CN101917150B - Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof - Google Patents

Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof Download PDF

Info

Publication number
CN101917150B
CN101917150B CN2010102094452A CN201010209445A CN101917150B CN 101917150 B CN101917150 B CN 101917150B CN 2010102094452 A CN2010102094452 A CN 2010102094452A CN 201010209445 A CN201010209445 A CN 201010209445A CN 101917150 B CN101917150 B CN 101917150B
Authority
CN
China
Prior art keywords
fuzzy
neural network
linear
generalized inverse
current
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.)
Active
Application number
CN2010102094452A
Other languages
Chinese (zh)
Other versions
CN101917150A (en
Inventor
刘国海
董蓓蓓
滕成龙
蒋彦
陈玲玲
赵文祥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
DAHANG YOUNENG ELECTRICAL CO.,LTD.
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to CN2010102094452A priority Critical patent/CN101917150B/en
Publication of CN101917150A publication Critical patent/CN101917150A/en
Application granted granted Critical
Publication of CN101917150B publication Critical patent/CN101917150B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Control Of Electric Motors In General (AREA)
  • Control Of Ac Motors In General (AREA)

Abstract

The invention discloses a robust controller of a permanent magnet synchronous motor based on a fuzzy-neural network generalized inverse and a construction method thereof. The construction method of the invention comprises the following steps of: combining an internal model controller and a fuzzy-neural network generalized inverse to form a compound controlled object; serially connecting two linear transfer functions and one integrator with the fuzzy-neural network with determined parameters and weight coefficients to form the fuzzy-neural network generalized inverse, serially connecting the fuzzy-neural network generalized inverse and the compound controlled object to form a generalized pseudo-linear system, linearizing a PMSM (permanent magnet synchronous motor), and decoupling and equalizing the linearized PMSM into a second-order speed pseudo-linear subsystem and a first-order current pseudo-linear subsystem; and respectively introducing an internal-model control method in the two pseudo-linear subsystems to construct the internal model controller. The robust controller of the invention has the advantages of overcoming the dependence and local convergence of the optimal gradient method on initial values and solving the problems of randomness and probability caused by using the simple genetic algorithm, obtaining the high performance control, anti-disturbance performance and adaptability of the motor and simplifying the control difficulty, along with simple structure and high system robustness.

