JP7183475B2 - power converter - Google Patents

Description

The present application relates to a power converter.

Conventionally, in the drive control of a polyphase AC rotating electrical machine by a power conversion device, there is instantaneous current control that directly determines the switching state of the power converter based on the driving state of the rotating electrical machine. control is known.

For example, a conventional power conversion device described in Patent Document 1 controls a rotating electrical machine using direct torque control based on model predictive control. This control method predicts the torque of the rotating electrical machine and the magnetic flux of the stator several steps ahead, and controls within a predetermined allowable range. Explore switching patterns. Therefore, it is expected that switching loss in a steady state can be reduced while maintaining a fast torque response time in a transient state by model predictive control.

Also, the conventional power conversion device described in Patent Document 2 controls an induction motor using a neural network. This neural network performs machine learning using data on optimal acceleration/deceleration rates in advance for various types of loads, inverter parameters, output current, main circuit voltage, and output speed as teacher data. Then, by utilizing the learned neural network when driving the induction motor, the induction motor can be driven with the optimum acceleration/deceleration rate signal.

Japanese Unexamined Patent Application Publication No. 2011-152038 Patent No. 3675505

In the conventional power converter described in Patent Document 1, in order to further reduce the switching loss, it is necessary to predict a longer interval, and the amount of calculation increases exponentially.

Further, in the conventional power conversion device described in Patent Document 2, the induction motor can be driven by an optimum acceleration/deceleration rate signal under pre-learned conditions. However, when a situation different from the learning conditions occurs, it is difficult to predict the output of the neural network, and the reliability of the control cannot be ensured.

The present application discloses a technique for solving the above-described problems, and provides a power converter that can suppress an increase in the amount of calculation in drive control of a rotating electric machine and realize highly reliable drive control. intended to

A power converter disclosed in the present application includes a plurality of switching elements, a power converter that converts DC power from a DC power supply into AC power and supplies it to a rotating electric machine, and a pattern A control device that determines switching states for one cycle in the plurality of switching elements using a generating function, generates a switching pattern composed of a combination of the switching states for one cycle, and controls the output of the power converter. and The control device prestores a first function and a second function for determining the switching pattern, generates a control command to be input to the pattern generation function, and generates the first function according to the range of the control command. , the second function is selected and used as the pattern generation function. The first function generates the switching pattern by inference using a trained model that performs inference based on input based on information obtained by machine learning based on teacher data, and the second function is The switching pattern is generated by a preset calculation based on the input.

According to the power conversion device disclosed in the present application, it is possible to provide a power conversion device capable of suppressing an increase in the amount of calculation in drive control of a rotating electrical machine and realizing highly reliable drive control.

1 is a block diagram showing the configuration of a power converter according to Embodiment 1; FIG. 2 is a hardware configuration diagram for realizing the power converter according to Embodiment 1. FIG. 4 is a diagram showing an example of switching states of the power converter according to Embodiment 1; FIG. FIG. 4 is a diagram for explaining switching patterns according to Embodiment 1; 3 is a block diagram of a V/f controller in the power converter according to Embodiment 1; FIG. 2 is a block diagram illustrating machine learning based on trained models and teacher data according to Embodiment 1. FIG. 2 is a hardware configuration diagram for generating a trained model according to Embodiment 1; FIG. 4 is a flowchart for explaining control operations in the power converter according to Embodiment 1; 2 is a block diagram showing the configuration of a power conversion device according to Embodiment 2; FIG. 9 is a flowchart for explaining control operations in the power conversion device according to Embodiment 2; FIG. 11 is a block diagram showing the configuration of a power converter according to Embodiment 3; FIG. 10 is a hardware configuration diagram for realizing a power conversion device according to Embodiment 3; FIG. 11 is a diagram for explaining model predictive control according to Embodiment 3; 10 is a flowchart for explaining control operations in a power conversion device according to Embodiment 3; 10 is a flowchart for explaining control operations in a power conversion device according to Embodiment 3; FIG. 12 is a block diagram showing the configuration of a power converter according to Embodiment 4; FIG. 11 is a hardware configuration diagram for realizing a power conversion device according to Embodiment 4; 10 is a flowchart for explaining control operations in the power conversion device according to Embodiment 4; 10 is a flowchart for explaining control operations in the power conversion device according to Embodiment 4;

Embodiment 1.
FIG. 1 is a block diagram showing the configuration of a power converter according to Embodiment 1. FIG.
As shown in the figure, the power conversion device 100 includes a power converter 1 as a main circuit and a control device 10 that controls the output of the power converter 1, and is connected between a DC power supply 2 and a rotating electric machine 3. .
The power converter 1 converts the DC power from the DC power supply 2 into AC power and supplies the AC power to the rotating electrical machine 3 to drive the rotating electrical machine 3 . The rotating electric machine 3 converts the AC power supplied from the power converter 1 into power. Various electric motors such as an induction motor and a synchronous motor can be used as the rotating electric machine 3 .

The control device 10 includes a V/f controller 11, a pattern determination section 12 that determines a switching pattern SWP, and a storage section 13 that holds a first function F1 and a second function F2.
The first function F1 generates a switching pattern SWP by inference using a trained model M that performs inference based on the input to the first function F1 based on information obtained by machine learning based on teacher data. . In this case, the trained model M is stored in the storage unit 13 and operates as the first function F1. The second function F2 also generates the switching pattern SWP using pulse width modulation (PWM) based on the inputs to the second function F2. The PWM method is a control method that normalizes the voltage command value Vref(k) using the bus voltage of the DC power supply 2 and compares it with a triangular wave that is a carrier wave.
The switching pattern SWP is a combination of switching states for one cycle, and the details will be described later.

The V/f controller 11 adjusts the three-phase (U-phase, V-phase, and W-phase) voltage command values Vref based on the speed command ωref(k), which is a given control target, for each set control cycle. (k) (: Vuref(k), Vvref(k), Vwref(k)). In addition, the control device 10 detects at least one switching state SW(k-1) (: SWu(k-1), SWv(k −1), SWw(k−1)). Then, the storage unit 13 uses the voltage command value Vref(k) and at least one switching state SW(k−1) in the previous cycle (k−1) as control commands, and based on the control commands, performs the first Either one of the function F1 and the second function F2 is output as the pattern generation function F to the pattern determination unit 12 . Note that the above control command serves as an input to the pattern generation function F. FIG.

The pattern determination unit 12, based on a control command consisting of a voltage command value Vref(k) and at least one switching state SW(k-1) in the previous cycle (k-1), and a pattern generation function F, That is, the control command is used as an input to the pattern generation function F to generate the switching pattern SWP(k) for the current cycle k.
In this manner, the control device 10 outputs the switching pattern SWP(k) to control the power converter 1 in each control cycle.

