CN110048657B - Decoupling control method for neural network inverse system of bearingless asynchronous motor - Google Patents
Decoupling control method for neural network inverse system of bearingless asynchronous motor Download PDFInfo
- Publication number
- CN110048657B CN110048657B CN201910431278.7A CN201910431278A CN110048657B CN 110048657 B CN110048657 B CN 110048657B CN 201910431278 A CN201910431278 A CN 201910431278A CN 110048657 B CN110048657 B CN 110048657B
- Authority
- CN
- China
- Prior art keywords
- asynchronous motor
- neural network
- rotor
- bearingless asynchronous
- torque
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P23/00—Arrangements or methods for the control of AC motors characterised by a control method other than vector control
- H02P23/0004—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control
- H02P23/0018—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control using neural networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P23/00—Arrangements or methods for the control of AC motors characterised by a control method other than vector control
- H02P23/0004—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control
- H02P23/0022—Model reference adaptation, e.g. MRAS or MRAC, useful for control or parameter estimation
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Magnetic Bearings And Hydrostatic Bearings (AREA)
- Control Of Ac Motors In General (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
A decoupling control method for a neural network inverse system of a bearingless asynchronous motor establishes a system state equation of the bearingless asynchronous motor on the basis of considering the dynamic characteristics of the stator current of a torque winding; then, according to the nonlinear coupling relation between the input variable and the output variable of the inverse system, using a neural network to identify the mathematical model of the inverse system of the bearingless asynchronous motor; secondly, a neural network inverse system mathematical model of the bearingless asynchronous motor is connected with an original system in series by adopting an inverse system decoupling method, and the bearingless asynchronous motor is decoupled into four independent second-order linear subsystems such as a rotor flux linkage, a rotating speed and two radial displacement components; and finally, configuring a closed-loop regulator for each subsystem to complete inverse system dynamic decoupling control of the bearingless asynchronous motor, belonging to the field of novel special motor driving and control and being particularly suitable for high-performance magnetic suspension operation control application occasions of the bearingless asynchronous motor.
Description
Technical Field
The invention relates to the technical field of special alternating current motor detection control, in particular to a decoupling control method for a neural network inverse system of a bearingless asynchronous motor.
Background
The bearingless motor is a novel motor which is developed in recent years and suitable for high-speed operation based on the similarity of the magnetic bearing and the stator structure of the alternating-current motor, and has wide application prospects in the fields of aerospace, material sealing transmission, advanced manufacturing and the like. The existing documents and patent retrieval find that a complex electromagnetic coupling relation exists in the bearingless asynchronous motor, and the research on inverse system decoupling control is already carried out at home and abroad, but the analysis inverse system model after considering the current dynamic state is relatively complex and is inconvenient for engineering application. The neural network has the capability of approximating a complex nonlinear function with any precision and self-learning, and if the neural network is adopted to identify the inverse system model of the bearingless asynchronous motor, the difficulty in realizing the 'analysis of the inverse system' can be effectively solved; however, the existing bearing-free asynchronous motor neural network inverse system control research does not consider the dynamic characteristics of stator current, and the improvement of the dynamic control performance of the bearing-free asynchronous motor is limited to a certain extent. Therefore, under the condition of considering the dynamic characteristics of the stator current, the patent provides a decoupling control method for a neural network inverse system of a bearingless asynchronous motor
Disclosure of Invention
In order to solve the technical problem, the invention provides a decoupling control method for a neural network inverse system of a bearingless asynchronous motor.
