CN114448311A - Fuzzy neural network prediction decoupling control system for bearingless permanent magnet synchronous generator - Google Patents
Fuzzy neural network prediction decoupling control system for bearingless permanent magnet synchronous generator Download PDFInfo
- Publication number
- CN114448311A CN114448311A CN202210077868.6A CN202210077868A CN114448311A CN 114448311 A CN114448311 A CN 114448311A CN 202210077868 A CN202210077868 A CN 202210077868A CN 114448311 A CN114448311 A CN 114448311A
- Authority
- CN
- China
- Prior art keywords
- neural network
- fuzzy neural
- input
- permanent magnet
- voltage
- 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.)
- Pending
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 152
- 230000001360 synchronised effect Effects 0.000 title claims abstract description 50
- 238000006073 displacement reaction Methods 0.000 claims abstract description 76
- 238000004804 winding Methods 0.000 claims abstract description 40
- 238000000034 method Methods 0.000 claims abstract description 24
- 239000000725 suspension Substances 0.000 claims abstract description 23
- 238000004364 calculation method Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 11
- 238000005339 levitation Methods 0.000 claims description 5
- 230000001537 neural effect Effects 0.000 claims description 4
- 230000005284 excitation Effects 0.000 claims description 2
- 238000003062 neural network model Methods 0.000 claims description 2
- 238000010248 power generation Methods 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 7
- 230000001276 controlling effect Effects 0.000 description 5
- 230000004044 response Effects 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/0003—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control
- H02P21/0014—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
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/0003—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control
- H02P21/001—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control using fuzzy control
-
- 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
- H02P25/00—Arrangements or methods for the control of AC motors characterised by the kind of AC motor or by structural details
- H02P25/02—Arrangements or methods for the control of AC motors characterised by the kind of AC motor or by structural details characterised by the kind of motor
- H02P25/022—Synchronous motors
- H02P25/024—Synchronous motors controlled by supply frequency
- H02P25/026—Synchronous motors controlled by supply frequency thereby detecting the rotor position
-
- 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
- H02P27/00—Arrangements or methods for the control of AC motors characterised by the kind of supply voltage
- H02P27/04—Arrangements or methods for the control of AC motors characterised by the kind of supply voltage using variable-frequency supply voltage, e.g. inverter or converter supply voltage
- H02P27/06—Arrangements or methods for the control of AC motors characterised by the kind of supply voltage using variable-frequency supply voltage, e.g. inverter or converter supply voltage using dc to ac converters or inverters
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Automation & Control Theory (AREA)
- Fuzzy Systems (AREA)
- Control Of Eletrric Generators (AREA)
Abstract
The invention discloses a fuzzy neural network prediction decoupling control system of a bearingless permanent magnet synchronous generator, which is formed by connecting a neural network prediction controller with a fuzzy neural network inverse controller, wherein the fuzzy neural network inverse controller is formed by a fuzzy neural network system, two analog switch signal modulation modules and an IGBT three-phase inverter; the neural network prediction controller predicts the eccentric displacement control quantity and the voltage control quantity at the next moment, the fuzzy neural network system outputs a suspension force winding reference voltage and a power generation winding reference voltage, the suspension force winding reference voltage is input into a first analog switch signal modulation module, and the power generation winding reference voltage is input into a second analog switch signal modulation module; the predicted displacement at the next moment is adopted, so that the method not only has the advantages of simple structure and strong robustness of an inverse system, but also combines the advantages of fuzzy logic control, and reliably realizes decoupling control between the radial suspension force and the generating voltage of the generator.
Description
Technical Field
The invention relates to a bearingless permanent magnet synchronous generator, in particular to a prediction decoupling control system of the bearingless permanent magnet synchronous generator, which is used for controlling the bearingless permanent magnet synchronous generator and is widely applied to the fields of wind driven generators, gas turbine generators, hybrid electric vehicles, electric vehicle flywheel energy storage systems, aerospace, power generation and the like.
Background
The traditional PID linear control method adopts direct torque control to carry out double closed loop decoupling control on the generator, although the decoupling can be carried out on the generator, the structure of the control system is complex, and in actual operation, due to the complexity of a generator system, accurate parameters are difficult to obtain and a satisfactory effect is difficult to obtain. The nonlinear decoupling control is a key for realizing the stable operation of the bearingless permanent magnet synchronous generator, and common intelligent nonlinear decoupling control methods comprise a neural network control method and a fuzzy logic control method, wherein the neural network control method has better self-learning and association capabilities, less manual intervention is performed, but the requirement on the number of samples is high, the learning period is long, the design cost is high, and fuzzy information cannot be processed. For example, the neural network adaptive inverse controller disclosed in chinese patent publication No. CN103647481 needs to adopt a method of separately decoupling control for two radial displacements x and y, and this method needs to perform segment processing on the rotor displacement, and is highly influenced by model accuracy. The fuzzy logic control method has a simple reasoning process, low sample requirement, more manual intervention and low reasoning learning speed. For example, the controller disclosed in chinese patent publication No. CN101630940 entitled "fuzzy neural network inverse robust controller for speed regulating system of induction motor and construction method" adopts an adaptive neural fuzzy inference system, which is a process of fuzzification of input and output variables, and the method has the disadvantages that each variable needs to correspond to many membership functions, which not only results in a complex structure, but also reduces the learning rate.