The notations (k-1) and (k) represent discrete time signals for each control cycle, where (k-1) is the previous value, (k) is the current value, and (k+1) is the next value. be. This also applies to FIG. 1 and subsequent figures.

FIG. 2 is a hardware configuration diagram for realizing the power converter 100. As shown in FIG.
The power converter 1 is composed of a three-phase inverter circuit that converts DC power from a DC power supply 2 into three-phase AC power, and drives a rotating electric machine 3 as a load. The power converter 1 includes a plurality of switching elements Q1 to Q6 each having a diode D connected in antiparallel. A connection point between the upper arm and the lower arm of each phase is connected to an input terminal of each phase of the rotary electric machine 3 by a bus bar. In this case, the U phase has switching elements Q1 and Q2, the V phase has switching elements Q3 and Q4, and the W phase has switching elements Q5 and Q6.

The control device 10 is composed of a processor 20 and a storage device 21 .
The storage device 21 includes a volatile storage device (not shown) such as a RAM (Random Access Memory), and a non-volatile auxiliary storage device (not shown) such as a HDD (Hard Disk Drive) and SSD (Solid State Drive). It has A flash memory may be used instead of the HDD as the non-volatile auxiliary storage device.

Processor 20 executes a control program input from storage device 21 .
The storage device 21 includes an auxiliary storage device and a volatile storage device. A control program 22 is input to the processor 20 from an auxiliary storage device via a volatile storage device.
The processor 20 outputs data 23 such as calculation results to the volatile storage device of the storage device 21, and stores these data 23 in the auxiliary storage device via the volatile storage device as necessary.

In FIG. 2, only the processor 20 and the storage device 21 constitute the control device 10, but the processor 20 and the storage device 21, and a dedicated circuit for PWM control, are not limited to this. It may also include a triangular wave generator and a PWM circuit or the like.

As described above, the control device 10 controls the power converter 1 by outputting the switching pattern SWP(k), which is a combination of switching states for one cycle, for each control cycle.
FIG. 3 is a diagram showing an example of switching states of the power converter 1. As shown in FIG. The switching state is a combination of ON (:1) and OFF (:0) signals for each switching element Q1-Q6. Eight switching states (SW1 to SW8) in which one of the switching elements Q1 to Q6 of the upper arm and the lower arm is turned on and the other is turned off, and all the switching elements Q1 to Q6 are turned off when the power converter 1 stops operating. There are 9 switching states of the switching state (SW0) to turn off. In the switching state SW1 (SWu1, SWv1, SWw1) of FIG. 3, the switching elements Q1, Q4, Q6 are on, and the switching elements Q2, Q3, Q5 are off.

FIG. 4 is a diagram for explaining the switching pattern SWP(k). The switching pattern SWP(k) is a combination of switching states for one cycle. One cycle T is divided into a plurality of sections, and the switching states assigned to each section are determined. In this case, the switching pattern SWP(k) is a combination in which the switching states SW3, SW4, SW2, SW2, SW6, and SW7 are set to switch in order.
One cycle may be divided into intervals of a predetermined width and switching states may be assigned to each, or timing information for switching between different switching states may be added to switching state information.

As shown in FIG. 4, the power converter 1 is operating in the switching state SW1 at the current time t(k), and the previous switching state is SW7. The switching state SW1 at the current time t(k) also continues from the previous period (k-1), and the switching states SW1 and SW7 are the switching states SW(k-1) in the previous period (k-1). .

The control device 10 detects the switching state SW(k-1), such as SW1 and SW7, in at least one previous period (k-1) including the switching state SW1 at the current time t(k), and the voltage command value Vref(k ), a switching pattern SWP(k) (:SW3, SW4, SW2, SW2, SW6, SW7) for one cycle of the current cycle k is generated by a pattern generation function F. The switching pattern SWP(k) is given to the power converter 1 as a switching state command for one cycle, and the switching elements Q1 to Q6 are on/off controlled.
Then, at time t(k+1), control device 10 generates switching pattern SWP(k+1) for the next cycle (k+1).

In this case, although the control device 10 receives and acquires the actual switching state SW(k-1) in the previous cycle (k-1) from the power converter 1, the control device 10 The switching state SW(k-1) in the previous period (k-1) may be obtained from the switching pattern SWP(k-1) generated in 1.

Note that one period T of the control period is usually much longer than the duration of one switching state, so the switching pattern SWP consists of a combination of a plurality of switching states. However, if the control cycle can be shortened to the same extent as the duration of the switching state, one switching state may correspond to one cycle.

Also, the switching state SW(k-1) in the previous cycle, which is part of the control command that is the input of the pattern generation function F, can be applied even if it is only the current switching state, but it is desirable that there are a plurality of them. Further, when the control cycle is short, the switching state in the previous cycle is not limited to the previous cycle (k-1), and for example, the switching state of the previous cycle (k-2) may also be adopted.

FIG. 5 is a block diagram of the V/f controller 11. As shown in FIG.
As shown in FIG. 5, the V/f controller 11 includes a Vδref calculator 111, a γδ/uvw coordinate converter 112, and an integrator 113. A Vδref calculator 111 calculates a δ-axis voltage command value Vδref(k) from the speed command ωref(k) based on a preset V/f ratio. The speed command ωref(k) is also input to the integrator 113, and the integrator 113 outputs the phase command θref(k). Based on the γ-axis voltage command value Vγref(k) to which 0 is input and the δ-axis voltage command value Vδref(k), the γδ/uvw coordinate converter 112 converts three phases (U phase, V phase, W phase) voltage command value Vref(k).
Note that the V/f control is a known technique, and the detailed description of the control is omitted.

By the way, the storage unit 13 holds a first function F1 and a second function F2, and the first function F1 is composed of a trained model M generated by machine learning based on teacher data.
The trained model M and its generation method will be described below. Note that the trained model M is stored in the storage unit 13 after being generated by completing machine learning in advance.
FIG. 6 is a block diagram illustrating machine learning based on the learned model M and teacher data.
As shown in FIG. 6, the learning unit 40 performs machine learning to generate a trained model M based on teacher data 47 obtained from learning data 41 prepared in advance.

The learning data 41 includes a voltage command value Vref(k) for each control cycle and a switching pattern SWP(k) that is a switching state SW for one cycle corresponding to the voltage command value Vref(k). In this case, the learning data 41 is a control method that reduces the switching loss of the power converter 1 compared to the PWM method. SWP(k)) is generated and obtained. Alternatively, the learning data 41 may be generated using a control method using selective harmonic elimination, low-order harmonic elimination, or an optimum pulse pattern.