The technical scheme adopted by the invention for solving the technical problems is as follows: the decoupling control method for the neural network inverse system of the bearingless asynchronous motor comprises the following steps:
step one, establishing a non-linear relation model of a bearingless asynchronous motor
Setting alpha-beta as a static two-phase orthogonal coordinate system, the alpha coordinate axis of which is consistent with the axial direction of the A-phase torque winding of the three-phase bearingless asynchronous motor, the beta coordinate axis is in the counterclockwise vertical direction of the alpha coordinate axis, setting d-q as a torque system rotor flux linkage orientation synchronous rotation coordinate system,
respectively selecting control variables of the system by considering the current dynamic differential equation of the torque windinguState variable ofxAnd output variablesyComprises the following steps:
then, a bearing-free asynchronous motor original system model can be established:
in the formulae (5) to (8),u 1 =u s1d,u 2 =u s1q,u 3 =i s2d,u 4 =i s2q;x 1 =α,x 2 =β,x 3 =,x 4 =,x 5 =i s1d,x 6 =i s1q,x 7 =ψ r1,x 8 =ω r ;y 1 =α,y 2 =β,y 3 =ψ r1,y 4 =ω r ;u s1d、u s1qis composed ofd-qOf torque windings in a coordinate systemd、qAn axis voltage component;i s1d、i s1qis composed ofd-qOf torque windings in a coordinate systemd、qAn axis current component;i s2d、i s2qis composed ofd-qOf windings suspended in a coordinate systemd、qAn axis current component;α、βis the rotor radial displacement component in the α - β coordinate system;ψ r1is the rotor flux linkage of the torque system;ω r is the rotor rotational angular velocity;K m is a magnetic suspension force coefficient determined by a motor structure,,μ 0in order to be the air gap permeability,lis the length of the stator iron core of the motor,rthe inner diameter of the stator is the inner diameter,L m2is a single-phase excitation inductor of a suspension winding,N 1、N 2effective series turns of the torque winding and the suspension winding are respectively;k sa radial displacement stiffness coefficient which is determined by the structure of the motor and the magnetic field intensity;mis the rotor mass;Jis the moment of inertia of the mechanical rotor;,, ,,,;R s1is a torque winding resistance;R r1is the rotor resistance;L m1is the mutual inductance of an equivalent two-phase torque system in a d-q coordinate system;L s1 is the self-inductance of the equivalent two-phase torque winding in the d-q coordinate system,L s1 = L m1 +L s l1;L r1is the self-inductance of the equivalent two-phase rotor winding in the d-q coordinate system,L r1 =L m1 +L r l1;L s l1for the stator leakage inductance of the torque system,L r l1rotor leakage inductance for torque systems;is the magnetic flux leakage coefficient;T L is the load torque;p 1is the number of pole pairs of the torque winding;
according to the implicit function theorem, a non-linear relation model of the inverse system of the bearingless asynchronous motor is established:
;Uis an output variable of an inverse system, comprisingu 1、u 2、u 3Andu 4four components;representing slave variablesTo inverse system output variablesUThe non-linear mapping relationship of (1);
step two, establishing a bearing-free asynchronous motor neural network inverse system model
Selecting a three-layer feedforward neural network, and adopting a hyperbolic tangent S-type transmission function as an excitation function of a hidden layer neuron:
sampling input variables and output variables of the bearingless asynchronous motor to obtain original data samples of the input variables and the output variables, performing smooth filtering processing on the sampled original data, and solving first and second derivatives of a rotor flux linkage, a rotating speed, an α radial displacement component and a β radial displacement component according to the smooth filtered data to form a neural network training sample setAnd;
optimizing and calculating a neural network training sample set through a genetic algorithm, taking a summation value of absolute values of prediction errors of training data as a judgment value of individual fitness of the neural network training sample set, selecting an optimal individual with the minimum judgment value as the individual fitness, assigning an initial weight and a threshold of the neural network by using the optimal individual, identifying and training the neural network through an LM (Linear modeling) algorithm, and ensuring that the output mean square error of the neural network is less than 0.001, so that the connection weight coefficient between each layer of neurons of the neural network is determined, and a bearingless asynchronous motor neural network inverse system model can be obtained according to the bearingless asynchronous motor inverse system nonlinear relation model;
step three, establishing a neural network inverse decoupling control system of the bearingless asynchronous motor
The output of the bearingless asynchronous motor neural network inverse system is connected with the input of the bearingless asynchronous motor neural network inverse system in series, the four closed loop regulators are respectively connected with the input of the bearingless asynchronous motor neural network inverse system in series, the numerical values of the rotor flux linkage, the rotating speed, the alpha radial displacement component and the beta radial displacement component output by the bearingless asynchronous motor original system are respectively compared with a given value of each closed loop regulator, and the four comparison operation difference values are respectively sent to each closed loop regulator to be used as the input of the bearingless asynchronous motor neural network inverse system, so that the bearingless asynchronous motor neural network inverse decoupling control system is formed, and the bearingless asynchronous motor neural network inverse system decoupling control is realized.
