12 is corrected from the difference between (U1, U2, U5) and (V1, V2, V5).

The learning operation of the processing unit of the neural network means 12 configured as described above will be described below. First, at the first timing, the switches 131a to 131a to
131e is connected to the “A” side, and the secondary magnetic flux command value λr ^*
And the torque current detection value iqs are determined by the switches 131a, 13
The signal is input to the neural network means 12 via 1b.

The input signals X1, X of the coefficient units 132, 135
5 is multiplied by the respective weighting factors W1 and W5, and is output.
Switches 131c and 131e are provided with coefficient units 132 and 135, respectively.
Is output to the back propagation circuit 138 by U1,
Input as U5. On the other hand, the secondary magnetic flux command value λr ^* is differentiated by a differentiator 133 via a switch 131a, and the differentiated value is multiplied by a weight coefficient W2 and output.
U to the back-propagation circuit 138 via 31d
Enter as 2.

That is, the back propagation circuit 1
In 38, the product of X1 and the weighting factor W1 is stored as U1, the product of X2 which is a differential value of X1 and the load W2 is stored as U2, and the product of X3 and the load W5 is stored as U5.

On the other hand, the induction motor 2 is controlled using the magnetic flux current command value ids ^* output from the adder 137 and the slip frequency command value ωs ^* output from the divider 136. The input current and the input voltage input to the induction motor 2 controlled by the respective command values generated at the first timing are detected, and the magnetic flux detector 7 calculates the secondary magnetic flux calculation value c.
λr and the calculated torque current value ciqs are calculated by the method described above. The secondary magnetic flux calculation value cλr and the torque current calculation value ciqs are given to the switches 131a and 131b.

Next, at the second timing, the switch 13
1a to 131e are connected to the “B” side, and the secondary magnetic flux calculation value cλr and the torque current calculation value ciqs are input to the neural network means 12. The neural network means 12 outputs the signals output in response to the input of the operation value to the back propagation circuit 138 by V1 and V1.
2. Imported as V5.

The back-propagation circuit 138 sets the evaluation function J based on the following equations (10) to (18), determines the correction amount Δw of each load, and corrects each load.

The evaluation function J is set as shown in equation (10), and J = 1/2 × {(V1-U1) 2+ (V2-U2) 2+ (V5-U5) 2} (10) The learning gain is Kd1. , Kd2, and Kd5, the load correction amount (change amount of the internal parameter) ΔWi (i = 1, 2, 2,
5) is expressed as in equations (11) and (12). ΔW1 = (LLJ / LLW1) · Kd1 = (V1−U1) X1 · Kd1 (11) ΔW5 = (LLJ / LLW5) · Kd5 = ( V5−U5) X5 · Kd5 (12) In equations (11) and (12), LL represents Laplacian.

Then, the new internal parameter is (13)
(15) W1 = W1 + ΔW1 (13) W5 = W5 + ΔW5 (14) W2 = 1 / W5 (15)

Next, at the third timing, the switch 13
1a to 131e are connected to the “C” side, and the switch 131
a, a secondary magnetic flux command value λr ^* is input, and the switch 1
A torque current command value iqs ^* is input from 31b.
The neural network means 12 calculates the magnetic flux current command value ids ^* and the slip frequency command value ωs ^* using the new load W corrected for these inputs as described above. And ids ^* , ω expressed by equations (16) and (17).
Output s ^* .

Ids ^* = (W1 · X1) + (W2 · X2) (16) ωs ^* = (W5 · X5) / λr ^* (17) Output from the coefficient unit 135 at the third timing. The signal is not used for learning, but is divided by the divider 136 with the secondary magnetic flux command value λr ^*, and the divided value is output as the slip frequency command value ωs ^* .

The values thus obtained at the first and second timings, ie, U1, U2, U5 and V1,
Learning is performed so that the deviation between V2 and V5 becomes zero, and the magnetic flux current command value ids ^* and the slip frequency command value ωs ^* are output at the third timing.

FIG. 5 shows various command values, calculated values, and changes in loads W1, W2, and W5 when the induction motor 2 is driven and controlled by the above-described second vector control device of the present invention. ing. As shown in FIG. 5, in the present embodiment, when the drive control of the induction motor 2 is performed by giving appropriate internal parameters, the weight coefficient (load) W of the neural network means is automatically learned to an optimum value, and the induction is performed. The generated torque τ of the electric motor 2 and the secondary magnetic flux calculation value cλr are corrected so that no error occurs with respect to the secondary magnetic flux command value λr ^* .
In this case, since the torque current detection value iqs is input to the neural network means separately from the torque current calculation value ciqs, the torque current command value i
Even if there is a delay between qs ^* and the feedback value, it is possible to respond.