The learning data 41 are input to the teacher data acquisition unit 42 . The teacher data acquisition section 42 includes an input data acquisition section 43 and a label data acquisition section 44 .
The input data acquisition unit 43 acquires the voltage command value Vref(k) for each control cycle and at least one switching state SW(k−1) in the previous cycle from the learning data 41 as teacher input data 45. and output to the learning unit 40 . At least one switching state SW(k-1) in the previous period includes the current switching state SW at the point of generation of the voltage command value Vref(k), and is obtained from the switching pattern SWP(k-1). can. The switching state at the end of the switching pattern SWP(k-1) corresponds to the current switching state SW.

The label data acquisition unit 44 acquires the switching pattern SWP(k) corresponding to the voltage command value Vref(k) for each control cycle from the learning data 41 as the teacher label data 46 and outputs it to the learning unit 40 .
The teacher data 47 consists of teacher input data 45 and teacher label data 46, and the learning unit 40 is based on the teacher data 47, which is a combination of the teacher input data 45 and the teacher label data 46 for each control cycle. to perform machine learning.

Learning with teacher data for drive control of the rotary electric machine 3 in the first embodiment is performed by a neural network configured by combining perceptrons. Specifically, the voltage command value Vref(k) and at least one switching state SW(k−1) in the previous period are used as teacher input data 45, and the switching pattern SWP(k) is used as teacher label data 46. and Then, the teacher data 47 based on these is given to the neural network, and learning is repeated while changing the weighting for each perceptron so that the output of the neural network becomes the same as the label data 46 for teacher.

In the learning process, weighting values are adjusted so as to reduce the error in the output of each perceptron by repeating error backpropagation. In other words, learning with teacher data eliminates errors between the teacher label data 46 and the output data of the neural network while adjusting the weighting values.
In this way, the learned model M is acquired for learning the features of the teacher data 47, making inferences based on the input, and deriving results.

The learned model M generated by machine learning in this way has the characteristics of the teacher data 47 . For example, if the learning data 41 to be transformed into the teacher data 47 uses model predictive control, the learned model M has a switching pattern that reduces the switching loss of the power converter 1 compared to the PWM method. Output SWP(k).

The neural network used by the learning unit 40 for learning may have three layers, or may have more layers, and may perform machine learning by deep learning.

FIG. 7 is a hardware configuration diagram for generating a trained model M. As shown in FIG. Machine learning for generating the learned model M is performed by a machine learning device 50 functioning as a neural network, and the machine learning device 50 is realized by the hardware configuration shown in FIG.
The machine learning device 50 is composed of a processor 51 and a storage device 52 .
The storage device 52 includes a volatile storage device such as a RAM 53 and a non-volatile auxiliary storage device such as an HDD 54 . Note that an SSD or flash memory may be used instead of the HDD as the non-volatile auxiliary storage device.

The HDD 54 holds a learning program 55, teacher data 56, and also holds a learning result 57 that is generated.
Various learning programs 55 are input to the processor 51 from the HDD 54 via the RAM 53, and the input various learning programs 55 are executed. The learning program 55 causes the processor 51 to perform learning with teacher data. That is, the teacher data 56 is also input from the HDD 54 to the processor 51 via the RAM 53 and learned according to the learning program 55 .
Further, the processor 51 outputs the data of the learning result 57 to the RAM 53 of the storage device 52, and stores the data in the HDD 54 via the RAM 53 as necessary.

The learning program 55 is a program including instructions for causing the processor 51 to execute learning with teacher data and to generate data as a result of machine learning (learning result 57).

The machine learning device 50 as described above can be realized by a PC (Personal Computer), a server device, or the like. However, since the amount of calculation is large, for example, a GPU (Graphics Processing Units) is installed in a PC, and a GPU (General-Purpose computing on Graphics Processing Units) technology is used for arithmetic processing of learning with teacher data. It may be possible to process at high speed.

Next, the control operation in the power converter 100 of this Embodiment 1 is demonstrated below.
FIG. 8 is a flow chart for explaining the control operation in the power conversion device 100. As shown in FIG. This control operation is performed by the control device 10 in each control cycle, and the operation at the current time t(k) is shown.

First, the control device 10 causes the V/f controller 11 to generate the voltage command value Vref(k) based on the speed command ωref(k), that is, acquires the voltage command value Vref(k) (step S1).
Next, the control device 10 acquires at least one switching state SW(k-1) within the previous period (k-1) including the switching state SW1 at the current time t(k) (step S2).
Note that the voltage command value Vref(k) obtained in step S1 and the switching state SW(k-1) in the previous cycle (k-1) obtained in step S2 are input to the pattern generation function F. becomes a directive. After obtaining each piece of information of the control command, the control device 10 inputs the information to the pattern determination unit 12 and the storage unit 13 .
Further, steps S1 and S2 are not limited to the order described above.

Next, the control device 10 outputs a control command consisting of the voltage command value Vref(k) and the switching state SW(k-1) in the previous cycle (k-1) to the learned model M in the storage unit 13. It is determined whether or not it is within the range of the input data for teacher 45 (step S3).

In step S3, if the control command is within the range of the teacher input data 45, the storage unit 13 outputs the first function F1 using the trained model M as the pattern generation function F. Then, the pattern determination unit 12 acquires the first function F1 as the pattern generation function F (step S4).

Then, the pattern determination unit 12 determines the switching pattern SWP(k), which is the switching state for one cycle, by the pattern generation function F based on the control command (step S5).
Control device 10 controls the switching state of power converter 1 according to the determined switching pattern SWP(k) to drive and control rotary electric machine 3 (step S6).

In step S3, when the control command exceeds the range of the teacher input data 45, the storage unit 13 outputs the second function F2 using the PWM method as the pattern generation function F. FIG. That is, the pattern determination unit 12 acquires the second function F2 as the pattern generation function F (step S7). After that, the process proceeds to step S5.

As described above, the pattern determination unit 12 acquires either one of the first function F1 and the second function F2 as the pattern generation function F, and the switching state for one cycle is determined by the pattern generation function F based on the control command. is determined as a switching pattern SWP(k).
Selection of either the first function F1 or the second function F2 is carried out according to the range of the control command. That is, the control command is compared with the input data for teacher 45, and if the control command is within the range of the input data for teacher 45, the first function F1 is selected. Select function F2.
The comparison between the control command that is the input of the pattern generation function F and the teacher input data 45 is such that, for example, the current switching state SW is the same in both, and the voltage command value Vref(k) in the control command is the teacher input. It is sufficient if it is within the range of the data 45 .

In this embodiment, when the control command exceeds the range of the input data for teacher 45, the second function F2 is applied. As a result, the first function F1 using the learned model M can be applied only within the range of the teacher data 47, and is not used under conditions different from the learning conditions. Therefore, the first function F1 does not exceed the expected inference and derives a fast and highly reliable result. When the model predictive control is used to generate the learned model M, the switching loss of the power converter 1 can be reduced as compared with the PWM method.
Further, even when the control command exceeds the range of the teacher input data 45, the switching pattern SWP(k) corresponding to the control command can be continuously generated by applying the second function F2. The second function F2 reliably derives a result based on preset calculations.

The input/output relationship of each of the first function F1 and the second function F2 prepared for the pattern generation function F is the same as the input/output relationship of the pattern generation function F. FIG. Therefore, the types of information input to each function (F, F1, F2) are common, namely the voltage command value Vref(k) and the switching state SW(k-1), which are control commands. The type of information output from each function (F, F1, F2) is also common and is the switching pattern SWP(k). However, regarding the second function F2, the switching pattern SWP(k) is output using only the voltage command value Vref(k) among the input control commands.

The teacher input data 45 is data input to the neural network to generate the first function F1, and is information of the same kind as the control command input to the pattern generation function F, that is, the voltage command value Vref(k). and the switching state SW(k-1) in the previous period (k-1). Similarly, the teacher label data 46 is the same type of information as the switching pattern SWP(k), which is the output of the pattern generation function F. FIG.

It is assumed that the storage unit 13 outputs the pattern generation function F based on the voltage command value Vref(k) and at least one switching state SW(k-1) in the previous period (k-1). However, the pattern determination unit 12 may extract one of the first function F1 and the second function F2 in the storage unit 13 as the pattern generation function F.
In that case, the voltage command value Vref(k) and at least one switching state SW(k-1) in the previous period (k-1) are not input to the storage unit 13, but are sent only to the pattern determination unit 12. is entered. Then, the pattern determination unit 12 extracts one of the first function F1 and the second function F2 as the pattern generation function F based on these control commands.

Further, in the above embodiment, the PWM method is used for the second function F2, and the learned model M based on the control that reduces the switching loss compared to the PWM method is used for the first function F1. The control on which these first and second functions F1 and F2 are based is not limited to the one described above.

Embodiment 2.
FIG. 9 is a block diagram showing the configuration of the power converter according to Embodiment 2. In FIG.
In this embodiment, the control device 10 includes a teacher data acquisition section 14 . If the generated control command, that is, the voltage command value Vref(k) and the switching state SW(k−1) exceed the range of the teacher input data 45, the teacher data acquiring unit 14 acquires the control command and the switching state SW(k−1). The pattern SWP(k) is acquired as new teacher data. Other configurations are the same as those of the first embodiment.

FIG. 10 is a flow chart for explaining the control operation in power converter 100 according to the second embodiment. This control operation is performed by the control device 10 in each control cycle, and the operation at the current time t(k) is shown. In this case, steps S1 to S7 are the same as in the first embodiment.

First, the control device 10 causes the V/f controller 11 to generate the voltage command value Vref(k) based on the speed command ωref(k), that is, acquires the voltage command value Vref(k) (step S1).
Next, the control device 10 acquires at least one switching state SW(k-1) within the previous period (k-1) including the switching state SW1 at the current time t(k) (step S2).
Note that the voltage command value Vref(k) obtained in step S1 and the switching state SW(k-1) in the previous cycle (k-1) obtained in step S2 are input to the pattern generation function F. becomes a directive. After obtaining each piece of information of the control command, the control device 10 inputs the information to the pattern determination unit 12 and the storage unit 13 .

Next, the control device 10 outputs a control command consisting of the voltage command value Vref(k) and the switching state SW(k-1) in the previous cycle (k-1) to the learned model M in the storage unit 13. It is determined whether or not it is within the range of the input data for teacher 45 (step S3).

In step S3, if the control command is within the range of the teacher input data 45, the storage unit 13 outputs the first function F1 using the trained model M as the pattern generation function F. Then, the pattern determination unit 12 acquires the first function F1 as the pattern generation function F (step S4), and uses the pattern generation function F (first function F1) based on the control command to perform switching state for one cycle. A certain switching pattern SWP(k) is determined (step S5).
Control device 10 controls the switching state of power converter 1 according to the determined switching pattern SWP(k) to drive and control rotary electric machine 3 (step S6).

In step S3, when the control command exceeds the range of the teacher input data 45, the storage unit 13 outputs the second function F2 using the PWM method as the pattern generation function F. FIG. That is, the pattern determination unit 12 acquires the second function F2 as the pattern generation function F (step S7).

Then, the pattern determination unit 12 determines the switching pattern SWP(k), which is the switching state for one cycle, by the pattern generation function F (second function F2) based on the control command (step S8).
Next, the teacher data acquisition unit 14 acquires the control command input to the pattern generation function F (second function F2) as teacher input data, and outputs the pattern generation function F (second function F2). The switching pattern SWP(k) is acquired as teacher label data. That is, the teacher input data and teacher label data are obtained as new teacher data (step S9).
The control device 10 controls the switching state of the power converter 1 according to the determined switching pattern SWP(k) to drive and control the rotating electric machine 3 (step S10).

In this embodiment, as in the first embodiment, the first function F1 using the trained model M is applied only within the range of the teacher data 47, and otherwise the second function F2 is applied. use. Then, the input/output information when using the second function F2, that is, the control command and the switching pattern SWP(k) are acquired as new teacher data. Therefore, the same effect as in the first embodiment can be obtained, and information collection for updating the learned model M or expanding the applicable range can be performed.

The control device 10 can be connected to the upper controller, and the upper controller receives information from the teacher data acquisition unit 14, so that the upper controller or a machine learning device connected thereto reconstructs or reconstructs the trained model M. , a new trained model can be generated with expanded coverage.

Embodiment 3.
FIG. 11 is a block diagram showing the configuration of a power converter according to Embodiment 3. In FIG.
Moreover, FIG. 12 is a hardware block diagram which implement|achieves the power converter device by Embodiment 3. As shown in FIG.
As shown in the figure, a power conversion device 100A includes a power converter 1 as a main circuit and a control device 10A that controls the output of the power converter 1, and is connected between a DC power supply 2 and a rotating electric machine 3. .
The power converter 1 converts the DC power from the DC power supply 2 into AC power and supplies the AC power to the rotating electrical machine 3 to drive the rotating electrical machine 3 . The rotating electric machine 3 converts the AC power supplied from the power converter 1 into power.

In this embodiment, a current detector (hereinafter referred to as detector 15 ) that acquires a detected value indicating the drive state of rotating electrical machine 3 is provided between power converter 1 and rotating electrical machine 3 . Detectors 15 are provided in the AC output lines of each phase, and detect current values I(k) (Iu(k), Iv(k), Iw(k)) as detection values. A current value I(k) is the value of the output current of the power converter 1 .
Various electric motors such as an induction motor and a synchronous motor can be used as the rotating electric machine 3 .

The control device 10A includes a state observation unit 16 that calculates position θ(k), magnetic flux ψ(k), and speed ω(k), which are state quantities of the rotary electric machine 3, based on the current value I(k), and a PI ( A proportional integral) flux controller 17 and a PI speed controller 18 are provided. Further, the control device 10A includes a pattern determination section 12A that determines the switching pattern SWP, and a storage section 13A that holds two first functions F1A, F1B and a second function F2A.
Note that, as shown in FIG. 12, the hardware configuration for realizing power converter 1 and control device 10A is the same as in the first embodiment.

Each of the first functions F1A and F1B uses learned models MA and MB that perform inference based on inputs to each of the first functions F1A and F1B based on information obtained by machine learning based on teacher data. The inference generates a switching pattern SWP. In this case, the trained models MA and MB are stored in the storage unit 13A and operate as the first functions F1A and F1B, respectively. The learned model MA and the learned model MB have different teacher data. For example, the learned model MA uses the information of the rotary electric machine 3 for low-speed rotation below the set speed as teacher data, and the learned model MB uses the above information. It is learned by using the information of the rotary electric machine 3 for high-speed rotation exceeding the set speed as teacher data.

Also, the second function F2 generates a switching pattern SWP using model predictive control based on the inputs to the second function F2.
Note that the switching pattern SWP is a combination of switching states for one cycle, as described in the first embodiment.

The detector 15 may detect current values (Iu(k), Iv(k), Iw(k)) of three phases (U phase, V phase, W phase) as the current value I(k). Alternatively, the current values for two phases may be detected and the remaining one phase calculated may be used. Also, a one-shunt current detection method that restores three-phase current values I(k) with one detector 15 may be used.
Here, for the detector 15, any current detector such as a CT (Current Transformer) detector or a shunt resistor may be used.

The state observation unit 16 calculates position θ(k), magnetic flux ψ(k), and speed ω(k), which are state quantities of the rotating electrical machine 3, based on the current value I(k). A parameter resistance value or inductance value may be used.

A control device 10A is provided with a speed command ωref(k) and a magnetic flux command ψref(k), which are control targets of the rotating electric machine 3 .
Based on the magnetic flux command ψref(k) and the magnetic flux ψ(k) calculated by the state observation unit 16, the PI magnetic flux controller 17 calculates the magnetic flux command ψref(k) and the magnetic flux ψ(k) for each control cycle. d-axis current command value Idref(k) for reducing the deviation of . Based on the speed command ωref(k) and the speed ω(k) calculated by the state observation unit 16, the PI speed controller 18 determines the speed command ωref(k) and the speed ω(k) for each control cycle. A q-axis current command value Idref(k) for reducing the deviation of is output.

Since the PI control is a well-known technique, detailed explanation is omitted.
Also, although the PI speed controller 18 that outputs the q-axis current command value Idref(k) is used, the rotary electric machine A PI speed controller that outputs three torque command values τref(k) may be used.

In addition, the control device 10A sets at least one switching state SW(k-1) (: SWu(k-1), SWv(k −1), SWw(k−1)).
The control device 10A has a dq-axis current command value Idqref(k) (Idref(k), Iqref(k)) as a current command value, a current value I(k) as a detected value, and a position θ (k) and speed ω(k) and at least one switching state SW(k-1) in the previous period (k-1) are used as control commands. Then, the storage unit 13A selects one of the first functions F1A, F1B and the second function F2 based on this control command and outputs it as the pattern generation function F to the pattern determination unit 12A. Note that the above control command serves as an input to the pattern generation function F. FIG.

Based on the control command and the pattern generation function F, that is, using the control command as an input to the pattern generation function F, the pattern determination unit 12A generates the switching pattern SWP(k) of the current cycle k.
In this manner, the control device 10A outputs the switching pattern SWP(k) to control the power converter 1 in each control cycle.

Note that at least one switching state SW(k-1) in the previous period (k-1), which is part of the input of the pattern generation function F, and the switching pattern SWP, which is the output of the pattern generation function F, are is the same as in form 1 of

Next, the model predictive control used for the second function F2 will be described below with reference to FIG.
In the case shown in the figure, the control device 10A determines the switching state so as to maintain the current value I (detected current value) within the set allowable current width. At the current time t(k), based on the current value I and the switching state SW(k-1) in at least one previous cycle (k-1) including the switching state SW1 at the current time t(k), the rotating electric machine 3, a candidate for the switching pattern SWP(k) (switching state for one cycle) for the current cycle is generated. Then, the predicted value of the current value I is calculated for each candidate. FIG. 13 shows the current t(k) and the previous current detection value 30, and the current predicted values 31, 31a, 31b, 31c, 31d, and 31e for each switching pattern SWP(k) candidate after the current time.

Then, among the predicted current values 31, 31a, 31b, 31c, 31d, and 31e, the switching pattern SWP(k) of the case (predicted current value 31) in which the time to reach the upper limit or lower limit of the allowable current width is the longest is searched. do.
As a result, the rotating electric machine 3 can be driven with the switching loss of the power converter 1 smaller than that of the PWM method.
Note that the model predictive control described above is an example, and at least one of the driving sound of the rotating electrical machine 3, mechanical vibration, current harmonics, and the follow-up time of the current value to the current command value is set to the PWM method. You may control to make it small.

Next, the control operation in the power converter 100A of this Embodiment 3 is demonstrated below.
14 and 15 are flowcharts for explaining the control operation in the power conversion device 100A. This control operation is performed by the control device 10A in each control cycle, and the operation at the current time t(k) is shown.
Also in this case, when the control device 10A acquires each piece of information as a control command, the control device 10A inputs the information into the pattern determination section 12A and the storage section 13A.

Detector 15 detects current value I(k) (Iu(k), Iv(k), Iw(k)), and control device 10A converts detected current value I(k) to one of control commands. It is acquired as a part and inputted to the pattern determination part 12A and the storage part 13A (step SS1).
Next, state observation unit 16 calculates position θ(k), magnetic flux ψ(k), and speed ω(k), which are state quantities of rotating electric machine 3, based on current value I(k) (step SS2). , the control device 10A acquires the position .theta.(k) and the velocity .omega.(k) as part of the control command, and inputs them to the pattern determining section 12A and the storing section 13A (step SS3).

Next, the PI flux controller 17 outputs a d-axis current command value Idref(k) based on the magnetic flux command ψref(k) and the magnetic flux ψ(k), and the PI speed controller 18 outputs Then, the q-axis current command value Idref(k) is output based on the speed command ωref(k) and the speed ω(k). Then, the control device 10A acquires the dq-axis current command value Idqref(k) (Idref(k), Iqref(k)) as part of the control command, and inputs it to the pattern determination unit 12A and the storage unit 13A ( step SS4).
Next, the control device 10A acquires at least one switching state SW(k-1) within the previous period (k-1) including the switching state at the current time t(k) (step SS5).
Note that steps SS1 to SS5 are not limited to the order described above.

Next, the control device 10A controls the dq-axis current command value Idqref(k), the current value I(k), the position θ(k), the speed ω(k), and at least Using one switching state SW(k-1) as a control command, it is determined whether or not the control command is within the range of teacher input data of the learned model MA in the storage unit 13A (step SS6).

In step SS6, if the control command is within the range of the teacher input data of the learned model MA, the storage unit 13A outputs the first function F1A using the learned model MA as the pattern generation function F. Then, the pattern determination unit 12A acquires the first function F1A as the pattern generation function F (step SS7).

Then, the pattern determination unit 12A determines the switching pattern SWP(k), which is the switching state for one cycle, by the pattern generation function F based on the control command (step SS8).
The control device 10A controls the switching state of the power converter 1 according to the determined switching pattern SWP(k) to drive and control the rotating electric machine 3 (step SS9).

In step SS6, if the control command exceeds the range of the teacher input data of the learned model MA, the control device 10A determines whether the control command is within the range of the teacher input data of the learned model MB in the storage unit 13A. It is determined whether or not (step SS10).
In step SS10, if the control command is within the range of the teacher input data of the learned model MB, the storage unit 13A outputs the first function F1B using the learned model MB as the pattern generation function F. Then, the pattern determination unit 12A acquires the first function F1B as the pattern generation function F (step SS11). After that, the process proceeds to step SS8.

In step SS10, when the control command exceeds the range of the teacher input data of the learned model MB, the storage unit 13A outputs the second function F2A using the model prediction method as the pattern generation function F. That is, the pattern determination unit 12A acquires the second function F2A as the pattern generation function F (step SS12). After that, the process proceeds to step SS8.

As described above, the control device 10A according to the third embodiment can control the current value I(k), the position θ(k), the magnetic flux ψ(k), and the speed ω(k) that indicate the driving state of the rotating electric machine 3. to determine the switching pattern SWP(k) for controlling the switching state of the power converter 1 . Therefore, at least one of the drive sound of the rotary electric machine 3, mechanical vibration, current harmonics, and the follow-up time of the current value to the current command value can be reduced compared to the PWM method.

In addition, in the third embodiment, the storage unit 13A includes two first functions F1A using two trained models MA and MB that perform inference based on information obtained by machine learning based on different teacher data, Retain F1B and a second function F2A using model predictive control. Then, the control device 10A selects one of the two first functions F1A and F1B and the second function F2A as the pattern generation function F according to the range of the generated control command, and selects the switching pattern SWP (k) is determined.

Selection of one pattern generation function F from among the first functions F1A, F1B and the second function F2A is performed according to the range of the control command, as in the first embodiment. In this case, the first function F1A is preferentially used over the first function F1B, and the control command is compared with the teacher input data of the learned model MA to determine whether the control command falls within the range of the teacher input data of the learned model MA. If within, select the first function F1A. If the control command exceeds the range of input data for teacher of learned model MA and is within the range of input data for teacher of learned model MB, first function F1B is selected. Then, when the control command exceeds the range of input data for teacher of the learned model MB, the second function F2 is selected.
The comparison between the control command that is the input of the pattern generation function F and the teacher input data 45 is, for example, that the current switching state SW is the same in both, and at least one of the other control commands, such as the dq-axis current command Alternatively, the value Idqref(k) may be compared.

Also in this embodiment, the first functions F1A and F1B using the trained models MA and MB can be applied only within the range of the respective teacher data, and are not used under conditions different from the learning conditions. Therefore, each of the first functions F1A and F1B can derive fast and highly reliable results without exceeding the expected inference. In addition, since two trained models MA and MB are prepared using different teacher data, highly reliable trained models MA and MB can be obtained by simple machine learning.
Further, even when the control command exceeds the range of the teacher input data of both learned models MA and MB, the switching pattern SWP(k) corresponding to the control command is continued by applying the second function F2A. can be generated by The second function F2A reliably derives a result based on preset calculations.

Also, the input/output relationship of the first functions F1A, F1B and the second function F2A prepared for the pattern generation function F is the same as the input/output relationship of the pattern generation function F. FIG. Therefore, the type of information input to each function (F, F1A, F1B, F2A) is common, and information that is a control command, that is, the dq-axis current command value Idqref(k) and the current value I(k) , position θ(k) and velocity ω(k), and at least one switching state SW(k−1) within the previous period (k−1). The type of information output from each function (F, F1A, F1B, F2A) is also common and is the switching pattern SWP(k). However, regarding the second function F2A, it is not necessary to use all of the input control commands, and for example, the switching pattern SWP(k) is output without using the switching state SW(k-1).

The teacher input data of the learned models MA and MB are data input to the neural network to generate the first functions F1A and F1B, respectively, and are the same kind of information as the control commands input to the pattern generation function F. is. Similarly, each of the teacher label data of the learned models MA and MB is the same type of information as the switching pattern SWP(k), which is the output of the pattern generation function F. FIG.

Also in this embodiment, the pattern determination unit 12A may extract one of the first functions F1A, F1B and the second function F2A in the storage unit 13A as the pattern generation function F.
In that case, the generated control command is input only to the pattern determination unit 12A without being input to the storage unit 13A. Then, the pattern determination unit 12A extracts the pattern generation function F from the storage unit 13A based on the control command.

In the above-described embodiment, the model prediction form is used for the second function F2A. Applicable.

Furthermore, although the two first functions F1A and F1B using the two trained models MA and MB are shown, the storage unit 13A may hold three or more first functions.

Further, by applying the second embodiment, the control device 10A may be provided with the teacher data acquisition unit 14 to acquire new teacher data.

Embodiment 4.
FIG. 16 is a block diagram showing the configuration of a power converter according to Embodiment 4. In FIG.
Moreover, FIG. 17 is a hardware block diagram which implement|achieves the power converter device by Embodiment 4. As shown in FIG.
As shown in the figure, the power conversion device 100B includes a power converter 1 as a main circuit and a control device 10B that controls the output of the power converter 1, and is connected between the DC power supply 2 and the rotating electric machine 3. .
The power converter 1 converts the DC power from the DC power supply 2 into AC power and supplies the AC power to the rotating electrical machine 3 to drive the rotating electrical machine 3 . The rotating electric machine 3 converts the AC power supplied from the power converter 1 into power.

In this embodiment, in order for the power conversion device 100B to acquire a detected value indicating the driving state of the rotating electric machine 3, a current detector (detector 15) similar to that in the third embodiment and a speed detector 15A and a position detector 15B.
Detector 15 is provided in the AC output line of each phase between power converter 1 and rotary electric machine 3, and detects current values I(k) (Iu(k), Iv(k), Iw(k )).