Preferably, the method for establishing the original system model of the bearingless asynchronous motor comprises the following steps:
establishing a nonlinear dynamic mathematical model of a torque system:
in formula (1):respectively representing voltage, current, resistance and flux linkage; subscript s1Andr 1stator and rotor respectively representing a torque system; subscriptd、qRespectively in a rotor flux linkage coordinate systemd、qAn axial component;respectively representing the rotor self-inductance, the stator self-inductance and the mutual inductance between the stator and the rotor of a torque system;L s l1、L r l1respectively, are torque systemsd-qLeakage inductance of the stator and the rotor in a coordinate system;p 1is the number of pole pairs of the torque winding;T L is the load torque;T r1 is the rotor time constant;ω ris the rotor rotational angular velocity;ω 1is the synchronous electrical angular velocity of the motor;Jis the moment of inertia of the mechanical rotor; other variables are defined in equation (8);
according to the magnetic suspension control principle, a controllable radial magnetic suspension force model of the bearingless asynchronous motor is established:
in formula (2):K m is a magnetic suspension force coefficient determined by a motor structure,F α、F βis at rest alongα、βThe controllable radial magnetic levitation force component of the coordinate axial direction;i s d2、i s q2respectively being two-pole suspension windingsd、qAn axis current component;of separate torque system air-gap flux linkagesd、qThe axial component of the magnetic flux is,the expression of (a) is:
rotor radial suspension equation of motion:
in formula (4):mis the rotor mass;f α、f βrespectively single-side electromagnetic tension components generated by unbalanced air gap magnetic field distribution when the rotor is eccentric,whereink sTo determine the radial displacement stiffness coefficient of the motor structure and magnetic field strength,α、βis the rotor radial displacement component in the α - β coordinate system;
substituting the expressions (5), (6) and (7) into the expressions (1) to (4) to obtain an eight-order state equation of the original system model of the bearingless asynchronous motor.
Preferably, in the second step, a three-layer feedforward neural network with 12 input nodes, 4 output nodes and 23 hidden layer nodes is selected.
Preferably, in step three, the closed-loop regulator adopts a PD regulator with a low-pass filtering link, and the transfer function of the closed-loop controller is:
in the formula (17), the compound represented by the formula (I),is a lead differential time constant;is a low pass filtering element time constant.
According to the technical scheme, the invention has the beneficial effects that:
1. in the neural network inverse system inverse model considering the current dynamic state, the influence of the stator current dynamic state characteristic objectively existing in the motor is fully considered, so that higher magnetic suspension operation dynamic decoupling control performance can be obtained compared with the condition when the stator current dynamic state characteristic is not considered.
2. The decoupling control method for the neural network inverse system of the bearingless asynchronous motor can simplify the structure of the whole control system to a certain extent. When the bearingless asynchronous motor neural network inverse system model is trained and identified, the cross coupling characteristics between stator current dynamic states and components are considered, and a load torque variable which is difficult to predict is not needed in the inverse system model, so that a torque winding current regulation closed loop in an original system and a real-time identification link of load torque can be omitted.
3. After a bearingless asynchronous motor inverse system model considering the current dynamic state is obtained by utilizing approximation and learning ability identification of a neural network to a nonlinear function, through inverse system decoupling, not only can excellent decoupling performance after the stator current dynamic state is considered be obtained, but also the problem that a complex analytic inverse system model is inconvenient to realize can be solved, and the method has the characteristics of small overshoot, good tracking performance, strong robustness to parameter change and the like.
Drawings
FIG. 1 is a topological structure diagram of an inverse system model for identifying a bearingless asynchronous motor by neural network training;
FIG. 2 is a schematic diagram of neural network inverse system decoupling for bearingless asynchronous motors;
fig. 3 is a schematic structure diagram of a neural network inverse decoupling control system of a bearingless asynchronous motor.
Detailed Description
A decoupling control method for a neural network inverse system of a bearingless asynchronous motor establishes a system state equation of the bearingless asynchronous motor on the basis of considering the dynamic characteristics of the stator current of a torque winding; then, according to the nonlinear coupling relation between the input variable and the output variable of the inverse system, using a neural network to identify the mathematical model of the inverse system of the bearingless asynchronous motor; secondly, a neural network inverse system mathematical model of the bearingless asynchronous motor is connected with an original system in series by adopting an inverse system decoupling method, and the bearingless asynchronous motor is decoupled into four independent second-order linear subsystems such as a rotor flux linkage, a rotating speed and two radial displacement components; and finally, configuring a closed-loop regulator for each subsystem to complete inverse system dynamic decoupling control of the bearingless asynchronous motor, belonging to the field of novel special motor driving and control and being particularly suitable for high-performance magnetic suspension operation control application occasions of the bearingless asynchronous motor.
The invention principle of the patent is based on:
1. the bearingless asynchronous motor is a multivariable, nonlinear and strongly coupled complex object, wherein a complex mechanical-electrical-magnetic coupling relation exists between a torque system and a magnetic suspension system, and the nonlinear relation between related variables can be deduced through mechanism analysis according to the working principle of the bearingless asynchronous motor.