Disclosure of Invention
The invention aims to solve the problems of the existing bearingless permanent magnet synchronous generator control technology, provides a bearingless permanent magnet synchronous generator fuzzy neural network prediction decoupling control system which is simple in structure, strong in robustness, low in requirement on samples and high in learning speed, and combines the advantages of fuzzy logic control, neural network prediction and inverse system.
The technical scheme adopted by the fuzzy neural network prediction decoupling control system of the bearingless permanent magnet synchronous generator is as follows: the fuzzy neural network inverse controller consists of a fuzzy neural network system, two analog switch signal modulation modules and an IGBT three-phase inverter; radial displacement x, y and generated voltage u of bearingless permanent magnet synchronous generator at current k moment and radial given displacement { x*,y*Given voltage u of motor*The two are used as input quantity of a neural network prediction controller together, and the neural network prediction controller predicts an eccentric displacement control quantity r and a voltage control quantity u at the k +1 momentcThe eccentric displacement control amount r and the voltage control amount ucThe eccentric angle theta of the permanent magnet synchronous generator without the bearing is input into a fuzzy neural network system together, and the fuzzy neural network system outputs the reference voltage of the k +1 moment suspension force winding under an alpha-beta coordinate system
And a reference voltage of the generating winding
Reference voltage for levitation winding
The reference voltage of the power generation winding is input into a first analog switch signal modulation module
Inputting the input signal into a second analog switch signal modulation module, and outputting a k +1 time switch signal S by the first analog switch signal modulation module1Switching signal S1Outputting three-phase current { i } of the suspension force winding after passing through the IGBT three-phase inverter1a,i1b,i1cThe second analog switch signal modulation module outputs a k +1 time switch signal S2Switching signal S2And outputting the generated voltage u after passing through a three-phase rectifier.
Furthermore, the fuzzy neural network system consists of a second order difference processor, a first order difference processor and a fuzzy neural network, wherein the eccentric displacement control quantity r is input into the second order difference processor, and is processed by the second order difference processor to obtain a first order difference quantity of the eccentric displacement control quantity r at the moment of k +1
Second order difference component
Eccentric displacement control quantity r and first order difference quantity thereof
Second order difference component
Inputting the signals into a fuzzy neural network together; the voltage control quantity ucInputting the voltage control quantity u to a first-order difference processor, and processing the voltage control quantity u by the first-order difference processor to obtain the voltage control quantity u at the k +1 momentcFirst order difference component of
Voltage control quantity ucAnd first order difference component
Are commonly input into the fuzzy neural network 11
The invention has the advantages that after the technical scheme is adopted:
1. the invention adopts the neural network prediction controller and the fuzzy neural network inverse controller, predicts the displacement of the next moment through the prediction capability of the neural network aiming at the rotor displacement control, then adopts the rotor eccentric displacement control quantity and the voltage control quantity as control signals through calculation, compared with the method of directly taking radial displacement as input quantity, because the predicted displacement of the next moment is adopted, the delay in the signal transmission process is eliminated, and the eccentric displacement reflects the rotor position more intuitively, the direct control of the rotor eccentricity can be realized, the suspension force control structure is simplified, and the influence of the model accuracy is reduced. The voltage control quantity is used as a control signal, and compared with a current signal, the voltage control has higher response speed and better dynamic performance.
2. The fuzzy neural network inverse controller adopted by the invention is constructed based on the inverse system principle, has the advantages of clear and visual physical concept and simple and clear mathematical analysis process, and simultaneously only comprises a fuzzy neural network system, an analog switch signal modulation module and an IGBT three-phase inverter, so that a large number of coordinate transformation modules and feedback modules are saved, the control cost can be effectively reduced, and the control efficiency is improved.
3. The neural network prediction controller adopted by the invention predicts the next moment aiming at the given signal and the detection signal to obtain the displacement variation and the voltage variation of the next moment, and compared with the simple difference of the given signal and the detection signal, the neural network prediction controller can effectively reduce the response delay of the system, improve the response speed of the control system and eliminate the static error.
4. The invention determines the weight value in the neural network structure and applies the T-S fuzzy inference model to carry out fuzzification processing, thereby reducing the number of membership function and simplifying the structure.
5. The invention not only has the advantages of simple structure and strong robustness of the inverse system, but also combines the advantages of low requirements of fuzzy logic control on samples and rapid learning and training of a neural network, thereby having great advantages in processing the complex control with high uncertainty of the bearingless permanent magnet synchronous generator, reliably realizing the decoupling control between the radial suspension force and the generating voltage of the bearingless permanent magnet synchronous generator, obtaining good control performance of the radial position of a rotor and the generating voltage, and simultaneously improving the capability of the control system for resisting parameter change and load disturbance.
Drawings
FIG. 1 is a block diagram of the present invention;
FIG. 2 is a block diagram of the architecture of the fuzzy neural network system of FIG. 1;
in the figure: 1. a bearingless permanent magnet synchronous generator; IGBT three-phase inverter; 3. a three-phase rectifier; 4. 5, simulating a switch signal modulation module; 6. a fuzzy neural network system; 7. a neural network predictive controller; 8. an angle calculation module; 9. a speed sensor; 10. a fuzzy neural network inverse controller; 11. a fuzzy neural network; 12. a second order difference processor; 13. a first order difference processor.
Detailed Description
Referring to fig. 1, the fuzzy neural network prediction decoupling control system of the bearingless permanent magnet synchronous generator is formed by connecting a neural network prediction controller 7 and a fuzzy neural network inverse controller 10, is connected with the bearingless permanent magnet synchronous generator 1, and is used for controlling the bearingless permanent magnet synchronous generator 1. The neural network prediction controller 7 is connected with a fuzzy neural network inverse controller 10 in series, and the fuzzy neural network inverse controller 10 is connected with the bearingless permanent magnet synchronous generator 1 in series.
The fuzzy neural network inverse controller 10 is composed of a fuzzy neural network system 6, two analog switch signal modulation modules 4 and 5 and an IGBT three-phase inverter 2. The input end of a fuzzy neural network system 6 is connected with a neural network prediction controller 7, the output end of the fuzzy neural network system 6 is respectively connected with the input ends of two analog switch signal modulation modules 4 and 5, the output end of the first analog switch signal modulation module 4 is connected with the input end of an IGBT three-phase inverter 2, the output end of the second analog switch signal modulation module 5 is connected with the input end of a three-phase rectifier 3, and the output end of the three-phase rectifier 3 is connected with the input end of the neural network prediction controller 7. The output end of the IGBT three-phase inverter 2 is connected with the bearingless permanent magnet synchronous generator 1, the input end of the three-phase rectifier 3 is connected with the output end of the bearingless permanent magnet synchronous generator 1, and the output end of the three-phase rectifier is connected with the neural network prediction controller 7.
The bearingless permanent magnet synchronous generator 1 adopts two displacement sensors to respectively acquire radial displacements x and y in x and y directions, and the two radial displacements x and y are input into the neural network prediction controller 7 and respectively used as a first input quantity and a second input quantity of the neural network prediction controller 7.
The bearingless permanent magnet synchronous generator 1 adopts the speed sensor 9 to acquire the rotating speed omega thereof, and the output end of the speed sensor 9 is connected with the input end of the fuzzy neural network system 6 through the angle calculation module 8. The rotation speed ω is calculated by the angle calculation module 8 through the angle calculation module 8, and the eccentricity θ is ω t, and t is time. The eccentricity angle θ is input to the fuzzy neural network system 6 as a first input quantity of the fuzzy neural network system 6, and is also a first input quantity of the fuzzy neural network inverse controller 10.
Three-phase generating winding voltage { u } of bearingless permanent magnet synchronous generator 12a,u2b,u2cAnd the generated voltage u is input into the neural network prediction controller 7 as a third input quantity of the neural network prediction controller 7.
Given radial displacement x in x and y directions at current k time*,y*Given voltage u of motor*The fourth, fifth and sixth input quantities respectively used by the neural network prediction controller 7, the radial displacement x and y at the time k detected by the displacement sensor and the generated voltage u are used as the input quantities of the neural network prediction controller 7 together, and the eccentric displacement control quantity r and the voltage control quantity u at the next time k +1 are predicted by the prediction calculation of the neural network prediction controller 7c. The eccentric displacement control amount r and the voltage control amount u at the moment k +1 are calculatedcThe input signals are input into the fuzzy neural network system 6 together as the second and third input quantities of the fuzzy neural network system 6, and are also the second and third input quantities of the fuzzy neural network inverse controller 10.
The input node of the fuzzy neural network system 6 is 3, and the output node is 4. The input quantity of the fuzzy neural network system 6 is the input quantity of the fuzzy neural network inverse controller 10, and the fuzzy neural network system6 outputting two groups of voltage signals which are respectively the reference voltage of the levitation force winding at the k +1 moment under the alpha-beta coordinate system
And generating winding reference voltage at the time k +1 under an alpha-beta coordinate system
Reference voltage of k +1 time suspension force winding
The reference voltage of the generating winding at the moment of k +1 is input into a first analog switch signal modulation module 4
Input into a second analog switching signal modulation module 5. The two groups of reference voltage signals respectively output corresponding switch signals after passing through the corresponding analog switch signal modulation modules 4 and 5: the first analog switch signal modulation module 4 outputs a switching signal S at the k +1 moment1＝(S1a,S1b,S1c) Switching signal S1＝(S1a,S1b,S1c) The input is to the IGBT three-phase inverter 2, and the IGBT three-phase inverter 2 is controlled. The second analog switch signal modulation module 5 outputs a switching signal S at the k +1 moment2＝(S2a,S2b,S2c) Switching signal S2＝(S2a,S2b,S2c) The voltage is input to the three-phase rectifier 3, and the three-phase rectifier 3 is controlled.
Switching signal S1＝(S1a,S1b,S1c) As the first input quantity of the IGBT three-phase inverter 2, the inverter is controlled to work to convert the DC bus voltage UDCThe second input quantity is input to the IGBT three-phase inverter 2 as the second input quantity of the IGBT three-phase inverter 2. The IGBT three-phase inverter 2 is controlled by a switching signal to work, and outputs three-phase current { i } of the k +1 moment suspension force winding1a,i1b,i1cAnd (4) the suspension control device is arranged in the bearingless permanent magnet synchronous generator 1 and used for controlling the suspension of the bearingless permanent magnet synchronous generator 1.
Three-phase rectifier 3 with switching signal S2＝(S2a,S2b,S2c) And the three-phase generation winding voltage { u } of the bearingless permanent magnet synchronous generator 12a,u2b,u2cAnd the generated voltage u is used as input and output to the neural network prediction controller 7 to control the stability of the generated voltage of the bearingless permanent magnet synchronous generator 1.
The output quantity of the fuzzy neural network inverse controller 10 is the switching signal S at the k +1 moment output by the second analog switching signal modulation module 52＝(S2a,S2b,S2c) And k +1 moment suspension force winding three-phase current { i) output by IGBT three-phase inverter 21a,i1b,i1cTriple-phase current { i) of suspension force winding1a,i1b,i1cThe input is to the bearingless permanent magnet synchronous generator 1 to control the suspension thereof.
The mathematical model of the bearingless permanent magnet synchronous generator 1 is a 5-order differential matrix equation with the relative vector order of {2,2,1 }. The intercector algorithm verifies that the bearingless permanent magnet synchronous generator 1 is reversible, namely a right inverse system exists, the right inverse system is constructed by referring to the structure of the neural network, and the fuzzy neural network inverse controller 10 can be constructed according to the method.
As shown in fig. 2, the fuzzy neural network system 6 is composed of a second order difference processor 12, a first order difference processor 13 and a fuzzy neural network 11, and the output terminals of the second order difference processor 12 and the first order difference processor 13 are respectively connected in series with the input terminal of the fuzzy neural network 11. The eccentric displacement control quantity r at the time of k +1 is input into a second-order difference processor 12, and the first-order difference quantity of the eccentric displacement control quantity r at the time of k +1 is obtained after the processing of the second-order difference processor 12
Second order difference component
The eccentric displacement control quantity r at the k +1 moment and the first order difference quantity thereof
Second order difference component
Are input together into the fuzzy neural network 11. Voltage control u at time k +1cInputting the voltage to a first order difference processor 13, and processing the voltage by the first order difference processor 13 to obtain a voltage control value u at the time of k +1cFirst order difference component of
Controlling the voltage at the moment k +1 by a quantity ucAnd first order difference component
Are input together into the fuzzy neural network 11. The eccentric displacement control quantity r at the k +1 moment and the first order difference quantity thereof
Second order difference component
And the voltage control amount u at the time k +1cAnd its first order difference component
And the eccentric angle theta signal output by the angle calculation module 8 is used as the input quantity of the fuzzy neural network 11 together, and two groups of voltages are obtained through calculation of the fuzzy neural network 11 and are respectively the reference voltage of the k +1 moment suspension force winding
And a reference voltage of the generation winding
Suspension force winding reference voltage at k +1 moment
Input to the first analog switching signalIn the modulation module 4, the first analog switch signal modulation module 4 modulates the input k +1 time suspension force winding reference voltage
The signal is processed, and the specific processing process is as follows:
based on the reference voltage of the suspension force winding
Three voltage intermediate variables V are calculated according to a formulaa,Vb,VcIn which
Then passing through three voltage intermediate variables Va,Vb,VcObtaining an intermediate variable Va,Vb,VcMaximum value of (V)maxMinimum value VminAnd the mean value Vcomm: maximum value Vmax＝max{Va,Vb,VcV, minimum value Vmin＝min{Va,Vb,VcMean value Vcomm＝(Vmax+Vmi)n2; finally, three voltage intermediate variables Va,Vb,VcAnd the mean value VcommCalculating to obtain a switching signal S1＝(S1a,S1b,S1c) In which S is1a＝Va-Vcom，S1b＝Vb-Vcomm，S1c＝Vc-Vcomm。
Reference voltage of generating winding at time k +1
The voltage is input into a second analog switch signal modulation module 5, and the second analog switch signal modulation module 5 generates a winding reference voltage at the input k +1 moment
Signal processing, second analog switch signal modulation moduleThe specific processing procedure of 5 is the same as that of the first analog switch signal modulation module 4, and only the reference voltage of the generating winding is used
Reference voltage substituted for levitation force winding
I.e. based on the reference voltage of the generating winding
Firstly, three voltage intermediate variables are calculated, then the maximum value, the minimum value and the average value of the intermediate variables are obtained through the three voltage intermediate variables, and finally the switching signal S is obtained through the calculation of the voltage intermediate variables and the average value2＝(S2a,S2b,S2c)。
Suspending the force winding reference voltage at the k +1 moment
And a reference voltage of the generating winding
Respectively input into the analog switch signal modulation module 4 and the analog switch signal modulation module 5, and switch control signals S output by the analog switch signal modulation module 4 and the analog switch signal modulation module 51,S2And respectively controlling the IGBT three-phase inverter 2 and the three-phase rectifier 3, thereby realizing the control of the bearingless permanent magnet synchronous generator 1.
The neural network prediction controller 7 adopts PI parameters to input quantity as radial given displacement { x*,y*Given voltage u*And the radial displacement x and y at the moment k and the signal of the generated voltage u are processed, and the specific processing process is as follows:
according to a given radial displacement x*,y*Calculating a displacement intermediate quantity e by radial displacements x and y at the moment k, wherein the calculation formula is as follows: e ═ x-x*)2+(y-y*)2(ii) a According to the middle of the displacementError proportionality coefficient K in quantity e and PI parameterPAnd the integral coefficient of error KiAnd (3) calculating the eccentric displacement control amount r at the moment of k +1, wherein the calculation formula is as follows:
given quantity u according to generated voltage*Generating voltage u at the moment of summation K and error proportionality coefficient KPAnd the integral coefficient of error KiCalculating the voltage control amount u at the moment of k +1cThe calculation formula is as follows:
error proportionality coefficient KPIs generally in the range of 10-30, and the error integral coefficient KiIs in the range of 0.01-0.9.
A fuzzy neural network 11 with 6 input nodes and 4 output nodes, plus 1 second order difference processor 12 and 1 first order difference processor 13 are used to construct a fuzzy neural system 6 with 3 input nodes and 4 output nodes, where: the fuzzy neural network 11 adopts a 5-layer self-adaptive neural fuzzy inference system (referred to as a fuzzy neural network for short), the number of input nodes is 6, the number of output layer nodes is 4, the mean square error of an error index selection sample, the membership functions of input and output variables adopt bell-shaped functions, each input adopts 15 membership functions, and the type of the output function is linear. The first output r of the second order difference processor 12 is used as the first input of the fuzzy neural network 11; second output of second order difference processor 12
As a second input to the fuzzy neural network 11; third output of second order difference processor 12
As a third input to the fuzzy neural network 11. At a step differenceFirst output u of processor 13cAs a fourth input to the fuzzy neural network 11; second output of first order difference processor 13
As a fifth input to the fuzzy neural network 11; the eccentric angle θ output by the angle calculation module 8 is used as the sixth input of the fuzzy neural network 11. The output of the fuzzy neural network 11 is the output of the fuzzy neural network system 6, namely the suspension force winding reference voltage at the k +1 moment
And a reference voltage of the generation winding
The fuzzy neural network 11 is obtained by training a training sample, and specifically comprises the following steps: 1) step excitation signal
And
the radial displacement x and the radial displacement y of the rotor of the bearingless permanent magnet synchronous generator 1 are obtained by respectively adding the radial displacement x and the radial displacement y to the input ends of the first analog switching signal modulation module 4 and the second analog switching signal modulation module 5, the rotating speed omega is detected by the speed sensor 9, and the three-phase rectifier 3 outputs the generated voltage u. 2) Processing the obtained rotating speed omega through an angle calculating module 8 to obtain an eccentric angle theta; inputting the obtained radial displacement x and y of the rotor and the generated voltage u into a neural network prediction controller 7 for processing, and acquiring the eccentric displacement control quantity r and the voltage control quantity u at the moment of k +1 processed by the neural network prediction controller 7cRespectively calculating the first derivative and the second derivative of the eccentric displacement control amount r, and calculating the voltage control amount ucThe first derivative is calculated. Controlling the eccentric displacement r and the first derivative thereof
And second derivative
Voltage control quantity ucAnd its first derivative
Training sample set forming fuzzy neural network together with eccentric angle theta
3) And training the fuzzy neural network 11 by adopting a hybrid algorithm, wherein the output mean square error of the fuzzy neural network is less than 0.001 after about 1000 times of training, so that the requirements are met, and all parameters and weight coefficients of the fuzzy neural network are determined. The model from which the fuzzy neural network 11 is derived is:
wherein U is the output quantity,
f is the fuzzy neural network model obtained by training.
The neural network prediction controller 7 calculates the calculated eccentric displacement control amount r and voltage control amount u at the time of k +1cThe input is input into the fuzzy neural network inverse controller 10. The eccentric displacement control amount r at the time k +1 is input to a second order difference processor 12 in the fuzzy neural network system 6, and the second order difference processor 12 processes the input eccentric displacement control amount r at the time k + 1: first order difference component of eccentric displacement control quantity r at k +1 time
The control method is obtained by calculating an eccentric displacement control quantity r (k-3) at the time of k-3, an eccentric displacement control quantity r (k-2) at the time of k-2, an eccentric displacement control quantity r (k) at the time of k and an eccentric displacement control quantity r (k +1) at the time of k +1, and the calculation formula is as follows:
second order difference component of eccentric displacement control quantity r at k +1 time
The control method is obtained by calculating an eccentric displacement control quantity r (k-3) at the moment k-3, an eccentric displacement control quantity r (k-2) at the moment k-2, an eccentric displacement control quantity r (k-1) at the moment k-1, an eccentric displacement control quantity r (k) at the moment k and an eccentric displacement control quantity r (k +1) at the moment k +1, and the calculation formula is as follows:
voltage control amount u at time k +1cThe voltage control amount u at the input time k +1 is inputted to a first-order difference processor 13 in the fuzzy neural network system 6, and the first-order difference processor 13 performs a voltage control on the voltage control amount u at the input time k +1cAnd (3) processing signals: voltage control amount u at time k +1cFirst order difference component of
Controlled by the voltage at time k-3cVoltage control amount u at time (k-3) and k-2c(k-2) and a voltage control amount u at the time kc(k) And the voltage control amount u at the time k +1c(k +1) is calculated, and the calculation formula is as follows:
the method comprises the steps of firstly establishing a mathematical model of a bearingless permanent magnet synchronous generator 1 of a controlled object, designing a neural network prediction controller 7, constructing a fuzzy neural network system 6, training the fuzzy neural network 11 by collecting sample data, forming a fuzzy neural network inverse controller 10 by two analog switch signal modulation modules 4 and 5 and the fuzzy neural network system 6, and connecting the neural network prediction controller 7 and the fuzzy neural network inverse controller 10 in series to form a complete fuzzy neural network prediction decoupling controller.