FIG. 6 shows a modification of FIG.
As shown in (a), the back propagation circuit 138
The same effect can be obtained even when the input of the input is set to the slip frequency. Also, as shown in FIG. 6B, the sum of a term proportional to the magnetic flux and a term proportional to the change in the magnetic flux can be used as the input to the back propagation circuit 138.

FIG. 7 shows a third embodiment of the present invention.
FIG. 8 is a partial detailed view of the neural network means 12 in FIG. 1 denote the same circuit elements or the same signals. The neural network unit 12 outputs the switches 121a to 121d, a coefficient unit 132 that outputs a first magnetizing current weight coefficient W1, a coefficient unit 133 that outputs a second magnetizing current weight coefficient W2, and a slip frequency weight coefficient W5. A coefficient unit 135 and a divider 136
, A differentiator or differentiator 146, and a back propagation circuit 138.

In FIG. 7, when the speed of the induction motor 2 is controlled, the speed command value ωr * is compared with the speed ωr.
A torque current command value iqs ^* is output based on a certain control theory. The secondary magnetic flux command value λr ^* is calculated from the actual speed and the magnetic flux command generator 13. On the other hand, the magnetic flux detector 7 estimates and calculates the actual secondary magnetic flux cλr and the torque current ciqs from the current i, the voltage v, or an equivalent signal applied to the induction motor 2.

The neural network means 12, which receives the secondary magnetic flux, the differential value of the secondary magnetic flux, the command value of the torque current, and the actual value, inputs the secondary magnetic flux command value λr ^* while learning the internal parameters. The slip frequency command value ωs ^*, which is obtained by dividing the value calculated based on the magnetic flux current command value ids ^* and the torque current command value iqs ^* , which are the output values of the above, by the secondary magnetic flux command value, is output. Therefore, the above output value, magnetic flux current command value ids ^* , torque current command value i
qs ^* and the slip frequency command value ωs ^* , the induction motor 2
Is optimally controlled.

In the neural network means 12, the following control is performed. First, at a certain timing, the switches 141a to 141d are connected to the A side, and inputs are X1, X2, and X5 based on the command value. X1 and weight function W
Let the output U1 be the sum of the product of X1 and the product of X2, which is the derivative of X1, and the weighting function W2, and let U2 be the product of X5 and the weighting function W5. At the next timing, switches 121a to 121a
1d is connected to the “B” side, and the input is X1 based on the measured value.
, X2, and X5, and V1 and V2 are calculated in the same manner as the command values. Learning of W1, W2, and W5 is performed using the following equation. The evaluation function J is set as in the equation (18), and J = 1/2 × {(U1-V1) 2+ (U2-V2) 2} (18) Assuming that the learning gains are Kd1 and Kd5, the internal parameters The change amount ΔWi (i = 1, 2, 5) is (19), (2
ＷW1 = (LL / LLW1) · Kd1 = (U1-V1) · X1 · Kd1 (19) ΔW5 = (LLJ / LLW5) · Kd5 = (U2-V2) · X2 · Kd5 (20) In equations (19) and (20), LL represents Laplacian. The new internal parameter is (2
W1 = W1 + ΔW1 (21) W5 = W5 + ΔW5 (22) W2 = 1 / W5 (23) As shown in equations (1) to (23), learning is performed.

Here, the magnetic flux current command value ids ^* and the slip frequency command value ωs ^* are calculated using the new internal parameters.

Ids ^* = (W1 · X1) + (W2 · X2) (24) ωs ^* = (W5 · X5) / λr ^* (25) The second embodiment of the vector control device for an induction motor of the present invention described above. According to the third embodiment, the weighting function W can be maintained even when the vehicle enters the operating state with appropriate internal parameters as shown in FIG.
Is automatically learned to an optimum value, and the generated torque τ of the motor and the secondary magnetic flux calculation value cλr are corrected so as not to cause an error with respect to the secondary magnetic flux command value λr ^* .