2. The neural network has the capability of approximating a complex nonlinear function with any precision and self-learning; if the inverse system mathematical model of the bearingless asynchronous motor considering the stator current dynamic state can be obtained through off-line training and identification learning according to the coupling relation between the related variables, the inverse system dynamic decoupling control of the bearingless asynchronous motor can be carried out, and the problem of difficulty in realizing the complex analytic inverse system mathematical model can be effectively solved. In order to achieve the purpose, the technical means adopted by the invention is as follows: a decoupling control method for a neural network inverse system of a bearingless asynchronous motor comprises the following steps:
1) establishing a non-linear relation mathematical model of a bearingless asynchronous motor
α - β are set as a static two-phase orthogonal coordinate system, the α coordinate axis of the system is consistent with the axial direction of the A-phase torque winding of the three-phase bearingless asynchronous motor, the β coordinate axis is in the anticlockwise vertical direction of the α coordinate axis, d-q are set as a rotor flux linkage orientation synchronous rotation coordinate system of the torque system, and the stator current dynamic equation and the rotor magnetic field orientation constraint condition are considered ""etc., to obtain a nonlinear dynamic mathematical model of the torque system:
in formula (1):respectively representing voltage, current, resistance and flux linkage; subscript s1Andr 1stator and rotor respectively representing a torque system; subscriptd、qRespectively in a rotor flux linkage coordinate systemd、qAn axial component;respectively representing the rotor self-inductance, the stator self-inductance and the mutual inductance between the stator and the rotor of a torque system;L s l1、L r l1respectively, are torque systemsd-qFixed and rotary in coordinate systemA sub leakage inductance;p 1is the number of pole pairs of the torque winding;T L is the load torque;T r1 is the rotor time constant;ω ris the rotor rotational angular velocity;ω 1is the synchronous electrical angular velocity of the motor.
According to the magnetic suspension control principle, a controllable radial magnetic suspension force model of the bearingless asynchronous motor can be obtained:
in formula (2):K m is a magnetic suspension force coefficient determined by a motor structure,F α、F βis at rest alongα、βThe controllable radial magnetic levitation force component of the coordinate axial direction;i s d2、i s q2respectively being two-pole suspension windingsd、qAn axis current component;of separate torque system air-gap flux linkagesd、qThe axial component of the magnetic flux is,the expression of (a) is:
rotor radial suspension equation of motion:
in formula (4):mis the rotor mass;f α、f βrespectively single-side electromagnetic tension components generated by unbalanced air gap magnetic field distribution when the rotor is eccentric,which isIn (1)k sTo determine the radial displacement stiffness coefficient of the motor structure and magnetic field strength,α、βis the rotor radial displacement component in the α - β coordinate system;
considering a torque winding current dynamic differential equation, respectively selecting a control variable, a state variable and an output variable of a system as follows:
wherein: being torque windingsd、qAn axis voltage component;i s2d、i s2qrespectively being two-pole suspension windingsd、qα and β are rotor radial displacement components in a α - β coordinate system;is the rotor flux linkage of the torque system; is the rotor rotational angular velocity.
Substituting equations (5), (6) and (7) into equations (1) to (4) can obtain the system with the following eight-order state equation:
in the formula (8), the reaction mixture is,,,,,;R s1 is a torque winding resistance;R r1 is the rotor resistance;L m1is the mutual inductance of an equivalent two-phase torque system in a d-q coordinate system;L s1 is the self-inductance of the equivalent two-phase torque winding in the d-q coordinate system,L s1 = L m1 +L s l1;L r1 is the self-inductance of the equivalent two-phase rotor winding in the d-q coordinate system,L r1 =L m1 +L r l1;L s l1、L r l1respectively the leakage inductance of the stator and the rotor of the torque system;is the magnetic flux leakage coefficient;T L is the load torque;p 1is the pole pair number of the torque winding.
And (4) carrying out system reversibility analysis by adopting an Interactor algorithm. To the outputThe time derivative is calculated successively until the derivative of the output variable contains the input control quantityUntil now. The specific derivation process is as follows:
meanwhile, according to the selected state variable, equation (3) becomes:
in the normal operation of the system,dthe rotor flux linkage and the air gap flux linkage of the shaft are both different from 0. Therefore, det (A) ≠ 0,rank(A) Relative order of the system is α = (α)1, α2, α3, α4) = 2, 2, and satisfies:that is, the sum of the relative steps of the system is equal to the order of the equation of the system state, it can be known from the inverse system theory that the bearingless asynchronous motor system described by the equation (8) is reversible.
According to the implicit function theorem and by combining equations (9) to (12), the non-linear relation model of the bearingless asynchronous motor inverse system considering the dynamic characteristics of the stator current can be expressed as follows:
2) training identification of neural network inverse system model
As shown in FIG. 1, a three-layer feedforward network with 12 input nodes, 4 output nodes and 23 hidden layer nodes is selected for inverse system model identification The connection weights between neurons in the input layer and the hidden layer, and between neurons in the hidden layer and the output layer are respectively.
The hyperbolic tangent S-type transmission function is adopted as the excitation function of the hidden layer neuron, and the method specifically comprises the following steps:
the method comprises the following specific training and identification steps of considering a mathematical model of a bearing-free asynchronous motor neural network inverse system of stator current dynamic states:
(1) sampling to obtain input and output original data samples. Based on a bearingless asynchronous motor analytic inverse control system considering stator current dynamic, random quantities distributed normally in a proper range are selected as excitation signals according to the actual physical running area of the motor, and the whole excitation time is 10 s. Sampling is carried out on input variables (two-phase current of a suspension winding and two-phase voltage of a torque winding) and output (two radial displacement components, rotor flux linkage and rotating speed) of the bearingless asynchronous motor, and the sampling step length is taken to be 0.001 s.
(2) A set of training and testing samples is obtained. After the sampled original data are subjected to smooth filtering, a high-precision five-point numerical differentiation method is adopted to obtain first and second derivatives of two radial displacement components, a rotor flux linkage and a rotating speed, so that a neural network training sample set is formedAnd. 6000 groups are selected as training sets at equal intervals in the obtained 10000 groups of data, and the rest 4000 groups are used as test sets. In practical applications, the raw data obtained by sampling and the "derivative" data obtained by numerical differentiation may not be in an order of magnitude, and therefore, the training sample set is often normalized. This facilitates convergence of the neural network training, which avoids the neural network being particularly sensitive or insensitive to a certain input quantity.
(3) And (5) training and identifying a neural network inverse system model. Firstly, optimizing an initial weight and a threshold of a BP neural network by adopting a genetic algorithm so as to avoid the influence of the initial weight and the threshold of the neural network on the training and identifying result of the bearingless asynchronous motor inverse system model. Each individual in the genetic algorithm population comprises all weights and thresholds of a network, the individual calculates the individual fitness through a fitness function, and the genetic algorithm finds out the individual corresponding to the optimal fitness value through selection, intersection and variation operations; wherein, the sum of absolute values of the prediction errors of the training data is used as the individual fitness, and the smaller the value of the individual fitness is, the better the individual is. And then, assigning values to the initial weight and the threshold value of the network by using the optimal individual. Finally, the static neural network in fig. 1 is trained using the LM algorithm. After about 1000 times of training, the output mean square error of the neural network is less than 0.001, and the requirements are met, so that each weight coefficient in the static neural network topological structure in the figure 1 is determined, and a bearingless asynchronous motor neural network inverse system model considering the stator current dynamic state is obtained.
3) Neural network inverse decoupling control system for constructing bearingless asynchronous motor
Firstly, neural network inverse system decoupling of the bearingless asynchronous motor is carried out. To output variableAcceleration variable ofAdding an integrator to obtainDecoupling the asynchronous motor system without bearing into four (pseudo) linear second-order integral subsystems including rotor flux linkage subsystem, rotation speed subsystem, α radial displacement component subsystem and β radial displacement component subsystem, with the equivalent transfer function of each subsystem being 1/s2Fig. 2 shows a schematic diagram of neural network inverse system decoupling.
Each subsystem is then configured with a closed-loop regulator. From linear system theory, it can be known that: for a transfer function of 1/s2The subsystem of (2) can obtain good control effect by adopting the PD regulator. However, in order to reduce the gain of the high-frequency section system and resist the influence of high-frequency noise, an improved PD regulator structure with a low-pass filtering link is adopted, and a specific closed-loop controllerThe transfer function of (a) is:
in the formula (17), the compound represented by the formula (I),is a lead differential time constant;is a low pass filtering element time constant.
With the addition of the closed-loop regulator, each linear subsystem is rectified to a typical type ii system, i.e., the open-loop transfer function of the subsystem is corrected to:
according to the dynamic performance index requirements of each subsystem and the structure of the regulator in the formula (17), a rotor flux linkage regulator, a rotating speed regulator, an alpha radial displacement component regulator and a beta radial displacement component regulator can be respectively designed.