The detector 15 may detect current values (Iu(k), Iv(k), Iw(k)) of three phases (U phase, V phase, W phase) as the current value I(k). Alternatively, the current values for two phases may be detected and the remaining one phase calculated may be used. Also, a one-shunt current detection method that restores three-phase current values I(k) with one detector 15 may be used. Any current detector such as a CT detector or a shunt resistor may be used for the detector 15 .

Speed detector 15A and position detector 15B are provided in rotating electric machine 3 . The speed detector 15A detects the rotational speed of the rotary electric machine 3 as speed ω(k). An encoder, resolver, Hall sensor, etc. are used for the speed detector 15A.
The position detector 15B detects the rotational position of the rotary electric machine 3 as the position θ(k). An encoder, a resolver, a Hall sensor, etc. are used for the position detector 15B.
Various electric motors such as an induction motor and a synchronous motor can be used as the rotating electric machine 3 .

The control device 10B includes a magnetic flux calculator 19 that calculates the magnetic flux ψ(k) of the rotating electric machine 3 based on the current value I(k), a pattern decision unit 12B that decides the switching pattern SWP, a first function F1C and a first function F1C. and a storage unit 13B that holds two functions F2B.
Note that, as shown in FIG. 17, the hardware configuration for realizing power converter 1 and control device 10B is the same as in the first embodiment.

The first function F1C generates a switching pattern SWP by inference based on information obtained by machine learning based on teacher data, using a learned model MC that performs inference based on inputs to the first function F1C. . In this case, the trained model MC is stored in the storage unit 13B and operates as the first function F1C.
The switching pattern SWP is, as described in the first embodiment, a combination of switching states for one cycle.

Also, the second function F2C combines PI flux control, PI speed control, PI current control, and PWM control to generate a switching pattern SWP based on the input to the second function F2C.
In FIG. 17, only the processor 20 and the storage device 21 constitute the control device 10B, but each PI controller for controlling magnetic flux, speed and current, and triangular wave generation for PWM control It may be provided with a device, a PWM circuit, and the like.

A control device 10B is provided with a speed command ωref(k) and a magnetic flux command ψref(k), which are control targets for the rotating electrical machine 3 . The control device 10B also stores the current value I(k) detected by the detector 15, the speed ω(k) detected by the speed detector 15A, and the position detected by the position detector 15B. θ(k) is input. The magnetic flux calculator 19 calculates the magnetic flux ψ(k) of the rotating electric machine 3 based on the current value I(k).
In addition, the control device 10B detects at least one switching state SW(k-1) (: SWu(k-1), SWv(k −1), SWw(k−1)).

The control device 10B outputs a speed command ωref(k) and a magnetic flux command ψref(k), current value I(k), position θ(k) and speed ω(k) which are detected values, and calculated magnetic flux ψ( k) and at least one switching state SW(k-1) in the previous cycle (k-1) are used as control commands. Then, the storage unit 13B selects one of the first function F1C and the second function F2B based on this control command and outputs it as the pattern generation function F to the pattern determination unit 12B. Note that the above control command serves as an input to the pattern generation function F. FIG.

Based on the control command and the pattern generation function F, that is, using the control command as an input to the pattern generation function F, the pattern determination unit 12B generates the switching pattern SWP(k) of the current cycle k.
In this manner, the control device 10B outputs the switching pattern SWP(k) to control the power converter 1 in each control cycle.

Note that at least one switching state SW(k-1) in the previous period (k-1), which is part of the input of the pattern generation function F, and the switching pattern SWP, which is the output of the pattern generation function F, are is the same as in form 1 of

Next, the control operation in the power converter 100B of this Embodiment 4 is demonstrated below.
18 and 19 are flowcharts for explaining the control operation in power conversion device 100B. This control operation is performed by the control device 10B in each control cycle, and the operation at the current time t(k) is shown.
Also in this case, when the control device 10B acquires each piece of information that serves as a control command, the control device 10B inputs the information to the pattern determination unit 12B and also to the storage unit 13B.

Detector 15 detects current value I(k) (Iu(k), Iv(k), Iw(k)), and control device 10B converts detected current value I(k) to one of control commands. is obtained as a part, and is input to the pattern determination part 12B and the storage part 13B (step ST1).
Next, the magnetic flux calculator 19 calculates the magnetic flux ψ(k) of the rotary electric machine 3 based on the current value I(k) (step ST2).

The speed detector 15A detects the speed ω(k), and the control device 10B acquires the detected speed ω(k) as part of the control command, and inputs it to the pattern determination unit 12B and the storage unit 13B ( step ST3). Similarly, the position detector 15B detects the position θ(k), the control device 10B acquires the detected position θ(k) as part of the control command, and stores it in the pattern determination unit 12B and the storage unit 13B. Input (step ST4).
The control device 10B acquires the calculated magnetic flux ψ(k) as a part of the control command, and inputs it to the pattern determination section 12B and the storage section 13B (step ST5).

The control device 10B acquires the speed command ωref(k) and the magnetic flux command ψref(k) as part of the control command, and inputs them to the pattern determination section 12B and the storage section 13B (step ST6).
Next, the control device 10B acquires at least one switching state SW(k-1) within the previous period (k-1) including the switching state at the current time t(k) (step ST7).
Note that steps ST1 to ST7 are not limited to the order described above.

Next, the control device 10B generates a speed command ωref(k) and a magnetic flux command ψref(k), a current value I(k), a position θ(k) and a speed ω(k), a magnetic flux ψ(k), With at least one switching state SW(k-1) in the previous cycle (k-1) as a control command, whether or not the control command is within the range of the teacher input data of the learned model MC in the storage unit 13B is determined (step ST8).

In step ST8, if the control command is within the range of the teacher input data of the learned model MC, the storage unit 13B outputs the first function F1C using the learned model MC as the pattern generation function F. Then, the pattern determination unit 12B acquires the first function F1C as the pattern generation function F (step ST9).

Then, the pattern determination unit 12B determines the switching pattern SWP(k), which is the switching state for one cycle, using the pattern generation function F based on the control command (step ST10).
Control device 10B controls the switching state of power converter 1 according to the determined switching pattern SWP(k) to drive and control rotary electric machine 3 (step ST11).

In step ST8, when the control command exceeds the range of input data for teacher of the learned model MC, the storage unit 13B outputs the second function F2B as the pattern generation function F. That is, the pattern determination unit 12B acquires the second function F2B as the pattern generation function F (step ST12). After that, the process proceeds to step ST10.