Description

Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and building method
Technical field
The present invention relates to the permagnetic synchronous motor controller, be applicable to that a voltage source inverter drives the robust control of a permanent magnet synchronous motors, belongs to the technical field of Electric Drive control appliance.
Background technology
Permagnetic synchronous motor (Permanent Magnet Synchronous Motor is called for short PMSM) is widely used in the fields such as blower fan water pump of Aero-Space, weapons national defence, Digit Control Machine Tool, industrial robot, Flexible Control, communications industry, oil field and chemical industry and year long operational time.
The control method of Permanent-magnet Synchronous-motor Speed Servo System mainly contains constant voltage and frequency ratio control, vector control, direct torque control and Differential Geometry STATE FEEDBACK CONTROL etc.Wherein, simple in structure based on the Permanent-magnet Synchronous-motor Speed Servo System under the constant voltage and frequency ratio control mode of steady-state model, cost is low, be easy to realize, can satisfy general speed governing requirement; But systematic function is not high, too relies on the system dynamics Mathematical Modeling, is a kind of open loop control; And carrying load ability is limited during low speed; When shock load or speed command, the step-out phenomenon takes place easily, can't obtain desirable dynamic control performance.And based on the Permanent-magnet Synchronous-motor Speed Servo System under the vector control mode of dynamic model have that dynamic property is good, speed-regulating range width, control precision advantages of higher, be a kind of steady state approximation decoupling zero, it is extensive gradually therefore to drag the application in field in industry; But; Vector control is because very big to the dependence of motor parameter, and only when magnetic linkage reached stable state and keep constant, rotating speed and magnetic linkage just satisfied the decoupling zero relation; Be difficult to guarantee full decoupled; The actual inaccessible result of theoretic analysis of control effect, and used vector rotating coordinate transformation is comparatively complicated in analog DC Electric Machine Control process, and system robustness reduces greatly.Be convenient to realize total digitalization based on the Permanent-magnet Synchronous-motor Speed Servo System under the stator magnetic linkage oriented direct torque control mode; Need not alternating current machine and direct current machine are done equivalence; Complicated rotating coordinate transformation and motor model have been saved; Needn't consider to control in the vector control problem that effect receives the rotor parameter variable effect; Only need to detect the stator magnetic linkage of stator resistance and observation motor, be to utilize torque and the stagnant chain rate of magnetic linkage to come the implementation part dynamic decoupling, but have defectives such as low-speed performance is poor, torque pulsation is big.The Differential Geometry method is to be a kind of method with non linear system linearisation decoupling zero control that instrument grows up with the Differential Geometry; Purpose is through after non linear system being carried out the exact linearization method processing; Complication system is transformed into simple linear system; Can under the situation of not losing system controllability and measurability and accuracy, in the working field of broad, use linear theory to analyze and design linear controller, but the method be in the exact linearization method and the asymptotic dynamic decoupling of input and output of the system of realization like this; Requirement obtains accurately Mathematical Modeling and utilization complicacy and abstract mathematical tool, and using on the engineering has certain difficulty.
At present; Though the neural net inverse system control method can be realized the linearisation decoupling zero of permagnetic synchronous motor; But several integral form linear subsystems that form after the decoupling zero are open-loop unstables; And have local minimum point, cross the defectives such as too dependence experience of selection of study and structure and type based on the minimized neural net of empiric risk, permagnetic synchronous motor exists that load changing, system's controllable parameter are many, not modeling dynamic effects and easy step-out etc. in actual motion simultaneously; These uncertain factors cause model mismatch, make system depart from the expection controlled target.
Summary of the invention
Because the dynamic model of permagnetic synchronous motor control is the MULTIVARIABLE TIME-VARYING SYSTEM of non-linear a, close coupling; In order to overcome the deficiency of several kinds of basic control methods of above prior art; The present invention provides a kind of robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse; Realize permagnetic synchronous motor linearisation decoupling zero control, suppress parameter perturbation and load disturbance well, overcome not modeling and disturb dynamically; Improve the governing system dynamic responding speed and the steady-state tracking precision of permagnetic synchronous motor, realize the high-performance robust control.
Another object of the present invention provides the building method of above-mentioned based on fuzzy-neural network generalized inverse robust controller, and some linear subsystems of decoupling zero are carried out integrated treatment, guarantees the Control of PMSM effect.
The technical scheme that controller of the present invention adopts is: combined by internal mode controller and based on fuzzy-neural network generalized inverse and form; Said internal mode controller has the speed internal mode controller and the electric current internal mode controller composes in parallel, and the speed internal mode controller has the speed internal model and speed control is formed, and the electric current internal mode controller has the electric current internal model and current controller connects to form; Said based on fuzzy-neural network generalized inverse and composite controlled object are composed in series the broad sense pseudo-linear system, and the equivalence of broad sense pseudo-linear system is 1 sub-linear system of speed and 1 sub-linear system of electric current; Based on fuzzy-neural network generalized inverse adds two 2 linear transfer functions by five layers of fuzzy neural network with 5 input nodes, 2 output nodes and 1 integrator is formed; Said composite controlled object comprises that rate of current detection and computing module connect to form with the expansion inverter control section that drives PMSM; Expansion inverter control section is connected to form by the voltage source inverter under contrary Park conversion and the SVPWM debud mode; Rate of current detects with computing module and comprises that rate of current calculating section, Park conversion, Clarke conversion and photoelectric encoder connect to form, and Clarke conversion and photoelectric encoder connect PMSM.