And finally, constructing a neural network inverse decoupling control system of the bearingless asynchronous motor, wherein a structure diagram of the neural network inverse decoupling control system of the bearingless asynchronous motor considering current dynamics is shown in fig. 3. Firstly, comprehensively comparing a given value of a rotor flux linkage with a feedback value thereof, then sending a rotor flux linkage error into a flux linkage regulator, and taking the output quantity of the flux linkage regulator as a second derivative given signal of the rotor flux linkageSending the signal to a corresponding signal input end of a neural network inverse system; the given value of the rotating speed is compared with the feedback value of the rotating speed comprehensively, the rotating speed error is sent to a rotating speed regulator, and the output quantity of the rotating speed regulator is taken as a second derivative given signal of the rotating speedComprehensively comparing the given value of α radial displacement component with its feedback value, sending the radial displacement error to α direction displacement regulator, and using the output of α direction displacement regulator as the second derivative given signal of α direction displacementThe connection method of β displacement regulator is similar to that of α displacement regulator, the closed loop regulator of rotor flux linkage, rotation speed, α displacement component β displacement component and mathematical model of neural network inverse system form a composite controller of bearingless asynchronous motor together, thus realizing dynamic decoupling control of neural network inverse system of bearingless asynchronous motor.
Claims (4)
1. The decoupling control method for the neural network inverse system of the bearingless asynchronous motor is characterized by comprising the following steps of:
step one, establishing a non-linear relation model of a bearingless asynchronous motor
Setting alpha-beta as a static two-phase orthogonal coordinate system, the alpha coordinate axis of which is consistent with the axial direction of the A-phase torque winding of the three-phase bearingless asynchronous motor, the beta coordinate axis is in the counterclockwise vertical direction of the alpha coordinate axis, setting d-q as a torque system rotor flux linkage orientation synchronous rotation coordinate system,
respectively selecting control variables of the system by considering the current dynamic differential equation of the torque windinguState variable ofxAnd output variablesyComprises the following steps:
then, a bearing-free asynchronous motor original system model can be established:
in the formulae (5) to (8),u 1 =u s1d,u 2 =u s1q,u 3 =i s2d,u 4 =i s2q;x 1 =α,x 2 =β,x 3 =,x 4 =,x 5 =i s1d,x 6 =i s1q,x 7 =ψ r1,x 8 =ω r ;y 1 =α,y 2 =β,y 3 =ψ r1,y 4 =ω r ;u s1d、u s1qis composed ofd-qOf torque windings in a coordinate systemd、qAn axis voltage component;i s1d、i s1qis composed ofd-qOf torque windings in a coordinate systemd、qAn axis current component;i s2d、i s2qis composed ofd-qOf windings suspended in a coordinate systemd、qAn axis current component;α、βis the rotor radial displacement component in the α - β coordinate system;ψ r1is the rotor flux linkage of the torque system;ω r is the rotor rotational angular velocity;K m is a magnetic suspension force coefficient determined by a motor structure,,μ 0in order to be the air gap permeability,lis the length of the stator iron core of the motor,rthe inner diameter of the stator is the inner diameter,L m2is a single-phase excitation inductor of a suspension winding,N 1、N 2effective series turns of the torque winding and the suspension winding are respectively;k sa radial displacement stiffness coefficient which is determined by the structure of the motor and the magnetic field intensity;mis the rotor mass;Jis the moment of inertia of the mechanical rotor;,, ,,,;R s1is a torque winding resistance;R r1is the rotor resistance;L m1is the mutual inductance of an equivalent two-phase torque system in a d-q coordinate system;L s1 is the self-inductance of the equivalent two-phase torque winding in the d-q coordinate system,L s1 = L m1 + L s l1;L r1is the self-inductance of the equivalent two-phase rotor winding in the d-q coordinate system,L r1 =L m1 +L r l1;L s l1for the stator leakage inductance of the torque system,L r l1rotor leakage inductance for torque systems;is the magnetic flux leakage coefficient;T L is the load torque;p 1is the number of pole pairs of the torque winding;
according to the implicit function theorem, a non-linear relation model of the inverse system of the bearingless asynchronous motor is established:
;Uis an output variable of an inverse system, comprisingu 1、u 2、u 3Andu 4four components;representing slave variablesTo inverse system output variablesUThe non-linear mapping relationship of (1);
step two, establishing a bearing-free asynchronous motor neural network inverse system model
Selecting a three-layer feedforward neural network, and adopting a hyperbolic tangent S-type transmission function as an excitation function of a hidden layer neuron:
sampling input variables and output variables of the bearingless asynchronous motor to obtain original data samples of the input variables and the output variables, performing smooth filtering processing on the sampled original data, and solving first and second derivatives of a rotor flux linkage, a rotating speed, an α radial displacement component and a β radial displacement component according to the smooth filtered data to form a neural network training sample setAnd;
optimizing and calculating a neural network training sample set through a genetic algorithm, taking a summation value of absolute values of prediction errors of training data as a judgment value of individual fitness of the neural network training sample set, selecting an optimal individual with the minimum judgment value as the individual fitness, assigning an initial weight and a threshold of the neural network by using the optimal individual, identifying and training the neural network through an LM (Linear modeling) algorithm, and ensuring that the output mean square error of the neural network is less than 0.001, so that the connection weight coefficient between each layer of neurons of the neural network is determined, and a bearingless asynchronous motor neural network inverse system model can be obtained according to the bearingless asynchronous motor inverse system nonlinear relation model;
step three, establishing a neural network inverse decoupling control system of the bearingless asynchronous motor
The output of the bearingless asynchronous motor neural network inverse system is connected with the input of the bearingless asynchronous motor neural network inverse system in series, the four closed loop regulators are respectively connected with the input of the bearingless asynchronous motor neural network inverse system in series, the numerical values of the rotor flux linkage, the rotating speed, the alpha radial displacement component and the beta radial displacement component output by the bearingless asynchronous motor original system are respectively compared with a given value of each closed loop regulator, and the four comparison operation difference values are respectively sent to each closed loop regulator to be used as the input of the bearingless asynchronous motor neural network inverse system, so that the bearingless asynchronous motor neural network inverse decoupling control system is formed, and the bearingless asynchronous motor neural network inverse system decoupling control is realized.
2. The decoupling control method for the neural network inverse system of the bearingless asynchronous motor according to claim 1, is characterized in that: the method for establishing the original system model of the bearingless asynchronous motor comprises the following steps:
establishing a nonlinear dynamic mathematical model of a torque system:
in formula (1):respectively representing voltage, current, resistance and flux linkage; subscript s1Andr 1stator and rotor respectively representing a torque system; subscriptd、qRespectively in a rotor flux linkage coordinate systemd、qAn axial component;respectively representing the rotor self-inductance, the stator self-inductance and the mutual inductance between the stator and the rotor of a torque system;L s l1、L r l1respectively, are torque systemsd-qLeakage inductance of the stator and the rotor in a coordinate system;p 1is the number of pole pairs of the torque winding;T L is the load torque;T r1 is the rotor time constant;ω ris the rotor rotational angular velocity;ω 1is the synchronous electrical angular velocity of the motor;Jis the moment of inertia of the mechanical rotor; other variables are defined in equation (8);
according to the magnetic suspension control principle, a controllable radial magnetic suspension force model of the bearingless asynchronous motor is established:
in formula (2):K m is a magnetic suspension force coefficient determined by a motor structure,F α、F βis at rest alongα、βThe controllable radial magnetic levitation force component of the coordinate axial direction;i s d2、i s q2respectively being two-pole suspension windingsd、qAn axis current component;of separate torque system air-gap flux linkagesd、qThe axial component of the magnetic flux is,the expression of (a) is:
rotor radial suspension equation of motion:
in formula (4):mis the rotor mass;f α、f βrespectively single-side electromagnetic tension components generated by unbalanced air gap magnetic field distribution when the rotor is eccentric,whereink sTo determine the radial displacement stiffness coefficient of the motor structure and magnetic field strength,α、βis the rotor radial displacement component in the α - β coordinate system;
substituting the expressions (5), (6) and (7) into the expressions (1) to (4) to obtain an eight-order state equation of the original system model of the bearingless asynchronous motor.
3. The decoupling control method for the neural network inverse system of the bearingless asynchronous motor according to claim 1, is characterized in that: in the second step, a three-layer feedforward neural network with 12 input nodes, 4 output nodes and 23 hidden layer nodes is selected.
4. The decoupling control method for the neural network inverse system of the bearingless asynchronous motor according to claim 1, is characterized in that: in the third step, the closed-loop regulator adopts a PD regulator with a low-pass filtering link, and the transfer function of the closed-loop controller is as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910431278.7A CN110048657B (en) | 2019-05-22 | 2019-05-22 | Decoupling control method for neural network inverse system of bearingless asynchronous motor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910431278.7A CN110048657B (en) | 2019-05-22 | 2019-05-22 | Decoupling control method for neural network inverse system of bearingless asynchronous motor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110048657A CN110048657A (en) | 2019-07-23 |
CN110048657B true CN110048657B (en) | 2020-10-16 |
Family
ID=67283314
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910431278.7A Expired - Fee Related CN110048657B (en) | 2019-05-22 | 2019-05-22 | Decoupling control method for neural network inverse system of bearingless asynchronous motor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110048657B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11577616B2 (en) * | 2020-10-27 | 2023-02-14 | GM Global Technology Operations LLC | Methods, systems, and apparatuses for torque control utilizing roots of pseudo neural network |
CN115001335A (en) * | 2022-05-25 | 2022-09-02 | 福州大学 | Bearing-free flux switching motor rotor suspension control method based on neural network |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2579119B2 (en) * | 1993-03-15 | 1997-02-05 | 方正 彭 | Vector controller for induction motor |
US7346462B2 (en) * | 2006-03-29 | 2008-03-18 | General Electric Company | System, method, and article of manufacture for determining parameter values associated with an electrical grid |
JP6210936B2 (en) * | 2014-05-30 | 2017-10-11 | ミネベアミツミ株式会社 | Self-excited resonance type power factor correction circuit and light source driving device |
CN105553365A (en) * | 2016-02-01 | 2016-05-04 | 四川长虹电器股份有限公司 | Automatic identification control method for parameters of permanent magnet brushless motor |
CN109217761A (en) * | 2018-09-26 | 2019-01-15 | 河南科技大学 | The inverse kinematics of current-control type induction-type bearingless motor decouple System with Sliding Mode Controller |
CN109217766B (en) * | 2018-09-26 | 2021-07-23 | 河南科技大学 | Independent inverse decoupling control system of bearingless asynchronous motor |
-
2019
- 2019-05-22 CN CN201910431278.7A patent/CN110048657B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN110048657A (en) | 2019-07-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110429881B (en) | Active-disturbance-rejection control method of permanent magnet synchronous motor | |
CN100433537C (en) | Method for controlling bearing-less AC asynchronous motor neural network inverse decoupling controller | |
Bu et al. | Sliding mode variable structure control strategy of bearingless induction motor based on inverse system decoupling | |
CN110048657B (en) | Decoupling control method for neural network inverse system of bearingless asynchronous motor | |
CN110705174B (en) | Five-degree-of-freedom magnetic suspension motorized spindle rotor displacement self-detection system and method | |
CN110138299A (en) | Induction-type bearingless motor reversed decoupling control system based on rotor resistance on-line identification | |
CN109217766A (en) | The independent reversed decoupling control system of induction-type bearingless motor | |
CN110649845B (en) | Photoelectric turntable position tracking control method based on robust generalized predictive control | |
CN105375848A (en) | Permanent magnet synchronous motor self-adaptive identification control method and control system thereof | |
CN105071730A (en) | Bearing-free asynchronous motor stator orientated reverse closed-loop control system taking current dynamics into consideration | |
CN102097986A (en) | Construction method for neural network generalized inverse decoupling controller of bearing-free synchronous reluctance motor | |
CN102136822B (en) | Five-DOF (freedom of degree) bearingless synchronous reluctance motor decoupling controller and construction method thereof | |
CN112422007A (en) | Construction method of least square support vector machine optimization control system of hybrid magnetic bearing | |
Feng et al. | Speed control of induction motor servo drives using terminal sliding-mode controller | |
CN102790580B (en) | The building method of induction-type bearingless motor SVMs inverse decoupling controller | |
CN102130647B (en) | Speed-sensor-less construction method for detecting rotating speed of bearing-less asynchronous motor | |
CN110131312B (en) | Five-degree-of-freedom alternating current active magnetic bearing active disturbance rejection decoupling controller and construction method | |
CN109600083A (en) | Two degrees of freedom bearing-free permanent magnet synchronous motor suspending power subsystem decoupled controller | |
CN109150045A (en) | The independent Inverse Decoupling method of induction-type bearingless motor | |
CN105071729A (en) | Bearing-free asynchronous motor stator magnetic flux linkage orientated reverse decoupling method taking current dynamics into consideration | |
CN202004708U (en) | Bearingless brushless direct current motor neural network alpha-order inverse controller | |
CN110504884A (en) | A kind of induction-type bearingless motor radial force suspension control system based on differential geometrical decoupled control | |
CN201928221U (en) | Neural net generalized inverse decoupling controller for bearingless synchronous reluctance motor | |
Bu et al. | A control method of bearingless induction motor based on neural network | |
CN214707585U (en) | Hybrid magnetic bearing least square support vector machine optimization decoupling controller |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20201016 |
|
CF01 | Termination of patent right due to non-payment of annual fee |