Instead of the neural network means 12 shown in FIG. 8, the following control can be performed by configuring as shown in FIG. First, at some timing, switch 1
21a to 121d are connected to the A side, and inputs are X1, X2, and X5 based on command values. The products of the respective inputs and the weighting factors are U1, U2, and U5. At the next timing, the switches 121a to 121d are connected to the "B" side, and based on the measured values, the inputs are X1, X2, and X5, Similarly, V1, V2, and V5 are calculated. Learning of W1, W2, and W5 is performed using the following equation. The evaluation function J is set as in the equation (25), and J = 1/2 × {(U1-V1) 2+ (U2-V2) 2+ (U5-V5) 2} (26) The learning gain is Kd1. , Kd2, and Kd5, the change amount ΔWi (i = 1, 2, 5) of the internal parameter is (2
ＷW1 = (LLJ / LLW1) · Kd1 = (U1-V1) · X1 · Kd1 (27) ΔW2 = (LLJ / LLW2) · Kd2 = (U2-V2) ) · X2 · Kd2 (28) ΔW5 = (LLJ / LLW5) · Kd5 = (U5-V5) · X1 · Kd5 (29) Expressions (27), (28), and (29) represent Laplacian. The new internal parameter is (3
W1 = W1 + ΔW1 (30) W2 = W2 + ΔW2 (31) W5 = W5 + ΔW5 (32) As shown in equations (0) to (32). Here, using the new internal parameters, the magnetic flux current command value ids ^* and the slip frequency command value ωs ^*
Is calculated.

Ids ^* = (W1 · X1) + (W2 · X2) (33) ωs ^* = (W5 · X5) / λr ^* (34) The slip frequency is calculated as shown in FIGS. 11 and 12. In this case, a similar result can be obtained even if the division of the magnetic flux is performed inside the neural network means 12. 11 and FIG. 12 are the same as FIG.

[0069]

According to the vector control apparatus for an induction motor of the present invention described above, even when the internal motor is operated under appropriate internal parameters, the torque generated by the induction motor and the secondary magnetic flux can be reduced to the command value. It is corrected so that no error occurs.

[Brief description of the drawings]

FIG. 1 shows a first embodiment of a vector control device for an induction motor according to the present invention.
FIG. 3 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a detailed view of the neural network means in FIG.

FIG. 3 shows a second embodiment of the vector control device for an induction motor according to the present invention.
FIG. 3 is a block diagram showing an embodiment of the present invention.

FIG. 4 is a detailed view of the neural network means in FIG. 3;

FIG. 5 is a view showing a simulation result of the embodiment of FIG. 3;

FIG. 6 is a diagram showing a modification of the embodiment of FIG. 3;

FIG. 7 is a block diagram showing a third embodiment of the control apparatus for the induction motor of the present invention.

FIG. 8 is a detailed view of the neural network means in FIG. 7;

FIG. 9 is a diagram showing a simulation result in the control method of FIGS. 7 and 8;

FIG. 10 is a diagram showing a modification of FIG. 7;

FIG. 11 is a view showing a modification of FIG. 7;

FIG. 12 is a view showing a modification of FIG. 7;

FIG. 13 is a block diagram showing an example of a conventional vector control device for an induction motor.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Power converter, 2 ... Induction motor, 3 ... Angle detector, 4
a, 4b, 4c: current detector, 5a, 5b, 5c: voltage detector, 6a, 6b: vector rotator, 7: magnetic flux detector, 8: integrator, 9a, 9b, 9c: proportional integrator, 10
a, 10b, 10c ... adder / subtractor, 11 ... adder, 12 ...
Neural network means, 13 ... magnetic flux command generator.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H02P 5 / 408-5 / 412 G05B 13/02

Claims (3)