The building method of controller of the present invention in turn includes the following steps: equivalence becomes PMSM with third-order model by Clarke conversion, Park conversion earlier, and PMSM forms composite controlled object through rate of current detection and integral body of computing module and expansion inverter control section formation; Connect with the fuzzy neural network of having confirmed each parameter and weight coefficient by 2 linear transfer functions and 1 integrator again and constitute based on fuzzy-neural network generalized inverse; Adopt based on fuzzy-neural network generalized inverse to connect with composite controlled object and constitute the broad sense pseudo-linear system, the broad sense pseudo-linear system becomes 1 second order speed linear subsystem and 1 single order electric current linear subsystem with PMSM linearisation and decoupling zero equivalence; At last second order speed linear subsystem and single order electric current linear subsystem are introduced internal model control method structure internal mode controller respectively; Internal mode controller combined with the broad sense pseudo-linear system form the based on fuzzy-neural network generalized inverse robust controller, the control composite controlled object.
Beneficial effect of the present invention is:
1, fuzzy neural network possesses the stronger advantages such as fuzzy reasoning ability of the stronger self-learning capability of neural net, computation capability, non-linear approximation capability and fuzzy logic simultaneously; It is combined with the linearisation decoupling zero characteristics of inverse system, utilize fuzzy logic technology to improve the learning ability of neural net; Utilize the learning ability of neural net to extract fuzzy rule or adjustment fuzzy rule parameter; Utilize neural net to realize fuzzy logic system and parallel fuzzy reasoning.This combination overcome control system not modeling influence dynamically; Have very strong robustness and fault-tolerance; The control problem of the multivariable of this complicacy of permagnetic synchronous motor two inputs two output System with Nonlinear Coupling is converted into the control problem of two stable linear subsystems; Further reasonably construct linear closed loop controller, can obtain high performance control and the anti-load disturbance and the adaptivity of motor, simplified the control difficulty greatly.
2, add the generalized inverse control system that transfer function and integrator are constructed composite controlled object with fuzzy neural network, be completely free of the dependence of traditional control method, overcome the unsteadiness of the fuzzy nerve network reverse system of integral form effectively for permagnetic synchronous motor controlled system Mathematical Modeling and parameter; Having solved former high-order controlled system partial status and be difficult for measuring the control problem of bringing, is the important breakthrough to traditional inverse system control method, the broad sense pseudo-linear system of generalized inverse system and the compound formation of original system; Not only can realize the linearisation decoupling zero of original system; And make limit reasonable disposition in complex plane of the linear subsystem that forms after the decoupling zero through the parameter of reasonable linear adjustment link, and obtain comparatively desirable open loop frequency characteristic, realize linearisation on a large scale; Decoupling zero and depression of order; Thereby can carry out High Accuracy Control according to linear control theory structure controller easily, help the comprehensive, simple in structure of system; System robustness is high, is easy to Project Realization.
3, fuzzy neural network parameter and weights confirms that reaching method of adjustment is that genetic algorithm combines with optimum gradient method.Traditional fuzzy neural network is used optimum gradient algorithm merely, though realize that simply local search ability is strong, its on-line study cycle is long, and algorithm the convergence speed is slow, is absorbed in defectives such as local minimum easily; Genetic algorithm is as a kind of global search and optimisation technique, though itself can not express knowledge, it has stronger learning ability and optimization ability; Genetic algorithm has parameter of overall importance and network configuration simultaneously, can search about 90% of optimal solution with fast speeds, but its later stage search variation probability is less; Be difficult to keep colony's diversity; And realize complicacy, the occasion in that needs are controlled in real time is not so good as optimum gradient method.Both are combined, learn from other's strong points to offset one's weaknesses, guarantee the global convergence of e-learning on the one hand by genetic algorithm, overcome dependence and the local convergence problem of optimum gradient method initial value; On the other hand, optimum gradient study has guaranteed local search ability, and online adjustment capability is strong, realizes simply having overcome randomness and probabilistic problem that simple genetic algorithm is brought simultaneously, helps to improve searching probability.
4, based on the dSPACE real-time emulation system as experiment porch, realized complete seamless link with MATLAB/Simulink/RTW.Advantages such as the dSPACE real-time system has real-time, and reliability is high, and extendibility is good.Processor in the dSPACE hardware system has high-speed computing, and has been equipped with abundant I/O support, and the user can make up as required; Software environment powerful and easy to use comprises and realizes that code generates automatically/downloads and test/kit debugged.Utilize its powerful software and hardware experiments platform can realize the high accuracy rotating speed control of permagnetic synchronous motor; The control algolithm of accomplishing motor is from the conceptual design to the mathematical analysis and test; From a cover concurrent engineering of the monitoring that is implemented to experimental result and the adjusting of real-time simulation test, the R&D cycle lack, economizes on resources, powerful, be easy to realization.
5. the present invention can be applied in the motor of other types such as synchronous machine, direct current machine, asynchronous machine equally, and is in the The synchronized Coordinative Control system of power set at a plurality of alternating current machines (many motors) with networking, has a extensive future.
Below in conjunction with accompanying drawing and embodiment the present invention is done further explain.
Description of drawings
Fig. 1 is that permagnetic synchronous motor body PMSM 1 and rate of current detect and computing module 31 connection layouts.
Fig. 2 is that PMSM 1 detects composite controlled object 3 structures that constituted with computing module 31 and simplifies equivalent model figure with expansion inverter control section 32 and rate of current.
Fig. 3 is based on fuzzy-neural network generalized inverse system 4 structures and equivalent model figure thereof;
Fig. 4 is two linear system diagrams of son that broad sense pseudo-linear system 5 structures and equivalence thereof become;
Fig. 5 is based on fuzzy-neural network generalized inverse robust controller 7 structure charts of the present invention.
Fig. 6 is that based on fuzzy-neural network generalized inverse robust controller 7 of the present invention uses the dSPACE experiment porch to carry out the theory diagram that control system is implemented.
Embodiment
As shown in Figure 5, based on fuzzy-neural network generalized inverse robust controller 7 control composite controlled object 3 of the present invention.Based on fuzzy-neural network generalized inverse robust controller 7 is through internal mode controller 6 and based on fuzzy-neural network generalized inverse 4 composition that combines.Internal mode controller 6 is made up of speed internal mode controller 61 and 62 parallel connections of electric current internal mode controller, and wherein, speed internal mode controller 61 is connected to form by speed internal model 611 and speed control 612; Electric current internal mode controller 62 is connected to form by electric current internal model 621 and current controller 622.Simultaneously; Based on fuzzy-neural network generalized inverse 4 in the based on fuzzy-neural network generalized inverse robust controller 7 is connected with composite controlled object 3 and is constituted broad sense pseudo-linear system 5, with System with Nonlinear Coupling decoupling zero equivalence 1 second order speed linear subsystem 51 of one-tenth and 1 single order electric current linear subsystem 52 of former high-order.Further; Based on fuzzy-neural network generalized inverse 4 is the equivalent mathematical model according to permagnetic synchronous motor PMSM 1; To the coupling between motor speed, voltage and the stator current; Analyzing on the reversible basis of PMSM 1 broad sense, adopting to have 5 input nodes, five layers of fuzzy neural network 41 of 2 output nodes add two 2 linear transfer functions and 1 integrator constitutes.Composite controlled object 3 is detected with computing module 31 and expansion inverter control section 32 by PMSM1, rate of current and connects to form; Be to drive PMSM1 against the Park conversion expansion inverter control section 32 that constitutes that combines with voltage source inverter under the SVPWM debud mode, connect the rate of current that constitutes by rate of current calculating section, Park conversion, Clarke conversion and photoelectric encoder 2 simultaneously and detect an integral body of forming with computing module 31.PMSM1 is formed by Clarke conversion, Park conversion and third-order model equivalence, and third-order model is three rank differential equation group under the d-q coordinate system.Wherein, the rate of current of formation detects the important component part of being not only composite controlled object 3 with computing module 31, also is electric current, rotating speed and the rotor displacement signal feedback link as internal model control and Park conversion simultaneously.Need to prove: the current signal of actual controlled device output is square
Figure BSA00000179183300051
of stator current, below all is called for short stator current.
The building method of above-mentioned based on fuzzy-neural network generalized inverse robust controller 7 is: at first based on permagnetic synchronous motor body PMSM1, the rate of current of forming via rate of current calculating section, Clarke, Park conversion and photoelectric encoder 2 detects with computing module 31 and constitutes an integral body by the expansion inverter control section 32 that the voltage source inverter under contrary Park conversion and the SVPWM modulation system is formed and forms composite controlled object 3 and come drive load.Secondly; Employing adds the based on fuzzy-neural network generalized inverse 4 with 2 input nodes, 2 output nodes that 2 linear transfer functions and 1 integrator the constitute formation broad sense pseudo-linear system 5 of connecting with composite controlled object 3 by the fuzzy neural network 41 (5 layer network) of 5 input nodes, 2 output nodes; Thereby the high-order nonlinear system linearization and the decoupling zero equivalence of PMSM 1 such multivariable, close coupling are become 1 second order speed linear subsystem 51 and 1 single order electric current linear subsystem 52, through the parameter a of reasonable linear adjustment transfer function J0, a J1...,
Figure BSA00000179183300052
, make limit reasonable disposition in complex plane of each linear subsystem that forms after the decoupling zero, realize of the transformation of the unstable subsystem of integral form to the stabistor system.On this basis; Second order speed linear subsystem 51 and single order electric current linear subsystem 52 are introduced internal model control method structure internal mode controller 6 respectively, and appropriate design internal mode controller 6 combines to form based on fuzzy-neural network generalized inverse robust controller 7 with control composite controlled object 3 with broad sense pseudo-linear system 5; Realization is to the high accuracy robust control of PMSM1; The system that makes overcomes not modeling and disturbs dynamically, has good sound attitude control performance, antijamming capability and high precision tracking performance.Can require to adopt different hardware or software to realize according to Different control.Specifically describe with following 7 steps:
The 1st step: as shown in Figure 1, the structure rate of current detects and computing module 31.Two phase stator current i in the control signal of PMSM1 SA, i SBDetect acquisition, two phase stator current i by Hall element SA, i SBAfter the Clarke conversion, the i that obtains through the Park conversion again Sd, i Sq, photoelectric encoder 2 detects tach signal and the i that PMSM1 obtains Sd, i SqAfter carrying out computing through the rate of current calculating section, the current signal i of output s 2=i Sd 2+ i Sq 2, rotor velocity ω rWith the output of angular displacement as rate of current detection and computing module 31.Rate of current detects with computing module 31 and as the output of composite controlled object 3 described below with for internal mode controller 6 feedback signal is provided simultaneously.
The 2nd step: as shown in Figure 2, the composite controlled object 3 of structure PMSM1.Composite controlled object 3 is to be connected and composed by the Mathematical Modeling of expansion inverter control section 32, equivalent PMSM1 and above-mentioned current detecting and computing module 31.Integral body to the body of expansion inverter control section 32 and PMSM1 is formed is carried out equivalence, and making it similarly, equivalence becomes controlled direct current machine; Expansion inverter control section 32 is by forming the Mathematical Modeling of the PMSM1 that connects thereafter against Park conversion and the voltage source inverter under the SVPWM modulation system; The Mathematical Modeling of PMSM1 is made up of Clarke conversion, Park conversion and DC Model series connection, but the actual still PMSM1 body that is connected with controller.Composite controlled object 3 be input as the stator voltage under the d-q two cordic phase rotators systems, i.e. u=[u 1, u 2] T=[u Sd, u Sq] T, be output as rotor velocity and two stator current signal, i.e. y=[y mutually 1, y 2] T=[ω r, i s 2] TU wherein Sd, u SqD axle and q shaft voltage under the two cordic phase rotators system as the input signal of composite controlled object 3, also is the output signal of system's reversibility Analysis here simultaneously respectively; ω r, i sBeing respectively the rotor velocity and the stator current signal of PMSM1 output, also is the important component part of system's reversibility Analysis input signal simultaneously.
The 3rd step: like Fig. 2~3, through analysis, equivalence and derivation, obtaining the Mathematical Modeling of whole permagnetic synchronous motor composite controlled object 3 under vector control mode is two cordic phase rotators systems; It is the Third-Order Nonlinear Differential Equations group under the d-q coordinate system; And exist according to the generalized inverse system of this three rank differential equation group of inverse system theoretical proof, and vector relatively rank be 2,1}; And then derive the generalized inverse of this system; Set up permagnetic synchronous motor generalized inverse system model,, confirm that simultaneously its 2 input variables are respectively for based on fuzzy-neural network generalized inverse 4 provides the foundation on the method 2 output variables are respectively the input u=[u of composite controlled object 3 1, u 2] T=[u Sd, u Sq] TWherein,
Figure BSA00000179183300062
With
Figure BSA00000179183300063
Be respectively two input variables of based on fuzzy-neural network generalized inverse system 4,
Figure BSA00000179183300064
With
Figure BSA00000179183300065
Be the 2nd step rotor angular speed and two stator current signal y mutually 1And y 2And the linear synthetic quantity of their all-order derivative, a 10, a 11, a 12, a 20And a 21Be respectively coefficient.
The 4th step: like Fig. 3, adopt fuzzy neural network 41 to add 2 linear transfer functions and 1 integrator structure based on fuzzy-neural network generalized inverse 4, for the learning training of fuzzy neural network 41 provides the foundation on the method.According to the concrete condition of PMSM1, reasonably regulate the parameter a of based on fuzzy-neural network generalized inverse 4 linear transfer functions 10, a 11, a 12, a 20, a 21Characterize the dynamic characteristic of generalized inverse system; Make limit reasonable disposition in complex plane of the single output of the single input linear subsystem that forms after the decoupling zero, realize of the transformation of the unstable subsystem of integral form, realize the stable control of open loop linearisation of non linear system to the stabistor system.Wherein fuzzy neural network 41 adopts five layer networks.Ground floor is an input layer, and input number of nodes is 5, and neuron is represented the input language variable for the input node, and this layer only is used to transmit signal and arrives one deck, i.e. f down 1=u i (1), a 1=f 1(f zAnd a zClean input and the activation primitive of representing the z node layer respectively, z=1,2,3,4,5; u i (1)In (1) and i represent neuronic i the input of ground floor node, below analogize), weight w Ij (1)=1 (representing that i input language variable arrives one deck j neuronic link weight coefficients down); Second layer obfuscation layer, node number are 15, and each node is represented a linguistic variable value, are used to calculate the membership function of each input component, and it is excitation function that this layer neuron chosen Gaussian function, promptly
Figure BSA00000179183300071
, f 2=-(u i (2)-m Ij) 2/ σ Ij 2(m IjAnd σ IjBe respectively the center and the width of j Gaussian function of i input language variable), each neuron is exported corresponding membership function, weight w Ij (2)=m IjThe 3rd layer is rules layer, and the node number is 9, is used to produce fuzzy logic ordination and former piece coupling, promptly calculates the fitness of every rule, and this layer neuron node carried out fuzzy and operated AND operation, i.e. f on the relevant position 3=min{u 1 (3), u 2 (3)... U 5 (3), a 3=f 3, weight w Ij (3)=1; The 4th layer of normalization layer, the node number is 9, network connects the conclusion that has defined regular node, produces every rule corresponding to the output that input is produced, and is the consequent coupling, carries out inclusive-OR operation, promptly
Figure BSA00000179183300072
a 4=min{1, f 4(p representes the input number of neuron node), weight w Ij (4)=1; Layer 5 ambiguity solution layer (output layer), the node number is 2, is used for ambiguity solution, realizes that sharpening calculates, and produces total output of control law, promptly
Figure BSA00000179183300073
Weight w Ij (5)=m Ijσ IjWherein import in the node first input of based on fuzzy-neural network generalized inverse 4 for 5
Figure BSA00000179183300074
First input as fuzzy neural network 41; It is through second-order system s/a 10s 2+ a 11S+a 12(second-order system is the second-order linearity link G that is connected with fuzzy neural network 41 1(s) s, a 10, a 11, a 12Coefficient for linear transfer function) is output as
Figure BSA00000179183300075
Be second input of fuzzy neural network 41; Again through 1 integrator s -1Output y 1, be the 3rd input of fuzzy neural network 41; Second input of based on fuzzy-neural network generalized inverse 4
Figure BSA00000179183300076
The 4th input as fuzzy neural network 41; It is through first-order system 1/a 20S+a 21(first-order system is the first-order linear link G that is connected with fuzzy neural network 41 2(s), a 20, a 21Coefficient for the single order link) is output as y 2, be the 5th input of fuzzy neural network 41.So fuzzy neural network 41 is formed based on fuzzy-neural network generalized inverse 4 with 2 linear transfer functions and 1 integrator, the output of fuzzy neural network 41 is exactly the output of based on fuzzy-neural network generalized inverse 4.
The 5th step: like Fig. 3, the adjustment of the parameter of fuzzy neural network 41 and weight coefficient value and definite.In conjunction with genetic algorithm and optimum gradient method, the study of fuzzy neural network 41 is divided into off-line learning and two stages of online adjustment weight coefficient.Specifically be divided into following steps: 1. with step excitation signal { u Sd, u SqBe added to 2 inputs of composite controlled object 3 respectively, gather the rotor velocity ω of PMSM1 with the sampling period of 5ms rAnd current i SA, i SB, detect and computing module 31 acquisition desired datas through rate of current And preserve; 2. with the data-signal of preserving
Figure BSA00000179183300078
Off-line is tried to achieve speed single order, second dervative respectively With the electric current first derivative
Figure BSA000001791833000710
Have this moment: y 1r,
Figure BSA000001791833000711
And then try to achieve according to the method in the 3rd step And signal done standardization processing, form the training sample set of fuzzy neural network 41
Figure BSA00000179183300081
3. at first use genetic algorithm off-line training fuzzy neural network 41, the parameter of its membership function of coarse adjustment and the initial weight of output, wherein crossover probability P cWith the variation probability P mAdopt adaptive mode, come the convergence situation (P of measure algorithm with suitable function c=k 1/ (f Max-f), P m=k 2/ (f Max-f), f MaxRepresent the maximum in the colony, average fitness, k respectively with f 1, k 2Be the positive real coefficient of size between 0~1), stop evolutionary generation and be set at G=300, so obtain an overall approximate solution, concrete training step and general genetic algorithm are similar, rough cover half is really stuck with paste each parameter and the weight coefficient of neural net 41; When the controling appliance running body, adopt the parameter of the error anti-pass optimum gradient method online in real time refinement adjustment fuzzy neural network 41 that drives quantifier and learning rate changing then, fuzzy neural network 41 output mean square error precision are remained in 0.0005.
The 6th step: formation fuzzy neural broad sense pseudo-linear system 5 as shown in Figure 4 becomes 1 sub-linear system 51 of speed and 1 sub-linear system 52 of electric current with former composite controlled object 3 linearisations and decoupling zero equivalence.At first, being connected with the fuzzy neural network 41 of having confirmed each parameter and weight coefficient by 2 linear transfer functions and 1 integrator constitutes based on fuzzy-neural network generalized inverse 4, shown in the little frame of broken lines of the left figure of Fig. 4; Then; This based on fuzzy-neural network generalized inverse 4 is composed in series broad sense pseudo-linear system 5 with composite controlled object 3; Shown in the big frame of broken lines of the left figure of Fig. 4; This broad sense pseudo-linear system 5 is that the electric current linear subsystem 52 parallelly connected equivalences by the speed linear subsystem 51 of 1 second order and 1 single order form, and shown in the right figure of Fig. 4,1 the second order speed linear subsystem 51 that equivalence becomes and the input of 1 single order electric current linear subsystem 52 are respectively Be two input variables of based on fuzzy-neural network generalized inverse 4, corresponding output is respectively ω r,
Figure BSA00000179183300083
Be that rate of current detects electric current and the rotor velocity with computing module 31 outputs, realized the control of the complicated nonlinear system of former high-order, coupling is converted into simple linear system control.
The 7th step: structure based on fuzzy-neural network generalized inverse robust controller 7.Can know according to the 2nd, 3 steps; The relative rank of system are { 2; 1}; Can know the interference that is faced in the character that is input as
Figure BSA00000179183300084
and combines respectively speed linear subsystem 51 and 52 two linear subsystems of electric current linear subsystem of the broad sense pseudo-linear system 5 that based on fuzzy-neural network generalized inverse 4 and composite controlled object 3 are combined into, the actual motion and the time-varying characteristics structure based on fuzzy-neural network generalized inverse robust controller 7 of parameter according to the 6th step.The present invention adopts designing method based on fuzzy-neural network generalized inverse robust controllers 7 such as internal model control principle in the linear system robust control theory, Lyapunov (Liapunov) theory.Wherein, internal mode controller 6 by linearisation speed internal mode controller 61 constitute with electric current internal mode controller 62.D 1(s) and D 2(s) be respectively the interference signal of two controllers, speed internal mode controller 61 is made up of speed internal model 611 and speed control 612, and electric current internal mode controller 62 is made up of electric current internal model 621 and current controller 622.Suitably select parameter a 10, a 11, a 12, a 20, a 21, make the inside expectational model 611 of second-order linearity speed subsystem be G 1m(s)=1/ (a 10s 2+ a 11S+a 12)=1/ (s 2+ 1.414s+1), so design obtains corresponding speed control 612 do
Figure BSA00000179183300091
The inside expectational model 621 of single order electric current linear subsystem is G 2m(s)=1/ (a 20S+a 21)=1/ (s+1) can be designed equally and obtains corresponding current controller 622 and do
Figure BSA00000179183300092
Wherein, a 10, a 12, a 11Transfer function G for speed inside expectational model 611 1m(s) coefficient, value are a 10=a 12=1, a 11=1.414, this moment internal model G 1m(s) be typical second order stable linear system; F 1(s) be a type low pass filter of corresponding speed controller 612, F 1(s)=1/ (0.5s+1) 2a 20, a 21Be the transfer function coefficients of electric current inside expectational model 621, value is a 20=a 21=1; F 2(s) be a type low pass filter of respective electrical stream controller 622, F 2(s)=1/ (2s+1)).The structure of whole based on fuzzy-neural network generalized inverse robust controller 7 and connection situation are as shown in Figure 5.
The whole enforcement sketch map of Permanent-magnet Synchronous-motor Speed Servo System on dSPACE real-time simulation and test macro experiment porch based on based on fuzzy-neural network generalized inverse robust controller 7 is as shown in Figure 6.PMSM1 and dSPACE 81 are arranged among Fig. 7, and subsidiary module comprises analog input ADC module, simulation output DAC module, input part, photoelectric coded disk 2, Hall element, magnetic brake unit, industry control display module 83 and SPM IPM 82; Software environment comprises that mainly real-time code generates downloaded software RTI, Comprehensive Experiment and test environment software ControlDesk and Simulink simulation software.Based on fuzzy-neural network generalized inverse robust controller 7 adopts dSPACE to realize controlling composite controlled object 3.The experiment control program downloads to the dSPACE control board by host computer, sends the experiment enabling signal, the control system independent operating through the visual control of ControlDesk interface; 6 road pwm control signals to the SPM drive motors of control board output; Test section collection electric current, voltage, speed and guard signal feed back to control board and store in order to the control effect analysis, but off-line or online modification parameter are controlled motor to reach high accuracy stable operation, shorten system development cycle.
The present invention is through the structure based on fuzzy-neural network generalized inverse; Realize this multivariable of permagnetic synchronous motor, close coupling the time become linearization of nonlinear system decoupling zero control; The control problem of the complication system that stator current, voltage and speed are intercoupled is converted into the control problem of linear stabistor system of simple second-order speed and the linear stabistor of single order electric current system, and combines internal model control principle, designs robust controller convenient and reasonablely; Realization is to the high accuracy robust control of permagnetic synchronous motor rotating speed; Overcome system not modeling disturb dynamically, make system have good dynamic and static performance, anti-interference and high precision tracking performance.

