KR102189823B1 - Automated torque ripple reduction apparatus of motor - Google Patents

Abstract

The present invention relates to an automated device for reducing torque pulsation of an electric motor, comprising: an adaptive control unit for outputting an estimated magnetic flux and an estimated torque for estimating the torque pulsation of an electric motor, and forward-compensating the estimated magnetic flux and the estimated torque, and from a nonlinear model of the motor. And a linear model generator for generating a linear model motor through feedback linearization, and an iterative learning control unit for reducing torque pulsation of the linear model motor through iterative learning control based on the estimated magnetic flux and the estimated torque. Accordingly, the present invention can construct an electric motor control system capable of reducing unwanted torque pulsation components without explicitly taking into account changes in surrounding environments, changes in parameters, and changes in operating frequency.

Description

Automatic torque pulsation reduction device of electric motor {AUTOMATED TORQUE RIPPLE REDUCTION APPARATUS OF MOTOR}

The present invention relates to an automation technology for reducing torque pulsation of an electric motor, and more particularly, an artificial neural network and repetitive learning control are applied to compensate for the unspecified torque pulsation component of the motor and control the speed to increase the robustness of the motor. It relates to a pulsation reduction automation device.

Today's elevator control systems mostly use vector control inverters to drive high performance and high efficiency. The driving performance and ride comfort of this system is determined by the performance of the current controller to control the torque component of the motor, and the cogging torque and magnetic flux imbalance due to the structural characteristics of the motor, current measurement error of the inverter, current offset, and speed controller The vibration of the elevator is caused by the pulsation and dead time. Existing methods for suppressing the vibration of elevator systems can be largely divided into two. The first is a motor design technique to minimize torque pulsation, and the second is a method of designing a controller to minimize torque pulsation. The technique of reducing torque pulsation through motor design can be said to be a fundamental solution that minimizes various factors of cogging torque and ripple. However, as the size of the motor increases, the economical efficiency decreases, and the system load, operating speed, and natural vibration frequency It can be said that it is a solution that is not flexible and effective because it cannot fully consider the pulsation. Therefore, it can be said that using the second method of designing a controller to reduce torque pulsation is a flexible and reasonable solution.

Conventionally, in order to remove the torque pulsation, the method for removing the pulsation component present in the rotation frequency of the motor has been used the most. This method is a method of measuring the magnitude and phase of the pulsation component within the rotational frequency generated when the motor rotates, and then controlling it with the opposite phase to cancel it. The implementation of the control method is simple, but the compensation gain for vibration compensation and There was a hassle of having to measure the phase directly. In addition, when a change in operating conditions such as a change in a load or rated speed occurs, the compensator must be set again from the beginning, and due to the characteristics of the compensation technique, abnormal frequencies other than the rotation frequency of the power component cannot be controlled.

Current controllers used in industrial inverters use vector control techniques for high-performance torque control to perfectly follow the torque of an actual motor. This technique can be controlled linearly only when the motor model to be controlled completely meets the given control environment, and it receives only the current command value for torque minutes corresponding to the power frequency.It is characterized by torque pulsation and inverter due to magnetic flux imbalance. It does not completely eliminate the torque pulsation of the elevator system due to the offset and measurement errors in the current, the pulsation of the speed controller, and the dead time.

Korean Patent Registration No. 10-0671958 (2007.01.19) relates to a control device for controlling the driving of a BLDC motor. In an inverter that applies power to the BLDC motor, the current measurement value and the reference value In comparison, by connecting a follower that generates a PWM (Pulse Width Modulation) signal, it is possible to efficiently reduce the error between the control command value and the actual output value while using a microcontroller unit (MCU) with a low specification processor.

Korean Patent Registration No. 10-1254941 (2013.04.16) relates to a method for reducing longitudinal vibration of an elevator by compensating for the speed pulsation of the rotational shaft of the motor, and by modeling the eccentricity of the rotational shaft of the motor, which is the cause of vibration inside the elevator, it compensates for the speed pulsation. It is an object of the present invention to provide a method for reducing longitudinal vibration of an elevator by compensating for the pulsation of the speed of a rotating shaft of an electric motor so as to improve ride comfort by suppressing the longitudinal vibration of an elevator car caused by. To this end, the method for reducing longitudinal vibration of an elevator through compensation for motor rotation shaft speed pulsation according to the present invention is to obtain an eccentric torque (Tec), which is a torque disturbance due to eccentricity of the motor rotation shaft, and then, through this A velocity pulsation calculation step of calculating a velocity pulsation; A phase detection step of a velocity pulsation to find a phase of the velocity pulsation obtained in the velocity pulsation calculation step by adding an integral controller to the low-pass filter; A step of multiplying the motor rotation speed by a sine function, and then finding the size of the speed pulsation through a motor rotation frequency component, a motor rotation frequency doubling component, and a DC component constituting the motor rotation speed; A torque-axis current command calculation step of obtaining a torque-axis current command for forward compensation for torque pulsation using the phase and magnitude of the speed pulsation; The current command for the torque axis, which is the output of the speed controller, divided by the torque constant, is added to the current command for the torque axis of the forward compensation current controller to compensate for the torque pulsation. It consists of; a speed pulsation compensation step in which the speed pulsation is compensated.

Korean Patent Registration No. 10-0671958 (2007.01.19) Korean Patent Registration No. 10-1254941 (2013.04.16)

An embodiment of the present invention is to provide an automated device for reducing torque pulsation of a motor capable of increasing the robustness of the motor by compensating for an unspecified torque pulsation component of the motor and controlling the speed by applying an artificial neural network and repetitive learning control.

An embodiment of the present invention is to provide an automated device for reducing torque pulsation of an electric motor that automates a repetitive learning controller according to a change in an operation and a load of an electric motor using an artificial neural network.

An embodiment of the present invention is to provide an automated device for reducing torque pulsation of a motor in which a motor model is linearized using feedback linearization and an automated control model through an artificial neural network and repetitive learning control is applied to the linearized motor model.

Among the embodiments, the apparatus for reducing torque pulsation of an electric motor includes an adaptive control unit that outputs an estimated magnetic flux and an estimated torque for estimating the torque pulsation of the motor, and a linear model that generates a linear model motor through feedback linearization from the nonlinear model of the motor. And a generator and an iterative learning control unit for reducing torque pulsation of the linear model motor through iterative learning control based on the estimated magnetic flux and the estimated torque.

The adaptive control unit may generate a state error model through an error between the actual current and the estimated current, and control the estimated current to be the same as the actual current to output the estimated magnetic flux and the estimated torque.

The iterative learning control unit includes an error calculation module that calculates an error signal between the command torque and the estimated torque, an iterative learning control module that performs iterative learning control with the error signal, and a PI control that performs PI control on signals other than the error signal. May contain modules.

The iterative learning control unit may perform the iterative learning control according to the following equation.

[Equation]

Here, k is the number of repetitions, t is time, _ is the control input at the kth repetition, _ is the output error of the reference signal _ and the current output y, and Q(q) is the plant excluding the frequency section dominated by the noise component. L(q), a low-pass filter for performing learning on a frequency section where dynamic characteristics dominate, is a learning function and means a function multiplied by the error of the last repetition period.

Among embodiments, the apparatus for reducing torque pulsation of an electric motor further includes an artificial neural network unit for automating the repetitive learning control through machine learning.

The artificial neural network unit may automate the iterative learning control through error backpropagation.

The disclosed technology can have the following effects. However, since it does not mean that a specific embodiment should include all of the following effects or only the following effects, it should not be understood that the scope of the rights of the disclosed technology is limited thereby.

The apparatus for reducing torque pulsation of an electric motor according to an exemplary embodiment of the present invention may compensate for an unspecified torque pulsation component of the motor by applying an artificial neural network and repetitive learning control and control the speed to increase the robustness of the motor.

The apparatus for reducing torque pulsation of an electric motor according to an exemplary embodiment of the present invention may automate and control a speed according to an operation and load change of the electric motor using an artificial neural network.

The apparatus for reducing torque pulsation of an electric motor according to an embodiment of the present invention may linearize a motor model using feedback linearization, and apply an automated control model through an artificial neural network and repetitive learning control to the linearized motor model.

1 is a view illustrating an automated device for reducing torque pulsation of an electric motor according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating an adaptation control unit of the automatic torque pulsation reduction apparatus of the electric motor shown in FIG. 1.
3 is a block diagram illustrating a linear model generation unit of the automatic torque pulsation reduction apparatus of the electric motor in FIG. 1.
FIG. 4 is a block diagram illustrating an iterative learning control unit of the automatic torque pulsation reduction apparatus of the electric motor shown in FIG. 1.
5 is a block diagram illustrating an artificial neural network unit of the automatic torque pulsation reduction apparatus of the electric motor in FIG. 1.
6 is a flowchart illustrating a motor control process for reducing torque pulsation of an automated apparatus for reducing torque pulsation of an electric motor according to an exemplary embodiment.

Since the description of the present invention is merely an embodiment for structural or functional description, the scope of the present invention should not be construed as being limited by the embodiments described in the text. That is, since the embodiments can be variously changed and have various forms, the scope of the present invention should be understood to include equivalents capable of realizing the technical idea. In addition, since the object or effect presented in the present invention does not mean that a specific embodiment should include all of them or only those effects, the scope of the present invention should not be understood as being limited thereto.

Meanwhile, the meaning of terms described in the present application should be understood as follows.

Terms such as "first" and "second" are used to distinguish one component from other components, and the scope of rights is not limited by these terms. For example, a first component may be referred to as a second component, and similarly, a second component may be referred to as a first component.

When a component is referred to as being "connected" to another component, it should be understood that although it may be directly connected to the other component, another component may exist in the middle. On the other hand, when it is mentioned that a certain component is "directly connected" to another component, it should be understood that no other component exists in the middle. On the other hand, other expressions describing the relationship between the constituent elements, that is, "between" and "just between" or "neighboring to" and "directly neighboring to" should be interpreted as well.

Singular expressions are to be understood as including plural expressions unless the context clearly indicates otherwise, and terms such as “comprise” or “have” refer to implemented features, numbers, steps, actions, components, parts, or It is to be understood that it is intended to designate that a combination exists and does not preclude the presence or addition of one or more other features or numbers, steps, actions, components, parts, or combinations thereof.

In each step, the identification code (for example, a, b, c, etc.) is used for convenience of explanation, and the identification code does not describe the order of each step, and each step has a specific sequence clearly in context. Unless otherwise stated, it may occur differently from the stated order. That is, each of the steps may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

The present invention can be embodied as computer-readable codes on a computer-readable recording medium, and the computer-readable recording medium includes all types of recording devices storing data that can be read by a computer system. . Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. Further, the computer-readable recording medium is distributed over a computer system connected by a network, so that the computer-readable code can be stored and executed in a distributed manner.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the field to which the present invention belongs, unless otherwise defined. Terms defined in commonly used dictionaries should be construed as having meanings in the context of related technologies, and cannot be construed as having an ideal or excessive formal meaning unless explicitly defined in the present application.

The automatic torque pulsation reduction apparatus 100 of a motor simplifies the control system by transforming a model of a permanent magnet synchronous motor, which is a nonlinear system, into a linear model through algebraic calculation, and by using a linear model and a controller, the torque pulsation of an unwanted frequency A control system for reducing components is proposed. The apparatus 100 for reducing torque pulsation of an electric motor is simply expressed as two state variables of magnetic flux and torque, and the linear model obtained through the feedback linearization technique has a fast response because it directly controls the magnetic flux and torque without using a current controller. The automatic torque pulsation reduction apparatus 100 of an electric motor repeatedly learns to build a system capable of reducing unwanted torque pulsations without explicitly considering changes in surrounding environment, changes in parameters, and changes in operating frequency. A controller is used to compensate for the torque pulsation component of a frequency that the user does not want other than the command frequency. In addition, the automatic torque pulsation reduction apparatus 100 of an electric motor uses an artificial neural network to secure the shortcomings of the existing speed control system that uses explicit programming techniques that are vulnerable to changes in surrounding environment and parameters when performing speed control. Control is performed using a controller Through the automatic torque pulsation reduction apparatus 100 of the motor, a system having a fast response with robustness to parameter changes, load changes, and disturbances can be implemented, and an unknown, predicted value that cannot be considered in the existing torque pulsation compensation system. Disturbances that cannot be disturbed can be compensated for with a simple control system.

1 is a view illustrating an automated device for reducing torque pulsation of an electric motor according to an embodiment of the present invention.

Referring to FIG. 1, the apparatus 100 for reducing torque pulsation of an electric motor includes a driving unit 10, an adaptive control unit 110, a linear model generation unit 120, an iterative learning control unit 130, and an artificial neural network unit 140. Include.

The driving unit 10 may include an inverter and an electric motor. Here, the inverter is a type of power conversion device, and refers to a device that converts DC power into AC power. Here, an electric motor is a device that converts electrical energy into mechanical energy by using the force received by a conductor through which an electric current flows in a magnetic field. In general, it may mean a motor. In one embodiment, the electric motor may be implemented as a permanent magnet synchronous motor (PMSM), and may be implemented as an elevator motor. In one embodiment, the driving unit 10 may include a space vector pulse width modulation (SVPWM) and a d-q conversion (inverse conversion) device. Here, SVPWM corresponds to a space vector PWM method as a method of PWM (pulse width modulation). PWM is a means of digitalizing an analog amount, and it samples an analog signal at a certain period and converts it into a pulse width proportional to the value. The inverter has a plurality of switching elements and can convert a DC voltage into a three-phase AC voltage by switching on and off of the switching element according to a PWM signal input from the SVPWM. Here, the dq conversion device may correspond to a device that converts a coordinate system of a three-phase circuit into a d-q axis rectangular coordinate system. The transformation of the d-q-axis Cartesian coordinate system can be expressed in two ways. It can be divided into a fixed coordinate system (or stator coordinate system) in which the axis is fixed and a rotation coordinate system. The drive unit 10 can express the physical quantity of the motor through two variables by converting (or inversely transforming) the coordinate system expressed in three phases to the dq-axis rectangular coordinate system through the dq conversion (inverse conversion) device, and can perform instantaneous control. have. For example, the d-axis may correspond to an axis where a magnetic flux of a motor is generated and may correspond to an axis that is a reference in vector control. The q-axis is an axis orthogonal to the d-axis, and may correspond to an axis that controls a current that generates torque in vector control.

The adaptive control unit 110 may output an estimated magnetic flux and an estimated torque for estimating the torque pulsation of the electric motor. In an embodiment, the adaptive control unit 110 may generate a state error model through an error between the actual current and the estimated current, and control the estimated current to be the same as the actual current to output the estimated magnetic flux and the estimated torque. For example, the adaptive control unit 110 may calculate the estimated magnetic flux and the estimated torque through Equation 1.

[Equation 1]

는 추정토크,

는 추정자속,

는 d축 쇄교자속,

는 q축 쇄교자속,

는 영구자석 자속,

는 회전자 좌표계의 고정자 d축 전류,

는 회전자 좌표계의 고정자 q축 전류, p는 미분 연산자를 의미할 수 있다.From here,

Is the estimated torque,

Is the estimated magnetic flux,

Is the d-axis flux linkage,

Is the q-axis flux linkage,

Is the permanent magnet magnetic flux,

Is the stator d-axis current of the rotor coordinate system,

Denotes the stator q-axis current of the rotor coordinate system, and p denotes a differential operator.

The linear model generator 120 may forward-compensate the estimated magnetic flux and the estimated torque, and generate a linear model motor through feedback linearization from a nonlinear model of the motor. Here, forward compensation is an active compensation method and may mean taking necessary corrections before affecting the output when a disturbance occurs in the control system. The control method through forward compensation can be referred to as feedforward control. In an embodiment, the linear model generator 120 may forward-compensate the estimated magnetic flux and the estimated torque previously set in the controller to prepare for the disturbance before the disturbance occurs.

The iterative learning control unit 130 may reduce the torque pulsation of the linear model motor through the iterative learning control based on the estimated magnetic flux and the estimated torque. In one embodiment, the iterative learning control unit 130 may be configured as an iterative learning controller (ILC). The iterative learning controller can reduce torque pulsation through the iterative learning control. Here, the iterative learning control (ILC) can be used to improve the reference signal tracking performance in a system that performs an operation that is repeated for a predetermined time interval. In order to apply iterative learning control, the value of the initial state should be the same for each repetition period, and the system should be performed by a reference signal that performs repetitive tasks for a fixed period of time.

The artificial neural network unit 140 may automate control of the motor speed based on the operation of the motor and a change in load through machine learning. In an embodiment, the artificial neural network unit 140 may automate the control of the motor speed through error backpropagation. For example, the artificial neural network unit 140 may determine a command torque for controlling the speed of the electric motor as necessary. Here, the error backpropagation method is used as a method of adjusting individual weights so that a desired value is output for the same input layer, and has an advantage of obtaining a stable result although the speed is slow.

FIG. 2 is a block diagram illustrating an adaptation control unit of the automatic torque pulsation reduction apparatus of the electric motor in FIG. 1.

In FIG. 2, the adaptive control unit 110 includes an actual current calculation model 112 for calculating the actual current, an estimated current calculation model 114 for calculating the estimated current, and the actual current and the It may include an adaptive algorithm model 116 that controls the estimated current to be the same. In one embodiment, the adaptive control unit 110 receives the actual current through the current sensor between the motor and the inverter, the estimated current through the control of the linear model generation unit 120 and the iterative learning control unit 130, and You can enter the angular velocity. More specifically, the adaptive control unit 110 may output the estimated magnetic flux and the estimated torque by controlling the addition or reduction of the actual current and the estimated current and correcting the error.

3 is a block diagram illustrating a linear model generation unit of the automatic torque pulsation reduction apparatus of the electric motor in FIG. 1.

In an embodiment, the linear model generator 120 may generate a linear model of an electric motor through Equations 2 to 7. More specifically, the voltage equation of the surface-attached permanent magnet synchronous motor (SPMSM) of the rotor reference coordinate system in a steady state is shown in Equation 2.

[Equation 2]

는 미분 연산자,

는 고정자 저항,

는 전체 고정자 인덕턴스,

는 영구자석 자속,

는 회전자 좌표계의 고정자 d축 전류,

는 회전자 좌표계의 고정자 q축 전류,

는 회전자 좌표계의 고정자 d축 전압,

는 회전자 좌표계의 고정자 q축 전압,

는 영구자석 동기기 전기적 각속도,

은 전동기의 각속도를 의미할 수 있다. From here,

Is the differential operator,

Is the stator resistance,

Is the total stator inductance,

Is the permanent magnet magnetic flux,

Is the stator d-axis current of the rotor coordinate system,

Is the stator q-axis current of the rotor coordinate system,

Is the stator d-axis voltage of the rotor coordinate system,

Is the stator q-axis voltage of the rotor coordinate system,

Is the electrical angular velocity of the permanent magnet synchronous machine,

Can mean the angular speed of the motor.

In an embodiment, Equation 2 is the same as Equation 3 when expressed in a vector form, and d and q-axis magnetic fluxes in Equation 3 can be expressed as Equation 4.

[Equation 3]

[Equation 4]

이며, 고정자 전류 벡터는

이며, 입력으로써 고정자 전압 벡터는

이다. 또한,

은 영구자석 자속이며,

은 회전자 좌표계의 영구자석 동기기 전기적 각속도,

는 고정자 저항,

는 비돌극 모터의 고정자 인덕턴스를 의미할 수 있다.Here, the stator flux vector is

And the stator current vector is

And the stator voltage vector as input is

to be. Also,

Is the permanent magnet magnetic flux,

Is the electrical angular velocity of the permanent magnet synchronizer in the rotor coordinate system,

Is the stator resistance,

May mean the stator inductance of the non-protruding motor.

In one embodiment, the electric torque is _ and the square value of the magnitude of the stator magnetic flux is equal to Equation 1. If Equation 1 is differentiated and arranged, Equation 5 can be obtained.

[Equation 5]

In order to linearize Equation 5, Equation 6 can be obtained by arranging the input as W = Wd + jWq.

[Equation 6]

는 선형시스템에서 제어되는 쇄교자속,

는 선형시스템에서 제어되는 토크를 의미할 수 있다.From here,

Is the flux linkage controlled in the linear system,

May mean the torque controlled in the linear system.

Finally, by substituting Equation 5 into Equation 6, a linear model in the form of Equation 7 can be obtained. Equation 7 may correspond to a linear model because the time-varying factor is not included in the right-hand term.

[Equation 7]

4 is a block diagram illustrating an iterative learning control unit of the automatic torque pulsation reduction apparatus of the electric motor shown in FIG. 1. 4(a) is a block diagram showing an iterative learning controller for reducing torque pulsation, and FIG. 4(b) is a block diagram of a general p-type iterative learning control system.

In FIG. 4(a), the iterative learning control unit 130 includes an error calculation module 132 for calculating an error signal between the command torque and the estimated torque, and an iterative learning control module 134 for performing iterative learning control with the error signal. And a PI control module 136 that performs proportional integral (PI) control on signals other than the error signal. More specifically, the iterative learning control unit 130 controls to reduce the torque pulsation component of a frequency that the user does not want through the iterative learning control module 134, and PI for command frequency components corresponding to signals other than the error signal. PI control may be performed through the control module 136.

In an embodiment, the iterative learning control unit 130 may perform the iterative learning control through Equation 8.

[Equation 8]

Here, γ may be a learning parameter, k may be the number of repetitions, t may be time, and uk may denote the ILC control input at the kth repetition time.

In another embodiment, the iterative learning controller 130 adds a Q-filter defined as a low-pass filter to Equation 8 to satisfy stability in the iteration domain, and a general learning gain L(q) than a learning parameter Iterative learning control can be performed through Equation 9 changed to.

[Equation 9]

는 k번째 반복 시 제어 입력,

는 기준 신호

와 현재 출력 y의 출력오차, Q(q)는 노이즈 성분이 지배적인 주파수 구간을 제외하고 플랜트 동특성이 지배적인 주파수 구간에 대하여 학습을 수행하도록 하기 위한 저역 통과 필터, L(q)는 학습함수로써 지난 반복시기의 오차에 곱해지는 함수를 의미할 수 있다. Where k is the number of iterations, t is the time,

Is the control input at the kth iteration,

Is the reference signal

And the output error of the current output y, Q(q) is a low-pass filter to perform learning on the frequency section where the plant dynamics dominates excluding the frequency section where the noise component is dominant, and L(q) is a learning function. It can mean a function that is multiplied by the error of the last iteration period.

In one embodiment, the iterative learning control unit 130 may determine the convergence condition and convergence speed of the error according to the determination of the L(q) value, and in each item to store the error value and the control input value of the last iteration time. As for the number of samples in the repeated profile, memory may be required.

5 is a block diagram illustrating an artificial neural network unit of the automatic torque pulsation reduction apparatus of the electric motor shown in FIG. 1.

In FIG. 5, the artificial neural network unit 140 controls the output through an error backpropagation method to obtain a desired output (eg, speed) for a given input of the neural network for supervised learning. I can. Here, supervised learning may correspond to a method of machine learning to infer a function from training data. For example, a supervised learner can correctly guess the value it wants to predict for a given data from the training data. More specifically, when an input pattern is given to each unit of the input layer, the artificial neural network unit 140 outputs an output layer signal through a weight and an activation function between each layer, and backpropagates at the upper layer, By adjusting the weight, the weight of the connection node can be adjusted so that the output pattern and the destination pattern are the same. The artificial neural network unit 140 may perform learning to output a desired value by repeating this process.

In an embodiment, the artificial neural network unit 140 may control a command torque capable of obtaining a target speed through Equations 10 to 13. The artificial neural network unit 140 can obtain a discrete time model of a single-input and single output system by arranging the electric equation, mechanical equation, torque and current equation of the motor as shown in Equation 10.

[Equation 10]

Here, the q-axis current can be expressed as Equation 11 by v_q^r (n) and ω_r (n) by discretizing the q-axis voltage equation.

[Equation 11]

Equation 11 can be obtained by using Equation 10 after arranging v_q^r (n).

[Equation 12]

Here, the torque can be expressed by Equation 13 with reference to Equation 12.

[Equation 13]

In one embodiment, the artificial neural network unit 140 may be implemented as an artificial neural network having three layers of an input layer, a hidden layer, and an output layer.

는 각각의 입력에 대한 (n-1)번째 층의 i번째와, j번째 요소들로 구성된 뉴런 사이의 가중치이다.

는 n번째 층의 j번째 요소로 구성된 바이어스이며,

은 (n-1)번째 층의 바이어스, 가중치 출력의 합을 더한것이며,

는 n번째 층의 i번째 신경망의 출력값,

는 인공 신경망의 출력,

은 n번째 층의 활성화 함수,

는 지령값에 해당할 수 있다.In Figure 5

Is the weight between the neurons composed of the i-th and j-th elements of the (n-1)th layer for each input.

Is the bias composed of the j-th element of the n-th layer,

Is the sum of the bias and weight outputs of the (n-1)th layer,

Is the output of the ith neural network of the nth layer,

Is the output of the artificial neural network,

Is the activation function of the nth layer,

May correspond to the command value.

That is, the components of the neural network may be summed with respect to the input pattern in consideration of the weight of the connection line as in Equation 14.

[Equation 14]

)Then, a signal is output to the connected node using an appropriate activation function (Equation 15). Several activation functions may be used, and the artificial neural network unit 140 may use a logistic sigmoid (f1) and a hyperbolic tangent function (f2) as activation functions of the hidden layer and the output layer, respectively. (

)

[Equation 15]

The artificial neural network unit 140 sets a cost function through Equation 16 from the difference between the neural network and the command value, and estimates the weight with a value at which the cost function is minimum using a gradient descent method. The backpropagation method of Equation 17 can be used.

[Equation 16]

[Equation 17]

6 is a flowchart illustrating a motor control process for reducing torque pulsation of an automated apparatus for reducing torque pulsation of an electric motor according to an exemplary embodiment.

In FIG. 6, the apparatus 100 for reducing torque pulsation of an electric motor may output an estimated magnetic flux and an estimated torque for estimating the torque pulsation of the motor through the adaptive control unit 110 (step S610).

The automatic torque pulsation reduction apparatus 100 of the motor may forward-compensate the estimated magnetic flux and the estimated torque through the linear model generation unit 120 and generate a linear model motor through feedback linearization from a nonlinear model of the motor (step S620). .

The automatic torque pulsation reduction apparatus 100 of the motor may reduce the torque pulsation of the linear model motor through iterative learning control based on the estimated magnetic flux and the estimated torque through the iterative learning control unit 130 (step S630).

Although the above has been described with reference to preferred embodiments of the present invention, those skilled in the art will variously modify and change the present invention within the scope not departing from the spirit and scope of the present invention described in the following claims. You will understand that you can do it.

100: motor torque pulsation reduction automation device
10: drive unit
110: adaptive control unit 112: actual current calculation model
114: estimated current calculation model 116: adaptive algorithm model
120: linear model generation unit
130: iterative learning control unit
132: error calculation module 134: iterative learning control module
136: PI control module
140: artificial neural network

Claims (6)

An adaptive control unit for outputting an estimated magnetic flux and an estimated torque for estimating the torque pulsation of the motor;
A linear model generator for forward-compensating the estimated magnetic flux and the estimated torque, and generating a linear model motor through feedback linearization from the nonlinear model of the motor; And
It includes an iterative learning control unit for reducing the torque pulsation of the linear model motor through iterative learning control based on the estimated magnetic flux and the estimated torque,
The iterative learning control unit is an error calculation module that calculates an error signal between a command torque and an estimated torque, an iterative learning control module that performs the iterative learning control with the error signal, and PI (Proportional Integral) control for signals other than the error signal An automated apparatus for reducing torque pulsation of an electric motor, comprising: a PI control module for performing the.
The method of claim 1, wherein the adaptation control unit
An automated apparatus for reducing torque pulsation of an electric motor, characterized in that for generating a state error model through an error between an actual current and an estimated current, controlling the estimated current to be equal to the actual current, and outputting the estimated magnetic flux and the estimated torque.
delete The method of claim 1, wherein the iterative learning control unit
An automated apparatus for reducing torque pulsation of an electric motor, characterized in that the iterative learning control is performed according to Equation 9 below.
[Equation 9]

Where k is the number of iterations, t is the time,

Is the control input at the kth iteration,

Is the reference signal

And the output error of the current output y, Q(q) is a low-pass filter to perform learning on the frequency section where the plant dynamics dominates excluding the frequency section where the noise component is dominant, and L(q) is a learning function. It means a function that is multiplied by the error of the last iteration period.
The method of claim 1,
An automatic apparatus for reducing torque pulsation of an electric motor, further comprising an artificial neural network unit for automating the speed control of the electric motor based on the operation and load change of the electric motor through machine learning.
The method of claim 5, wherein the artificial neural network
An automated apparatus for reducing torque pulsation of an electric motor, characterized in that the speed control of the electric motor is automated through error backpropagation.

Images