As described above, the control device 10B according to the fourth embodiment can control the driving state of the rotating electric machine 3 based on the current value I(k), the position θ(k), the magnetic flux ψ(k), and the speed ω(k). to determine the switching pattern SWP(k) for controlling the switching state of the power converter 1 . Therefore, at least one of the drive sound of the rotary electric machine 3, mechanical vibration, current harmonics, and the follow-up time of the current value to the current command value can be reduced compared to the PWM method. In addition to the current value I(k), the position θ(k) and the speed ω(k) are also detected and used. Calculations in the control device 10B are easy and highly reliable.

Furthermore, when the first function F1C using the learned model MC is used as the pattern generation function F, the control does not use the PI control, so adjustment such as parameter setting in the PI control is unnecessary, which is highly convenient.

Also, to select one pattern generation function F from the first function F1C and the second function F2B, as in the first embodiment, it is performed according to the range of the control command. The control command is compared with the teacher input data of the learned model MC, and if the control command is within the range of the teacher input data of the learned model MC, the first function F1C is selected. If the control command exceeds the range of the teacher input data of the learned model MC, the second function F2B is selected.
The comparison between the control command that is the input of the pattern generation function F and the teacher input data 45 is performed by comparing, for example, at least one of the other control commands that have the same current switching state SW in both. Also good.

Also in this embodiment, the first function F1C using the trained model MC can be applied only within the scope of the teacher data, and is not used under conditions different from the learning conditions. Therefore, the first function F1C can derive fast and highly reliable results without exceeding the expected inference.
Further, even when the control command exceeds the range of the teacher input data of the learned model MC, the switching pattern SWP(k) corresponding to the control command can be continuously generated by applying the second function F2B. The second function F2B reliably derives a result based on preset calculations.

Also, the input/output relationship of the first function F1C and the second function F2B prepared for the pattern generation function F is the same as the input/output relationship of the pattern generation function F. FIG. Therefore, the type of information input to each function (F, F1C, F2B) is common, and information that is a control command, that is, the speed command ωref(k), the magnetic flux command ψref(k), and the current value I( k), position θ(k) and velocity ω(k), magnetic flux ψ(k) and at least one switching state SW(k−1) in the previous period (k−1). The type of information output from each function (F, F1C, F2B) is also common and is the switching pattern SWP(k). However, regarding the second function F2B, it is not necessary to use all of the input control commands, and for example, the switching pattern SWP(k) is output without using the switching state SW(k-1).

Each teacher input data of the learned model MC is data input to the neural network to generate the first function F1C, and is the same kind of information as the control command input to the pattern generation function F. Similarly, each teacher label data of the learned model MC is the same type of information as the switching pattern SWP(k), which is the output of the pattern generation function F. FIG.

Also in this embodiment, the pattern determination unit 12B may extract one of the first function F1C and the second function F2B in the storage unit 13B as the pattern generation function F.
In that case, the generated control command is input only to the pattern determination unit 12B without being input to the storage unit 13B. Then, the pattern determination unit 12B extracts the pattern generation function F from the storage unit 13B based on the control command.

In the above-described embodiment, the second function F2B uses a combination of PI flux control, PI speed control, PI current control, and PWM control. can apply various controls related to the driving of the rotary electric machine 3 .
Further, by applying the second embodiment, the control device 10B may be provided with the teacher data acquisition unit 14 to acquire new teacher data.

While this application describes various exemplary embodiments and examples, various features, aspects, and functions described in one or more embodiments may not apply to particular embodiments. can be applied to the embodiments singly or in various combinations.
Therefore, countless modifications not illustrated are envisioned within the scope of the technology disclosed in the present application. For example, modification, addition or omission of at least one component, extraction of at least one component, and combination with components of other embodiments shall be included.

1 power converter, 2 DC power supply, 3 rotary electric machine, 10, 10A, 10B control device, 12, 12A, 12B pattern determination unit, 13, 13A, 13B storage unit, 14 teacher data acquisition unit, 45 teacher input data, 46 label data for teacher, 47 teacher data, 100, 100A, 100B power converter, F pattern generation function, F1, F1A, F1B, F1C first function, F2, F2A, F2B second function, Idqref dq-axis current command value , M, MA, MB, MC learned model, Q1 to Q6 switching elements, SWP switching pattern, SW switching state, Vref voltage command value.

Claims (11)

a power converter including a plurality of switching elements, converting DC power from a DC power supply into AC power and supplying the power to a rotating electric machine;
For each set control cycle, a pattern generation function is used to determine the switching states of the plurality of switching elements for one cycle, and a switching pattern composed of a combination of the switching states for the one cycle is generated. and a control device that controls the output of the power converter,
The control device prestores a first function and a second function for determining the switching pattern, generates a control command to be input to the pattern generation function, and generates the first function according to the range of the control command. , selecting and using one of the second functions as the pattern generating function,
The first function uses a learned model that performs inference based on input based on information obtained by machine learning based on teacher data, and generates the switching pattern by the inference,
The second function generates the switching pattern by a preset operation based on the input.
Power converter. The teacher data consists of teacher input data and teacher label data, the control command and the teacher input data are the same kind of information, and the switching pattern and the teacher label data are the same kind of information. is information,
The control device selects the pattern generation function by comparing the control command and the teacher input data.
The power converter according to claim 1. The control device selects the first function as the pattern generation function when the control command is within the range of the teacher input data, and when the control command exceeds the range of the teacher input data, selecting the second function as the pattern generation function;
The power converter according to claim 2. When the second function is selected and used as the pattern generation function, the control device uses the control command and the generated switching pattern as the teacher input data and the teacher label data, respectively. obtained as training data,
The power converter according to claim 2 or 3. The control device is connectable to a higher controller and transmits the new teacher data to the higher controller in response to a request.
The power converter according to claim 4. The control device acquires an output command value of the power converter for each control cycle, and at least one switching state within a previous cycle including the current switching state of the plurality of switching elements and the output command value and as the control command,
The power converter according to any one of claims 1 to 5. The output command value is a voltage command value,
The second function generates the switching pattern using pulse width modulation by comparing the voltage command value and a carrier wave.
The power converter according to claim 6. The control device acquires at least a current value as a detected value indicating the drive state of the rotating electric machine for each control cycle, and further uses the detected value as the control command.
The power converter according to claim 6. The control device includes a state observation unit that calculates a state quantity of the rotating electric machine based on the detected value, and further uses the state quantity as the control command.
The power converter according to claim 8. the second function generates the switching pattern using model predictive control based on the control command;
The power converter according to claim 9. The control device holds a plurality of functions each serving as the first function, and the trained models used in the functions are generated based on the different teacher data,
The control device selects one of the plurality of functions and the second function as the pattern generation function according to the range of the control command.
The power converter according to any one of claims 1 to 10.

Images