(57) [Claims] An input of a power converter for driving an induction motor is provided.
A parameter detecting means for detecting the force voltage and output current, and calculating means for calculating a secondary flux calculation value and the torque current calculation value of the induction motor based on an input voltage and output current of the power converter, wherein the induction Secondary magnetic flux command value and torque current for motor
A command value, a secondary magnetic flux calculation value calculated by the calculation means,
And a torque current calculation value are input, and a set of the secondary magnetic flux command value and the torque current command value or
Select a set of the secondary magnetic flux calculation value and torque current calculation value
The second magnetic flux command value or the second magnetic flux calculation value
Magnetic flux current command value obtained by multiplying the magnetizing current weight coefficient of
And outputs the secondary magnetic flux command value or the reciprocal of the secondary magnetic flux calculation value
Multiplied by a second magnetizing current weighting factor and the torque current
Flow command value or the torque current calculation value for the first slip
All values obtained by multiplying by the value multiplied by the frequency weighting factor
Output the frequency command value, and select the set of the secondary magnetic flux command value and the torque current command value.
The magnetic flux current command value and the slip frequency command when
Value and a set of the secondary magnetic flux calculation value and the torque current calculation value.
The magnetic flux current command value and the slip frequency when selected
The first magnetizing current so that the difference from the command value becomes zero.
Changing the weight coefficient and the first slip frequency weight coefficient.
A neural network means that the detected and actual actual torque current value and the flux current value of the induction motor on the basis of said slip frequency command value output from the neural network means and the output current of the power converter Vector control for controlling the induction motor in accordance with a deviation between the detected actual magnetic flux current value and the magnetic flux current command value, and a deviation between the detected actual torque current value and the torque current command value. Means for controlling an induction motor vector. 2. An input of a power converter for driving an induction motor.
A parameter detecting means for detecting the force voltage and output current, and calculating means for calculating a secondary flux calculation value and the torque current calculation value of the induction motor based on an input voltage and output current of the power converter, wherein the induction Secondary magnetic flux command value and torque current for motor
A command value, a secondary magnetic flux calculation value calculated by the calculation means,
And the calculated torque current value and the actual torque current value
The combination of the input secondary magnetic flux command value and the actual torque current value
Or the set of the secondary magnetic flux calculation value and the torque current calculation value is
Or the set of the secondary magnetic flux command value and the torque current command value
And the first magnetic flux command value or the second magnetic flux calculation value
And the secondary magnetic flux command value
Alternatively, the second magnetizing current is added to the differential value of the secondary magnetic flux operation value.
Magnetic flux current finger obtained by adding the value multiplied by the weight coefficient
And outputs the actual torque current value or the calculated torque current value.
Alternatively, the first slip frequency weight is added to the torque current command value.
Divided by the secondary magnetic flux command value
The obtained slip frequency command value is output, and the set of the secondary magnetic flux command value and the actual torque current is selected.
The magnetic flux current command value and the slip frequency command when
Value and the calculated value of the secondary magnetic flux calculated value and the torque current calculated value
The magnetic flux current command value and the total
So that the difference from the frequency command value becomes zero.
A magnetizing current weighting factor, a second magnetizing current weighting factor and the first
Neural network means for changing the slip frequency weighting coefficient of, the actual magnetic flux current value of the induction motor and the actual magnetic flux current value of the induction motor based on the slip frequency command value output from the neural network means and the output current of the power converter A torque current value and a deviation between the detected actual magnetic flux current value and the magnetic flux current command value, and a deviation between the detected actual torque current value and the detected torque current command value. Vector control means for controlling the induction motor; and a vector control device for the induction motor, comprising: 3. An input of a power converter for driving an induction motor.
A parameter detecting means for detecting the force voltage and output current, and calculating means for calculating a secondary flux calculation value and the torque current calculation value of the induction motor based on an input voltage and output current of the power converter, wherein the induction Secondary magnetic flux command value and torque current for motor
A command value, a secondary magnetic flux calculation value calculated by the calculation means,
And a torque current calculation value are input, and a set of the secondary magnetic flux command value and the torque current command value or
Select a set of the secondary magnetic flux calculation value and torque current calculation value
The second magnetic flux command value or the second magnetic flux calculation value
And the secondary magnetic flux command value
Alternatively, the second magnetizing current is added to the differential value of the secondary magnetic flux operation value.
Magnetic flux current finger obtained by adding the value multiplied by the weight coefficient
Command value and output the torque current calculation value or the torque current command value.
A value multiplied by a first slip frequency weighting factor and the secondary magnetic field;
Output the slip frequency command value obtained by dividing the
And select the set of the secondary magnetic flux command value and the torque current command value.
The magnetic flux current command value and the slip frequency command when
Value and a set of the secondary magnetic flux calculation value and the torque current calculation value.
The magnetic flux current command value and the slip frequency when selected
The first magnetizing current so that the difference from the command value becomes zero.
A weight coefficient, the second magnetizing current weight coefficient, and the first switch.
Neural network means for changing the slip frequency weighting coefficient , based on the slip frequency command value output from the neural network means and the output current of the power converter , the actual magnetic flux current value and the actual torque of the induction motor. Current value and a deviation between the detected actual magnetic flux current value and the magnetic flux current command value, and the induction according to a deviation between the detected actual torque current value and the detected torque current command value. Vector control means for controlling an electric motor, and a vector control device for an induction motor, comprising: