KR101122673B1 - Method for speed control of ac motor using the radial basis function network observer - Google Patents

Abstract

The present invention relates to an intelligent speed control system of an induction motor using a radial basis function network (RBFN) observer. It is to build a robust high performance, low cost induction motor servo system. In practical induction motor systems, system-specific uncertainties necessarily exist. In order to overcome the inherent uncertainties of these parameters, such as disturbances and dynamic factors not included in the design, in the present invention, RBFN was used as an uncertainty observer to approximate the inherent uncertainty, and the structural error of the system An additional robust control term has been added in place of the large number of rules and additional update parameters in the RBFN to compensate for this problem. Therefore, the induction motor without speed sensor is designed to have Lyapunov stability in the whole closed loop system. Accordingly, the present invention overcomes the inherent uncertainty inherent in low and low cost induction motors without a speed sensor to reduce the error of the output to the input.

Induction Motor, Radial Basis Function Network (RBFN), Speed Control System, Uncertainty Observer, Adaptive Control, Liafnov Stability, Direct Torque Control (DTC)

Description

METHOD FOR SPEED CONTROL OF AC MOTOR USING THE RADIAL BASIS FUNCTION NETWORK OBSERVER}             

1 illustrates the structure of a typical RBFN uncertainty observer.

2 is a block diagram of a basic direct torque control system for a three-level inverter system.

3 shows a simplified structure of a general direct torque control system.

4 is a block diagram for implementing an induction motor control system according to a preferred embodiment of the present invention.

5 is a block diagram showing the overall block structure of the induction motor control system according to a preferred embodiment of the present invention.

6 is a view showing a control result in an induction motor using a conventional IP controller.

7 is a view showing a control result in an induction motor using another RBFN in a preferred embodiment of the present invention.

The present invention relates to an induction motor, and more particularly, to effectively control an induction motor using a radial basis function network (hereinafter referred to as 'RBFN') as an operation error observer.

When constructing a servo control system without a speed sensor of an induction motor, if a proper assumption is used, a speed control loop having a desired performance can be designed relatively simply using only the proportional integral and integral proportional design method using the mechanical equation of the motor. However, this simple approach has the disadvantage of showing sensitive characteristics depending on the uncertainty of the model parameters and unknown load conditions. In addition, in the induction motor system without the actual speed sensor, the speed estimation error is necessarily present. Therefore, the effect analysis on the generated error must be considered in the stability analysis viewpoint of the entire closed loop control system.

If accurate mathematical modeling of the inherent uncertainty of the control system is possible, robust control characteristics of the controller and estimator can be obtained, but it is also difficult to obtain accurate modeling information due to the uncertainty and the strong nonlinearity of the uncertainty. Therefore, recently, various types of intelligent approaches such as adaptive purge logic, fuzzy neural network, recurrent purge neural network, and the like have been actively used for high power induction motor control.

As an example, R.J Wai applied a recurrent fuzzy neural network for on-line estimation of system sharing uncertainty, and applied to speed controller design problem of induction motor, and obtained better control performance than integrative speed controller. However, the complex structure and large number of parameters and design constants of the recurrent fuzzy neural networks used caused computational complexity. Moreover, the design of the controller requires accurate intuition of the system dynamics, and the unknown parameter set through trial and error of the system designer makes the design more difficult. In another study conducted by B.K Bose, a simple static fuzzy logic was used to design the speed and torque controllers. However, due to the static mapping of the fuzzy logic used, there is no ability to adapt adaptively to changes in the environment of the system and thus cannot guarantee robustness and stability. Moreover, designing a static fuzzy system requires expert knowledge, and therefore, accurate knowledge of the system under control is required for effective controller design.

In the control of the induction motor by the conventional PI and PI technique, which is operated as described above, it is unavoidable in the process of converting electromagnetic energy into mechanical energy. Nonlinearity of the model (5th order nonlinear differential equations) and multivariable system (multiple currents as input variables and torques, velocity and flux as output variables), as well as changing system parameters (severely affected by operating temperature) Unknown load torque is caused by electronic resistance, coefficient of friction and inductance depending on the magnitude of the magnetic flux, and the variable behavior of the non-modeled system and the measurement of partial state variables (measurable rotor flux Present), and in the induction motor control by the intelligent control system, fuzzy neural network or other intelligent The type control algorithm is complicated and requires accurate expert knowledge for system design.

Accordingly, an object of the present invention, which was devised to solve the problems of the prior art operating as described above, is to provide a low-cost induction motor system having the advantages of the existing high priced high-performance induction motor.

Another object of the present invention is to effectively control an induction motor that exhibits nonlinear characteristics by using RBFN as an operation error observer for overcoming system uncertainty in controlling a induction motor that is not equipped with a low-cost speed sensor and for stable system operation. It is for.

Another object of the present invention is to use RBFN as an approximator to model unknown states and uncertainties, and to implement robust speed controllers in induction motor systems with strong nonlinearity that are inaccessible by classical control techniques. will be.

In order to achieve the above object, an embodiment of the present invention provides a method of controlling a speed of an induction motor using a radial basis function network observer, wherein the input torque value is obtained through a robust and stable speed controller and a torque comparator. Comparing the, and the process of compensating for the error with the input value, the process of comparing the input magnetic flux value in the magnetic flux comparator, and generating a gating signal from the torque comparator and the output values from the magnetic flux comparators Transferring the three-level inverter to the adaptive observer; and outputting the inverter value to the adaptive observer, outputting an approximation of the input value, and controlling the induction motor through the inverter value. It is done.

Hereinafter, the operating principle of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily flow the gist of the present invention, the detailed description thereof will be omitted. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intentions or customs of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

1 is a diagram showing the structure of a typical RBFN uncertainty observer.

RBFN is a design tool for the instar-outstar model and consists of an input layer 110, an output 130, and a hidden layer 120 of standardized Gaussian activation functions. RBFN is based on a partially tuned, redundantly accepted layer structure. A schematic structure of a simple type of RBFN consisting of one input 110, one output 130, and one hidden layer 120 is shown in FIG. 1.

The linear combination of the variable values transmitted from the input 110 is processed by the non-linear function in the hidden layer 120 and transferred to the output 130 to output the value. Here, the hidden layer 120 nodes of the RBFN normalize the Gaussian activation functions as shown in Equation (1).

는 중심,

는 q번째 가우시안 함수의 폭.x is the input vector,

Is the center,

Is the width of the qth Gaussian function.

은

로 중심화된 층이다. 그리고

는 q번째 가우시안 함수의 변화이다. 그러므로 은닉층 노드q는 입력 벡터들에게

와 비슷한, 큰 응답 값을 준다. RBFN의 출력은 단순하게 은닉층 노드출력의 가중화된 합이다. 그리고 본 발명에서 은닉층 출력 선형 조합의 단순한 형태를 이용하여 RBFN을 불확실 관측기로서 사용한다. (수학식 2)는 이러한 은닉층 출력의 선형 조합을 보이고 있다.For each q th hidden node, its receptive field,

silver

As the centered layer. And

Is the change of the qth Gaussian function. Therefore, the hidden layer node q

Similarly, it gives a large response. The output of the RBFN is simply the weighted sum of the hidden layer node outputs. In the present invention, RBFN is used as an uncertain observer using a simple form of hidden layer output linear combination. Equation (2) shows a linear combination of these hidden layer outputs.

는 i번째 노드와 RBFN 출력 사이의 가중치이다. 그리고

은

's의 벡터이다.

Is the weight between the i th node and the RBFN output. And

silver

is the vector of 's.

FIG. 2 is a diagram illustrating the structure of a basic direct torque control (DTC) system for a three-level inverter system.

The induction motor DTC system of FIG. 2 includes an integral proportional (IP) controller 205, a torque comparator 210 and a flux comparator 215, a switching logic generator 220, an adaptive observer 225, and a three-level inverter system 230. ) Is included.

The torque value is subtracted from the value from the adaptive observer 225 via the IP controller 210 to compare the value in the torque comparator 210. Here, the IP controller 210 has an overshoot that is small so that the step tracking response is negligible compared to the PI control and the zero steady state error and has good adjustment characteristics. In addition, the magnetic flux value is subtracted from the value from the adaptive observer 225 to compare the value in the magnetic flux comparator 215, the values output from the comparators (210 to 215) through the switching logic generator 220 through the gating signal ( Gating signal is generated, and the gating signal is transmitted to the three-level inverter 230 and the adaptive observer 225. The three-level inverter 230 controls the induction motor 235 according to the gating signal value. The observation value is generated and output from the gating signal transmitted to the adaptive observer 225 and the output values of the three-level inverter 230, subtracted from the input torque value and the magnetic flux value, and the value is converted to the switching logic generator 220. To pass.

In FIG. 2, an IP controller 210 having an RBFN uncertainty observer is used to stabilize the speed control loop and to induce the parameters Ki and Kp of the controller to obtain the required control performance. Applying some reasonable assumptions to the nominal model dynamics to ensure that rotor speed, flux and torque measurements and regulation of components work perfectly, the schematic diagram of the DTC system is shown in FIG. It can be expressed briefly as follows.

3 is a schematic structural diagram of a general direct torque control system.

(315)를 통해 적응 토크값을 출력하여 입력 토크값과 감산한다. The torque value is output through the IP controller 305 is passed through the torque constant K (310), the T disturbance value by adding a gating signal to generate a three-level inverter

The adaptive torque value is outputted through 315 and subtracted from the input torque value.

를

가 따라가도록 하는 것이다. The control objective of the system is a limited speed given under inherent uncertainty with the constraint that all signals in the Peruvian system must be in a specific area.

To

To follow.

In FIG. 3, a simple mechanical model equation can be obtained by the following Equation 3.

Equation (3) can be written in the form of appropriate variables.

Although expressed as an initial value in Equation (4), inherent uncertainty exists in the induction motor model in most practical cases. It is therefore reasonable to include the inherent uncertainty in Equation 4 above, and this results in Equation 5 being obtained.

는

의 모델링 에러,

는 제어 입력

는 모델링되지 않은 불확실성이다. here,

Is

Modeling error,

Control input

Is the unmodeled uncertainty.

Equation 5 may be expressed again in the form of Equation 6.

)을 안다면, 상기 (수학식 6)에서 표현된 폐루프 시스템에 대한 완벽한 제어 입력이 점근적으로 안정해짐을 (수학식 7)과 같이 계산 할 수 있다. If the exact inherent uncertainty (

), It can be calculated as Equation (7) that the complete control input to the closed loop system represented by Equation (6) becomes asymptotically stable.

는 시스템을 안정화시키기 위해 설계된 상수이다.here

Is a constant designed to stabilize the system.

Substituting Equation 7 into Equation 6, an error dynamic condition is obtained as shown in Equation 8 below.

이다. Equation (8) means that the control input represented by Equation (7) induces the entire Perup system to be stable with an appropriate design constant Kx. In other words,

to be.

In sensorless speed control using RBFN, the predicted rotor speed is fed back to generate a control signal instead of the sensing value. Using the predicted velocity, the state equation (Equation 6) can be rewritten as (Equation 9).

그리고

는 전체 제어 입력이다.here,

And

Is the full control input.

와 알려지지 않은 회전자속도 에러

는 (수학식 10)과 같이 정의된다.For most control systems it is assumed that the predicted speed and the actual speed are the same. However, in real systems, prediction errors exist. Rotor speed error measured by speed meter

And unknown rotor speed error

Is defined as (10).

와

는 급격히 감소하는 예상에러다.here,

Wow

Is an unexpected decline.

와 전체 불확실성을 나타내는 변수에 대한 가정이 (수학식 11)과 같이 얻어진다.remind

And assumptions about the variables representing the total uncertainty are obtained as shown in Equation (11).

는 유한한 양의 상수이다,

,

은 전체 불확실성으로 최적화되고 유도되었다.

은 유한한 양의 상수이다.here,

Is a finite positive constant,

,

Is optimized and derived from the overall uncertainty.

Is a finite positive constant.

The DTC machine structure allows for fast torque response with increasing robustness to variations in variables. In addition, switching voltage selection and torque ripple reduction algorithms for three-level inverter systems in high voltage applications can reliably solve large torque fluctuations and leakage element problems in the low speed range. However, the speed control behavior is still influenced by closed system-specific uncertainties, such as parameter uncertainties, external load disturbances and unmodeled mechanical dynamics problems.

을 RBFN의 출력

으로 바꾸고, 근사화 방법을 이용해서, 모든 신호들은 한결같이 일정하다는 가정 하에서 RBFN의 변수들과 일정한 상수들을 조정할 적절한 법칙과 추적 에러를 0으로 만들기 위한 제 어 법칙들을 유도 할 수 있다.Unknown uncertainty

Output of RBFN

By using the approximation method, we can derive the appropriate law to adjust the RBFN's variables and constants, and the control laws to zero the tracking error, assuming all signals are uniformly constant.

의 출력을 이용해서 (수학식 9)의 역방향 힘의 제어 입력을 (수학식 12)와 같이 표현될 수 있다.RBFN,

Using the output of Equation 9, the control input of the reverse force of Equation 9 can be expressed as Equation 12.

라면, 상기 (수학식 12)의 제어 입력은 전체 시스템을 점근적으로 안정화 시킬 수 있다.If a universal solution completely uncovers an unknown uncertainty, for example

In this case, the control input of Equation 12 may stabilize the whole system gradually.

In practice, however, structural errors are inevitable. So additional compensation control is required.

는 W를 위한 상수 집합,

설계자에 의해 구체화된 양의 상수. 그리고

는

과 같은 특정 제어가 가능한 영역이다.here,

Is a set of constants for W,

Positive constant specified by the designer. And

Is

This is an area where specific control is possible.

을 포함한

로 나타내면 (수학식 14)와 같다.Full control input

Including

It is represented by Equation (14).

는 상기 (수학식 12)에서 결정되어진다. 그리고

는 RBFN의 구조적 오차를 보상하기 위한 추가 제어 입력이다.here,

Is determined by Equation (12). And

Is an additional control input to compensate for the structural error of the RBFN.

은 (수학식 15)와 같이 예측 할 수 있다.We also determine additional rules for the weight W of RBFN. And constant

Can be predicted as (Equation 15).

In Equation 15, the tracking error and other signals related to the closed loop system are finally integrated.

을 위한 적합한 법칙의 유도와 보상을 위한 제어 입력,

를 위한 적응 규칙을 도출하기 위해 리아프노프(Lyapunov) 함수를 (수학식 16)로 정의 한다.Unknown constant limits,

Control input for derivation and compensation of suitable law for

In order to derive the adaptation rule for, Lyapunov function is defined as (Equation 16).

는 미지의 값이므로 다음의 (수학식 17)이 요구된다.In Equation 16,

Since is an unknown value, the following Equation 17 is required.

The time derivative of Equation 17 is expressed as Equation 18.

라 하고, 리아프노프 방정식의 시간 미분을 취하면 (수학식 19)와 같다.

If we take the time derivative of the Liafnov equation, we get

Then, the following equation can be derived from Equation 14 and Equation 15.

의 시간 도 함수(Time derivative)는 (수학식 21)과 같이 나타낼 수 있다.And,

The time derivative of can be expressed as (Equation 21).

와

는 다음의 (수학식 22)와 같이 정의된다.a constant

Wow

Is defined as Equation 22 below.

,

, 과

을 위한 상하한(upper bounds)은 (수학식 23)과 같이 유도될 수 있다.(20) and the direct method of Liafnov,

,

, And

The upper bounds for may be derived as shown in Equation 23.

This indicates that all signals are Unified Ultimately Bounded.

5 is a view showing the entire block structure of the induction motor control system according to a preferred embodiment of the present invention.

The induction motor DTC system of FIG. 5 includes a robust and stable speed controller 505, a torque comparator 210, a flux comparator 215, a switching logic generator 220, an adaptive observer 225, and 3. A level inverter system 230.

)을 감산한 값이다. 방사형 기저 함수 네트워크 관측기는 토크 제어값(e_x)을 이용하여 가중 제어값((1-z^-1)W) 및 상수 제한 제어값((1-z^-1)ζ)을 생성한다. 방사형 기저 함수 네트워크 관측기는 생성한 가중 제어값((1-z^-1)W) 및 상수 제한 제어값((1-z^-1)ζ)을 적응 법칙 제어 블록으로 출력한다.
방사형 기저 함수 네트워크 관측기는 적응 법칙 제어 블록으로부터 가중치(W) 및 상수 제한값(ζ)을 입력 받는다. 방사형 기저 함수 네트워크 관측기는 입력 받은 가중치(W) 및 상수 제한값(ζ)을 이용하여 함수 출력값(ε)을 생성한다. 방사형 기저 함수 네트워크 관측기는 생성한 함수 출력값(ε)을 속도 제어 블록으로 출력한다.
적응 법칙 제어 블록은 방사형 기저 함수 네트워크 관측기로부터 가중 제어값((1-z^-1)W) 및 상수 제한 제어값((1-z^-1)ζ)을 입력 받는다. 또한, 적응 법칙 제어 블록은 함수 제어값(e, σ₁, σ₂)을 입력 받는다. 적응 법칙 제어 블록은 함수 제어값(e, σ₁, σ₂), 방사형 기저 함수 네트워크 관측기로부터 입력 받은 가중 제어값((1-z^-1)W) 및 상수 제한 제어값((1-z^-1)ζ)을 이용하여 가중치(W). 상수 제한값(ζ) 및 오차 보상 제어 입력값(U_p)을 생성한다. 적응 법칙 제어 블록은 생성한 가중치(W) 및 상수 제한값(ζ)을 방사형 기저 함수 네트워크 관측기로 출력한다. 적응 법칙 제어 블록은 생성한 오차 보상 제어 입력값(U_p)을 합산기로 출력한다.
속도 제어 블록은 토크 제어값(e_x)을 입력 받고, 방사형 기저 함수 네트워크 관측기로부터 함수 출력값(ε)을 입력 받는다. 속도 제어 블록은 입력 받은 토크 제어값(e_x)과 함수 출력값을 이용하여 역방향 힘 제어 입력값(U_n)을 생성한다. 속도 제어 블록은 생성한 역방향 힘 제어 입력값(U_n)을 합산기로 출력한다.Here, the speed controller 505 uses an algorithm of torque reduction and voltage vector selection for a three-level inverter system and an adaptive observation concept for speed estimation proposed by Kubota, which is commercially preferred. Determine the torque.
To this end, the speed controller 505 includes a radial basis function network observer (RBFN observer), adaptive laws control block, speed controller block and summer.
The radial basis function network observer receives the torque control value e _x . At this time, the torque control value is the first feedback value (inputted from the adaptive observer at the input value ω _{r_ref} ) for controlling the speed of the induction motor 235 (

) Is subtracted. The radial basis function network observer uses the torque control value e _x to generate a weighted control value ((1-z ⁻¹ ) W) and a constant limit control value ((1-z ⁻¹ ) ζ). The radial basis function network observer outputs the generated weighting control value ((1-z ^-1 ) W) and the constant limiting control value ((1-z ^-1 ) ζ) to the adaptive law control block.
The radial basis function network observer receives a weight (W) and a constant limit (ζ) from the adaptive law control block. The radial basis function network observer generates a function output value ε using the input weight W and the constant limit value ζ. The radial basis function network observer outputs the generated function output value epsilon to the speed control block.
The adaptive law control block receives a weighted control value ((1-z ^-1 ) W) and a constant limiting control value ((1-z ^-1 ) ζ) from the radial basis function network observer. In addition, the adaptive law control block receives the function control values e, σ ₁ , and σ ₂ . The adaptive law control block consists of a function control value (e, σ ₁ , σ ₂ ), a weighted control value ((1-z ^-1 ) W) input from a radial basis function network observer, and a constant limit control value ((1-z- ^{). 1} ) ζ) using the weight (W). A constant limit value ζ and an error compensation control input value U _p are generated. The adaptive law control block outputs the generated weight W and the constant limit ζ to the radial basis function network observer. The adaptive law control block outputs the generated error compensation control input value U _p to a summer.
The speed control block receives a torque control value e _x and a function output value ε from the radial basis function network observer. The speed control block generates the reverse force control input value U _n by using the received torque control value e _x and the function output value. The speed control block outputs the generated reverse force control input value U _n to the summer.

)을 감산한 토크값을 입력 받는다. 토크 비교기(210)는 토크값과 토크 기준값을 비교하여 토크 비교값을 생성한다. 토크 비교기(210)는 생성한 토크 비교값을 스위칭 로직발생기(220)로 출력한다.
또한 자속값은 적응 관측기(225)로부터 나온 값과 감산하여 자속 비교기(215)에서 값을 비교하여, 상기 비교기(210 내지 215)들로부터 출력된 값들을 스위칭 로직발생기(220)를 통해 게이팅 신호(Gating signal)를 발생시킨다. 상기 게이팅 신호는 3-레벨 인버터(230)와 적응 관측기(225)로 전달된다. 다시 말하면, 자속 비교기(215)는 자속값을 입력 받는다. 이때, 자속값은 자속 비교기(215)로 입력 되는 자속 제어값(