When the invention works, the radial displacement is given by a given amount { x*,y*V. voltage given quantity u*Inputting the radial displacement x and y and the voltage u at the moment k into a neural network prediction controller 7, and outputting an eccentric displacement control quantity r and a voltage control quantity u at the moment k +1c. Then the eccentric displacement control amount r and the voltage control amount u at the time of k +1 outputted from the neural network prediction controller 7cInput into the fuzzy neural network system 6. The first analog switch modulation module 4 and the second analog switch modulation module 5 output control signals at the k +1 moment according to the fuzzy neural network 11
And
and the IGBT three-phase inverter 2 and the three-phase rectifier 3 are respectively controlled, so that the radial suspension force and the generating voltage of the bearingless permanent magnet synchronous generator 1 are controlled. Because the fuzzy neural network 1 adopts the fuzzification algorithm, a fuzzy function similar to human expert brain reasoning is added in the system, and a part of membership function in the common neural network is replaced, so that the fuzzy neural network reduces the number of the membership function and simplifies the structure compared with the common neural network. The neural network system 6 is connected in series before the original system, the definition of the inverse system is known, if the inverse system can be constructed, the original nonlinear original system can realize pseudo linearization by connecting the inverse system in series before the original system, and the input-output relationship of the original nonlinear original system is simplified from the original complex function mapping relationship into the simple integral linear relationship, so that the decoupling control between the generating voltage and the radial suspension force is realized.
Claims (10)
1. A fuzzy neural network prediction decoupling control system of a bearingless permanent magnet synchronous generator is characterized in that: the fuzzy neural network inverter is formed by connecting a neural network prediction controller (7) and a fuzzy neural network inverse controller (10), wherein the fuzzy neural network inverse controller (10) is formed by a fuzzy neural network system (6), two analog switch signal modulation modules (4 and 5) and an IGBT three-phase inverter (2); radial position of bearingless permanent magnet synchronous generator at current k momentShift x, y and generated voltage u and radial given displacement x*,y*Given voltage u of motor*The two are used as input quantity of a neural network prediction controller (7) together, and the neural network prediction controller (7) predicts an eccentric displacement control quantity r and a voltage control quantity u at the moment of k +1cThe eccentric displacement control amount r and the voltage control amount ucThe eccentric angle theta of the permanent magnet synchronous generator without the bearing is input into a fuzzy neural network system (6) together, and the fuzzy neural network system (6) outputs the reference voltage of the k +1 moment suspension force winding under an alpha-beta coordinate system
And a reference voltage of the generating winding
Reference voltage for levitation winding
Input into a first analog switch signal modulation module (4) to generate a winding reference voltage
The input signal is input into a second analog switch signal modulation module (5), and the first analog switch signal modulation module (4) outputs a k +1 time switch signal S1Switching signal S1Outputs three-phase current { i) of the suspension force winding after passing through an IGBT three-phase inverter (2)1a,i1b,i1cThe second analog switch signal modulation module (5) outputs a k +1 time switch signal S2Switching signal S2And the generated voltage u is output after passing through the three-phase rectifier (3).
2. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 1, characterized in that: the fuzzy neural network system (6) consists of a second-order difference processor (12), a first-order difference processor (13) and a fuzzy neural network (11), and consists ofThe eccentric displacement control quantity r is input into a second-order difference processor (12), and the first-order difference quantity of the eccentric displacement control quantity r at the k +1 moment is obtained after the processing of the second-order difference processor (12)
Second order difference component
Eccentric displacement control quantity r and first order difference quantity thereof
Second order difference component
The input is jointly input into a fuzzy neural network (11); the voltage control quantity ucInput into a first order difference processor (13), and processed by the first order difference processor (13) to obtain a voltage control quantity u at the time of k +1cFirst order difference component of
Voltage control quantity ucAnd first order difference component
Are jointly input into a fuzzy neural network (11).
3. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 1, characterized in that: the first analog switch signal modulation module (4) calculates a switch signal S according to a formula1＝(S1a,S1b,S1c) In which S is1a＝Va-Vcomm，S1b＝Vb-Vcomm，S1c＝Vc-Vcomm，
Average value Vcomm＝(Vmax+Vmin) /2, maximum value Vmax＝max{Va,Vb,VcV, minimum value Vmin＝min{Va,Vb,Vc}; the second analog switch signal modulation module (5) calculates the switch signal S2＝(S2a,S2b,S2c) The method is similar to that of the first analog switching signal modulation module (4), and only the reference voltage of the generating winding is used
Replacing levitation force winding reference voltage
4. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 1, characterized in that: the neural network predictive controller (7) is based on the formula
Calculating the eccentric displacement control amount r at the moment of k +1 according to the formula
Calculating the voltage control amount u at the moment of k +1cThe intermediate displacement e ═ x-x*)2+(y-y*)2Error proportionality coefficient KPIs in the range of 10-30, and an error integral coefficient KiIs in the range of 0.01-0.9.
5. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 2, characterized in that: the second order difference processor (12) calculates the first order difference component according to a formula
Calculating the second order difference component according to the formula
r (k-3) is the eccentric displacement control quantity at the time of k-3 r (k-3), r (k-2) is the eccentric displacement control quantity at the time of k-2, the eccentric displacement control quantity at the time of k, r (k-1) is the eccentric displacement control quantity at the time of k-1, and r (k +1) is the eccentric displacement control quantity at the time of k + 1.
6. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 2, characterized in that: the first order difference processor (13) calculates the first order difference component according to a formula
uc(k-3) is a voltage control amount at the time k-3, uc(k-2) is a voltage control amount at the time of k-2, uc(k) Is the voltage control quantity at the time k, ucAnd (k +1) is a voltage control amount at the time k + 1.
7. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 2, characterized in that: the fuzzy neural network (11) adopts a 5-layer self-adaptive neural fuzzy inference system, the number of input nodes is 6, the number of output layer nodes is 4, the mean square error of an error index selection sample, the membership functions of input and output variables adopt bell-shaped functions, each input adopts 15 membership functions, and the type of the output function is linear.
8. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 7, wherein: step excitation signal
And
respectively added to the input ends of a first analog switch signal modulation module (4) and a second analog switch signal modulation module (5) to acquire an eccentric displacement control quantity r and a voltage control quantity u at the k +1 moment which are obtained by processing of a neural network prediction controller (7)cForming a training sample set of the fuzzy neural network
Training the fuzzy neural network (11) to obtain a model thereof:
output quantity
f is the fuzzy neural network model obtained by training.
9. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 1, characterized in that: the rotating speed omega of the bearing permanent magnet synchronous generator is acquired by adopting a speed sensor (9), the output end of the speed sensor (9) is connected with the input end of a fuzzy neural network system (6) through an angle calculation module (8), the rotating speed omega is calculated to obtain an eccentric angle theta which is omega t through the angle calculation module (8), and t is time.
10. The fuzzy neural network predictive decoupling control system of the bearingless permanent magnet synchronous generator according to claim 1, characterized in that: the input end of the three-phase rectifier (3) is respectively connected with the bearingless permanent magnet synchronous generator and a second analog switch signal modulation module (5), and the switch signal S2And three-phase generating winding voltage { u } of bearingless permanent magnet synchronous generator2a,u2b,u2cAnd the three-phase rectifier (3) is used as an input quantity together.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210077868.6A CN114448311A (en) | 2022-01-24 | 2022-01-24 | Fuzzy neural network prediction decoupling control system for bearingless permanent magnet synchronous generator |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210077868.6A CN114448311A (en) | 2022-01-24 | 2022-01-24 | Fuzzy neural network prediction decoupling control system for bearingless permanent magnet synchronous generator |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114448311A true CN114448311A (en) | 2022-05-06 |
Family
ID=81369598
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210077868.6A Pending CN114448311A (en) | 2022-01-24 | 2022-01-24 | Fuzzy neural network prediction decoupling control system for bearingless permanent magnet synchronous generator |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114448311A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050237034A1 (en) * | 2004-04-23 | 2005-10-27 | Patterson Stanley C | Fault tolerant architecture for permanent magnet starter generator subsystem |
| CN101546978A (en) * | 2009-05-06 | 2009-09-30 | 南京航空航天大学 | Control method of complete-period generator of bearing-free switch reluctance motor |
| CN103414428A (en) * | 2013-08-05 | 2013-11-27 | 江苏大学 | Bearingless synchronous reluctance motor rotor eccentric displacement controller and construction method thereof |
| CN106330041A (en) * | 2016-09-13 | 2017-01-11 | 江苏大学 | Mixed excitation type bearingless generator direct prediction controller based on magnetic flux linkage observation |
| CN107276473A (en) * | 2017-06-29 | 2017-10-20 | 江苏大学 | Permanent-magnet synchronous motor with five degrees of freedom without bearing Fuzzy Neural Network Decoupling controller |
| CN107370429A (en) * | 2017-06-29 | 2017-11-21 | 江苏大学 | Bearing-free permanent magnet synchronous motor fuzzy neural network inverse decoupling controller |
| CN113090461A (en) * | 2021-05-13 | 2021-07-09 | 曲阜师范大学 | Low wind speed vertical axis wind turbine suspension control method based on sliding mode neural network model prediction |
-
2022
- 2022-01-24 CN CN202210077868.6A patent/CN114448311A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050237034A1 (en) * | 2004-04-23 | 2005-10-27 | Patterson Stanley C | Fault tolerant architecture for permanent magnet starter generator subsystem |
| CN101546978A (en) * | 2009-05-06 | 2009-09-30 | 南京航空航天大学 | Control method of complete-period generator of bearing-free switch reluctance motor |
| CN103414428A (en) * | 2013-08-05 | 2013-11-27 | 江苏大学 | Bearingless synchronous reluctance motor rotor eccentric displacement controller and construction method thereof |
| CN106330041A (en) * | 2016-09-13 | 2017-01-11 | 江苏大学 | Mixed excitation type bearingless generator direct prediction controller based on magnetic flux linkage observation |
| CN107276473A (en) * | 2017-06-29 | 2017-10-20 | 江苏大学 | Permanent-magnet synchronous motor with five degrees of freedom without bearing Fuzzy Neural Network Decoupling controller |
| CN107370429A (en) * | 2017-06-29 | 2017-11-21 | 江苏大学 | Bearing-free permanent magnet synchronous motor fuzzy neural network inverse decoupling controller |
| CN113090461A (en) * | 2021-05-13 | 2021-07-09 | 曲阜师范大学 | Low wind speed vertical axis wind turbine suspension control method based on sliding mode neural network model prediction |
Non-Patent Citations (2)
| Title |
|---|
| YIZHOU HUA ET AL.: "IEEE Transactions on Applied Superconductivity", pages 1 - 5 * |
| 姜宏伟 等: "基于改进粒子群优化的主动磁悬浮轴承模糊PID控制", no. 32, pages 35 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hannan et al. | Switching techniques and intelligent controllers for induction motor drive: Issues and recommendations | |
| CN107370429B (en) | Fuzzy neural network inverse decoupling controller for bearingless permanent magnet synchronous motor | |
| CN103532459A (en) | Linear servo motor control method for numerically-controlled machine tool driving | |
| CN112039394A (en) | PMSM servo control system based on fuzzy active disturbance rejection | |
| CN109067275A (en) | A kind of permanent-magnetism linear motor chaotic control method based on decoupling self-adaptive sliding formwork | |
| Jie et al. | Speed regulation based on adaptive control and RBFNN for PMSM considering parametric uncertainty and load fluctuation | |
| CN112910359A (en) | Improved permanent magnet synchronous linear motor model prediction current control method | |
| Zhao et al. | Neuroadaptive dynamic surface control for induction motors stochastic system based on reduced-order observer | |
| Beladjine et al. | A comparative study of BLDC motor speed control using PI and ANN regulator | |
| CN115890668A (en) | Distributed optimization learning control method and system for robot joint module | |
| CN113659895B (en) | Permanent magnet synchronous motor full-state constraint finite time control method based on instruction filtering | |
| KR20200033478A (en) | Automated torque ripple reduction apparatus of motor | |
| CN114448311A (en) | Fuzzy neural network prediction decoupling control system for bearingless permanent magnet synchronous generator | |
| Wen-long et al. | Ship electrical propulsion control system based on improved model-free adaptive control | |
| CN115833690A (en) | Six-phase permanent magnet synchronous motor parameter-free model prediction current control system and method | |
| Elyazid et al. | New fuzzy speed controller for dual star permanent magnet synchronous motor | |
| Benbouhenni | A novel switching tables of twelve sectors DTC for induction machine drive using artificial neural networks | |
| CN114665766A (en) | Permanent magnet synchronous motor force and position hybrid control system based on load moment estimation | |
| CN114448310B (en) | Five-degree-of-freedom bearingless permanent magnet synchronous generator neural network prediction decoupling controller | |
| Mao et al. | Nonlinear decoupling sliding mode control of permanent magnet linear synchronous motor based on α-th order inverse system method | |
| Jegan et al. | Simulation and Validation of Permanent Magnet Synchronous Motor Drives Using Reinforcement Learning | |
| Moutchou et al. | Comparative study of SMC and PI control of a permanent magnet synchronous generator decoupled by singular perturbations | |
| Mahmoudi et al. | Neuro-Genetic sensorless sliding mode control of a permanent magnet synchronous motor using Luenberger observer | |
| Kalamian et al. | Design of a nonlinear model-based predictive controller for a wind turbine based on PMSG using an augmented extended kalman filter | |
| Liu et al. | CMAC neural network application for induction motor drives |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20240229 Address after: 518000 1104, Building A, Zhiyun Industrial Park, No. 13, Huaxing Road, Henglang Community, Longhua District, Shenzhen, Guangdong Province Applicant after: Shenzhen Hongyue Enterprise Management Consulting Co.,Ltd. Country or region after: China Address before: Zhenjiang City, Jiangsu Province, 212013 Jingkou District Road No. 301 Applicant before: JIANGSU University Country or region before: China |
|
| TA01 | Transfer of patent application right |
Effective date of registration: 20240606 Address after: 211135 room 815, building 1, artificial intelligence Industrial Park, No. 266, Chuangyan Road, Qilin science and Technology Innovation Park, Jiangning District, Nanjing, Jiangsu Province Applicant after: Nanjing Jijing micro Semiconductor Co.,Ltd. Country or region after: China Address before: 518000 1104, Building A, Zhiyun Industrial Park, No. 13, Huaxing Road, Henglang Community, Longhua District, Shenzhen, Guangdong Province Applicant before: Shenzhen Hongyue Enterprise Management Consulting Co.,Ltd. Country or region before: China |