)에서 적응 관측기(225)로부터 입력 받은 제2 궤환값(

)을 감산한 값이다. 자속 비교기(215)는 자속값과 자속 기준값을 비교하여 자속 비교값을 생성한다. 자속 비교기(215)는 생성한 자속 비교값을 스위칭 로직발생기(220)로 출력한다.
스위칭 로직발생기(220)는 토크 비교기(210)로부터 토크 비교값을 입력 받고, 자속 비교기(215)로부터 자속 비교값을 입력받는다. 스위칭 로직발생기(220)는 토크 비교값과 자속 비교값을 이용하여 게이팅 신호(Gating signal)을 생성한다. 스위칭 로직발생기(220)는 게이팅 신호를 3-레벨 인버터(230)로 출력한다.
상기 3-레벨 인버터(230)는 게이팅 신호값에 따라 유도 전동기(235)를 제어하게 된다. 즉, 3-레벨 인버터(230)는 게이팅 신호값을 이용하여 유도 전동기(235)를 제어하기 위한 전류 제어값을 생성한다. 3-레벨 인버터(230)는 전류 제어값을 이용하여 유도 전동기(235)를 제어한다.
상기 적응 관측기(225)로 전달된 게이팅 신호와 3-레벨 인버터(230)의 출력값들로부터 관측값을 생성, 출력하여 입력되는 토크값과 자속값에 감산하여 시스템의 구조적인 오차를 보상하게 된다. 즉, 적응 관측기(225)는 게이팅 신호를 통해 연산한 인버터 전압값(V_qs, V_ds)을 입력 받는다. 적응 관측기(225)는 전류 제어값을 통해 연산한 인버터 전류값(I_qs, I_ds)을 입력 받는다. 적응 관측기(225)는 입력 받은 인버터 전압값(V_qs, V_ds) 및 인버터 전류값(I_qs, I_ds)을 이용하여 제1 궤환값(

), 제2 궤환값(

) 및 제3 궤환값(

)을 생성한다.
도 4는 본 발명의 바람직한 실시예에 따른 유도 전동기 제어 시스템을 구현하기 위한 블록선도를 나타낸 도면이다.The summer receives the reverse force control input value U _n from the speed control block and the error compensation control input value U _p from the adaptive law control block. The summer generates a torque value by summing the received reverse force control input value U _n and the error compensation control input value U _p . The summer outputs the generated torque value to the torque comparator 210.
The torque value is subtracted from the value from the adaptive observer 225 before input and output to the speed controller 505 and compared through the torque comparator 210 to compensate for the error with the input value. That is, the torque comparator 210 has a third feedback value (a) in the torque value output from the summer of the speed controller.

Input the torque value after subtracting). The torque comparator 210 generates a torque comparison value by comparing the torque value with the torque reference value. The torque comparator 210 outputs the generated torque comparison value to the switching logic generator 220.
In addition, the magnetic flux value is subtracted from the value from the adaptive observer 225 to compare the value in the magnetic flux comparator 215, the value output from the comparators 210 to 215 through the switching logic generator 220 gating signal ( Generate a gating signal. The gating signal is passed to a three-level inverter 230 and an adaptive observer 225. In other words, the magnetic flux comparator 215 receives the magnetic flux value. At this time, the magnetic flux value is a magnetic flux control value input to the magnetic flux comparator 215 (

), The second feedback value () received from the adaptive observer 225

) Is subtracted. The magnetic flux comparator 215 compares the magnetic flux value with the magnetic flux reference value to generate a magnetic flux comparison value. The magnetic flux comparator 215 outputs the generated magnetic flux comparison value to the switching logic generator 220.
The switching logic generator 220 receives a torque comparison value from the torque comparator 210 and receives a magnetic flux comparison value from the magnetic flux comparator 215. The switching logic generator 220 generates a gating signal using the torque comparison value and the magnetic flux comparison value. The switching logic generator 220 outputs the gating signal to the three-level inverter 230.
The three-level inverter 230 controls the induction motor 235 according to the gating signal value. That is, the three-level inverter 230 generates a current control value for controlling the induction motor 235 using the gating signal value. The three-level inverter 230 controls the induction motor 235 using the current control value.
An observation value is generated and output from the gating signal transmitted to the adaptive observer 225 and the output values of the three-level inverter 230 to compensate for the structural error of the system by subtracting the input torque value and the magnetic flux value. That is, the adaptive observer 225 receives the inverter voltage values V _qs and V _ds calculated through the gating signal. The adaptive observer 225 receives the inverter current values I _qs and I _ds calculated through the current control values. The adaptive observer 225 uses the input inverter voltage values V _qs and V _ds and the inverter current values I _qs and I _ds to determine the first feedback value (

), The second feedback value (

) And the third feedback value (

).
4 is a block diagram for implementing an induction motor control system according to a preferred embodiment of the present invention.

delete

4 is a schematic diagram of a control system based on the DS1003 control boards 410, 415, and 420 in an implementation of a control system, including a three-level inverter system and a main controller board equipped with a magnetic flux observer for speed control. have. The PC installed with the motor speed control software is provided with a DS1003 control board (410, 415, 420), the DS1003 control board (410, 415, 420) is a TMS320C40 (410) and an analog-to-digital converter (415) that acts as a control And a digital input / output port 420. The command input from the PC calculates a control value from the TMS320C40 410, and the value output from the digital input / output port 420 passes through a gate pulse generator 425 to generate a pulse and generate a gate drive ( Gate drive 430 is connected to the three-level inverter system to control the induction motor 435. The values output from the three-level system are transmitted to the analog-to-digital converter 415, converted to digital, and compensated for structural errors of the system through the TMS320C40 (410).

6 is a view showing a control result in an induction motor using a conventional IP controller, Figure 7 is a view showing a control result in an induction motor using RBFN according to a preferred embodiment of the present invention.

FIG. 6 illustrates a schematic simulation result of a speed control system including a main controller substrate equipped with a three-leval interleaver system and a magnetic flux observer for speed control as shown in FIG. 4.

J= 4.0 x J,

B= 4.0 x B, Tl=8로서 J는 관성계수, B는 점성 마찰 계수이며, Tl은 외부 부하의 토크를 의미한다. 토크 변동 감소 알고리즘과 제안된 RBFN 속도 제어기를 위해 제어 주기의 샘플링시간은 200

s로 맞춰졌다. 제안된 알고리즘이 고출력 유도 전동기 DTC 시스템을 위한 것이기 때문에, 교대 주파수는 500Hz~1.0kHz 사이의 값으로 유지되었다. 그리고 제안된 RBFN제어기의 효과를 보이기 위해, RBFN 관측기가 없는 공칭 제어 입력에 의한 상대적인 결과와, 세밀하게 조정된 IP 제어를 사용한다. 상기 도 6내지 도 7은 희망 속도 500rpm에서의 회전 속도(speed), 상전류(Lqs)와 제어 입력(Control Input)이 제시되었다. 이 경우에는 불확실성이 없으므로, 성능 추적을 통한 시뮬레이션 결과는 효과적인 추적 성능을 보여주고 있다. 그러나 2.5초에 파라미터 변화와 급격한 외란이 일어났을 경우, 도 7은 도 6과 비교하여 더 강인한 성능을 보여준다. 상기 도 6은 2.5초에서 IP제어가 약 1.5초 정도 걸리는데 비해, RBFN제어기를 통한 제어는 약 0.2초가 걸린다. 또한 RBFN에 비해 IP제어기는 오차의 변동과 함께 더 큰 추적 오차를 보여준다. The induction motor used here has a rated speed of 220Vac and 1740rpm in three phases, and to show the relative results of the proposed approach, an abnormal condition including disturbance of the load was used. The conditions used for testing

J = 4.0 x J,

B = 4.0 x B, Tl = 8, where J is the inertia coefficient, B is the viscous friction coefficient, and Tl is the torque of the external load. For the torque fluctuation reduction algorithm and the proposed RBFN speed controller, the sampling time of the control period is 200

set to s Since the proposed algorithm is for high power induction motor DTC systems, the alternating frequency was maintained between 500Hz and 1.0kHz. And to show the effectiveness of the proposed RBFN controller, we use the relative results of the nominal control input without the RBFN observer and the finely tuned IP control. 6 to 7 show a rotation speed, a phase current Lqs, and a control input at a desired speed of 500 rpm. Since there is no uncertainty in this case, the simulation results from the performance tracking show effective tracking performance. However, when a parameter change and a sudden disturbance occurs in 2.5 seconds, Figure 7 shows a more robust performance compared to Figure 6. In FIG. 6, the IP control takes about 1.5 seconds in 2.5 seconds, whereas the control through the RBFN controller takes about 0.2 seconds. In addition, compared to RBFN, IP controller shows greater tracking error with variation of error.

In the above simulation results, since the overall DTC system including the three-level inverter is approximated by the simple DC motor shown in Fig. 4, the approximation error and the unknown uncertainty are inevitable and affect the performance. However, in the simulation phase, no change occurs because the designed AC motor is used with the IP control gain calculated by the mechanical parameters. Through these results, the feasibility, robust and stable characteristics are confirmed, and the proposed RBFN controller shows a relatively small tracking error even if the inherent tracking error is severely applied, compared to the conventional IP controller structure.

The present invention shows an embodiment of a direct torque control system (DTC) of a three-level induction motor, and is not limited to the system implemented in the present invention, but implements various types of controllers or motor systems such as vector control or PM motors. It is also possible.

While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments, but is capable of various modifications within the scope of the invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the scope of the following claims, but also by those equivalent to the scope of the claims.

In the present invention operating as described in detail above, the effects obtained by the representative of the disclosed invention are briefly described as follows.

The present invention uses RBFN as an observer in an induction motor control system without a speed sensor to approximate the inherent uncertainty and estimates the error, and considering the influence of the speed estimation error, a direct torque control system of a three-level induction motor (DTC) has the stability of the entire control system.

In addition, by providing the inexpensive induction motor system with the advantages of the existing high priced high-performance induction motor, the advanced neural network control method, which has remained mainly in theory until now, is applied to the induction motor widely used in the actual industrial field. This can be of great help to the motor control industry.

Claims (4)

In the speed control method of an induction motor using a radial basis function network observer, Generating a torque value for controlling the speed of the induction motor using the torque control value; Generating a torque comparison value by comparing the torque value output by the speed control with the torque reference value, and generating a magnetic flux comparison value by comparing the magnetic flux value and the magnetic flux reference value; Generating a gating signal by inputting the torque comparison value and the magnetic flux comparison value; Generating a current control value for controlling the induction motor by the gating signal; Generating a first feedback value, a second feedback value, and a third feedback value by using the gating signal and the current control value; Here, the torque control value is a value obtained by subtracting the first feedback value from an input value for controlling the speed, and the magnetic flux value is a value obtained by subtracting the second feedback value from a magnetic flux control value, and the torque comparison value. The torque value input to generate the value is a value obtained by subtracting the third feedback value from the torque value output by the speed control, The process of controlling the speed, Generating a weighted control value and a constant limit control value using the torque control value; Generating a weight, a constant limit value, and an error compensation control input value through an adaptive law using the weighted control value, the constant limit control value, and a function control value input from the outside; Generating a function output value using the weight and the constant limit value; Generating a reverse force control input value using the torque control value and the function output value; And generating the torque value by adding the error compensation control input value and the reverse force control input value to generate a torque value. In the speed control apparatus of an induction motor using a radial basis function network observer, A speed controller for generating a torque value for controlling the speed of the induction motor by using the torque control value; A torque comparator for generating a torque comparison value by comparing the torque value output by the speed control with the torque reference value; A magnetic flux comparator for generating a magnetic flux comparison value by comparing the magnetic flux value with the magnetic flux reference value, A switching logic generator configured to generate a gating signal by inputting the torque comparison value and the magnetic flux comparison value; A three-level inverter for generating a current control value for controlling the induction motor by the gating signal; Including an adaptive observer for generating a first feedback value, a second feedback value and a third feedback value using the gating signal and the current control value, Here, the torque control value is a value obtained by subtracting the first feedback value from an input value input to the speed controller to control the speed, and the magnetic flux value is the second control value from the magnetic flux control value input to the magnetic flux comparator. A value obtained by subtracting the feedback value, the torque value input to the switching logic generator is a value obtained by subtracting the third feedback value from the torque value output from the speed controller, The speed controller, A radial basis function network observer generating a weighted control value and a constant limiting control value using the torque control value, and generating a function output value using the weight and the constant limiting value; Receiving the weighted control value and the constant limiting control value from the radial basis function network observer, and using the weighted control value, the constant limiting control value, and a function control value input from the outside, the weight, An adaptive law control block for generating constant limit and error compensation control inputs; A speed control block receiving a function output value from the radial basis function network observer and generating a reverse force control input value using the torque control value and the function output value; An adder for generating the torque value by summing the error compensation control input value input from the adaptive law control block and the reverse force control input value input from the speed control block, and outputting the generated torque value to the torque comparator. Speed control device of induction motor using a radial basis function network observer, characterized in that it comprises a. delete delete

Images