Claims (3)

1. robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse, it is characterized in that: this based on fuzzy-neural network generalized inverse robust controller (7) is combined by internal mode controller (6) and based on fuzzy-neural network generalized inverse (4) and forms; Said internal mode controller (6) is composed in parallel by speed internal mode controller (61) and electric current internal mode controller (62); Speed internal mode controller (61) is made up of speed internal model (611) and speed control (612), and electric current internal mode controller (62) is connected to form by electric current internal model (621) and current controller (622); Said based on fuzzy-neural network generalized inverse (4) is composed in series broad sense pseudo-linear system (5) with composite controlled object (3), and broad sense pseudo-linear system (5) equivalence is 1 sub-linear system of speed (51) and 1 sub-linear system of electric current (52); Based on fuzzy-neural network generalized inverse (4) adds 2 linear transfer functions by the five layers of fuzzy neural network (41) with 5 input nodes, 2 output nodes and 1 integrator is formed; Said composite controlled object (3) is detected by permagnetic synchronous motor (1), rate of current and the expansion inverter control section (32) of computing module (31) and driving permagnetic synchronous motor (1) connects to form; Expansion inverter control section (32) is connected to form by the voltage source inverter under contrary Park conversion and the SVPWM debud mode; Rate of current detects with computing module (31) and is connected to form by rate of current calculating section, Park conversion, Clarke conversion and photoelectric encoder (2), and Clarke conversion and photoelectric encoder (2) connect permagnetic synchronous motor (1).
2. the building method of a robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse is characterized in that successively as follows:
1) equivalence becomes permagnetic synchronous motor (1) with third-order model by Clarke conversion, Park conversion, and permagnetic synchronous motor (1) detects with computing module (31) and integral body of expansion inverter control section (32) formation through rate of current and forms composite controlled object (3);
2) connect with the fuzzy neural network (41) of having confirmed each parameter and weight coefficient by 2 linear transfer functions and 1 integrator and constitute based on fuzzy-neural network generalized inverse (4); Adopt based on fuzzy-neural network generalized inverse (4) to connect with composite controlled object (3) and constitute broad sense pseudo-linear system (5), broad sense pseudo-linear system (5) becomes 1 second order speed linear subsystem (51) and 1 single order electric current linear subsystem (52) with permagnetic synchronous motor (1) linearisation and decoupling zero equivalence;
3) second order speed linear subsystem (51) and single order electric current linear subsystem (52) are introduced internal model control method structure internal mode controller (6) respectively; Internal mode controller (6) combined with broad sense pseudo-linear system (5) form based on fuzzy-neural network generalized inverse robust controller (7), control composite controlled object (3).
3. building method according to claim 2 is characterized in that:
In the step 1), with permagnetic synchronous motor (1) two phase stator current through detecting electric current, rotor velocity and angular displacement that the rotating speed that obtains export output through electric current that the Park conversion obtains and by photoelectric encoder (2) through the computing of rate of current calculating section after again after the Clarke conversion as rate of current detection and computing module (31); Composite controlled object (3) be input as the stator voltage under the d-q coordinate system, be output as rotor velocity and two stator currents mutually;
Step 2) in, 2 input variables of based on fuzzy-neural network generalized inverse (4) be respectively rotor velocity with two mutually stator current signal with and the linear synthetic quantity of all-order derivative, 2 output variables be respectively the input of composite controlled object (3); The parameter of fuzzy neural network (41) and definite method of weight coefficient value are following steps: 1. with step excitation signal { u Sd, u SqBe added to 2 inputs of composite controlled object (3) respectively, gather the rotor velocity ω of permagnetic synchronous motor (1) with the sampling period of 5ms rAnd current i SA, i SB, detect and computing module (31) acquisition desired data through rate of current
Figure FSB00000716500300021
And preserve; 2. with the data-signal of preserving
Figure FSB00000716500300022
Off-line is tried to achieve speed single order, second dervative respectively
Figure FSB00000716500300023
With the electric current first derivative The training sample set of forming fuzzy neural network (41); 3. use genetic algorithm off-line training fuzzy neural network (41) earlier; The parameter of its membership function of coarse adjustment and the initial weight of output obtain an overall approximate solution; Rough each parameter and the weight coefficient of confirming fuzzy neural network (41); And then when concrete operation, adopt the error anti-pass optimum gradient method refinement that drives quantifier and learning rate changing to adjust the parameter of fuzzy neural network (41), fuzzy neural network (41) output mean square error precision is remained in 0.0005;
In the step 3); The input of 1 second order speed linear subsystem (51) and 1 single order electric current linear subsystem (52) is respectively two input variables of based on fuzzy-neural network generalized inverse (4), and output is respectively rate of current and detects electric current and the rotor velocity of exporting with computing module (31); Based on fuzzy-neural network generalized inverse robust controller (7) adopts dSPACE to realize controlling composite controlled object (3).
CN2010102094452A 2010-06-24 2010-06-24 Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof Active CN101917150B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2010102094452A CN101917150B (en) 2010-06-24 2010-06-24 Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2010102094452A CN101917150B (en) 2010-06-24 2010-06-24 Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof

Publications (2)

Publication Number Publication Date
CN101917150A CN101917150A (en) 2010-12-15
CN101917150B true CN101917150B (en) 2012-06-20

Family

ID=43324568

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2010102094452A Active CN101917150B (en) 2010-06-24 2010-06-24 Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof

Country Status (1)

Country Link
CN (1) CN101917150B (en)

Families Citing this family (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102437816B (en) * 2011-10-25 2014-05-07 武汉鑫通科创科技发展有限公司 Adaptive motor motion control apparatus based on neural network
CN103023398B (en) * 2012-11-19 2015-01-21 西安理工大学 Internal model control method of permanent magnet synchronous motor
CN103259479B (en) * 2013-05-28 2016-08-10 江苏大学 A kind of permanent magnet synchronous electric machine neural network left inverse state observation method
CN103345159B (en) * 2013-07-03 2015-08-05 江苏大学 Based on the hybrid electric vehicle BSG system control method of neural Network Adaptive Inversion
US9390368B2 (en) 2013-10-21 2016-07-12 International Business Machines Corporation Coupling parallel event-driven computation with serial computation
CN104142626B (en) * 2014-04-22 2017-05-24 广东工业大学 Ship dynamic positioning control method based on inverse system and internal model control
CN104155877A (en) * 2014-08-19 2014-11-19 江苏科技大学 Brushless DC motor fuzzy control system based on genetic algorithm and control method thereof
CN104852639A (en) * 2015-05-26 2015-08-19 江南大学 Parameter self-tuning speed controller of permanent magnet synchronous motor based on neural network
KR102459677B1 (en) * 2015-11-05 2022-10-28 삼성전자주식회사 Method and apparatus for learning algorithm
CN106130431B (en) * 2016-07-21 2018-09-28 宿迁学院 A kind of linear motor RBF neural generalized inverse internal model control method
CN106249599B (en) * 2016-09-28 2020-01-10 河南理工大学 Neural network prediction-based networked control system fault detection method
CN106849793B (en) * 2017-03-01 2019-03-01 西安交通大学 A kind of Over Electric Motor with PMSM fuzzy Neural Network Control System
CN106934174B (en) * 2017-03-27 2020-06-05 上海海事大学 Modeling and stability analysis method for electric propulsion ship experiment platform
CN107490760A (en) * 2017-08-22 2017-12-19 西安工程大学 The circuit breaker failure diagnostic method of fuzzy neural network is improved based on genetic algorithm
CN109324503B (en) * 2018-08-28 2022-02-15 南京理工大学 Multilayer neural network motor system control method based on robust integration
CN109861618B (en) * 2019-01-11 2020-11-20 江苏大学 Construction method of anti-interference composite controller of BSG alternating current motor for hybrid electric vehicle
CN113574795A (en) * 2019-03-19 2021-10-29 松下知识产权经营株式会社 Motor control method, method for changing control model of motor, motor control system, system for changing control model of motor, and program for changing control model of motor
CN111948946B (en) * 2020-08-24 2022-05-17 淮阴工学院 Robust evaluation inversion control system based on HJI theory and design method thereof
CN112019111B (en) * 2020-08-27 2022-01-18 青岛大学 State constraint-based fuzzy self-adaptive control method for random system of permanent magnet synchronous motor
CN112936270B (en) * 2021-02-04 2022-05-03 合肥工业大学 Mechanical arm internal model control method based on neural network
CN113067506B (en) * 2021-03-09 2022-07-26 中国地质大学(武汉) Permanent magnet synchronous motor periodic disturbance suppression method based on inner model equivalent input interference
CN113093546B (en) * 2021-04-02 2022-11-11 中国科学院光电技术研究所 Design method of second-order fuzzy-dynamic high-profile controller for photoelectric tracking system

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1212702C (en) * 2003-02-20 2005-07-27 东南大学 Neural network reversal control frequency converter of induction motor and structure method
CN100536311C (en) * 2007-11-30 2009-09-02 江苏大学 Neural network generalized inverse permanent magnetism synchronous machine decoupling controller structure method without bearing
CN101299581B (en) * 2008-03-10 2010-09-15 江苏大学 Neural network generalized inverse coordination control frequency transformer for two induction machines and construction method thereof
CN101299580A (en) * 2008-03-10 2008-11-05 江苏大学 Cooperative controller for two-motor vector control frequency control system and construction method thereof
CN101630940B (en) * 2009-08-12 2012-07-04 江苏大学 Fuzzy neural network inverse robust controller of induction motor speed regulation system and construction method
CN101630936B (en) * 2009-08-12 2011-09-14 江苏大学 Neural network inverse controller of brushless DC motor and construction method thereof
CN101741297B (en) * 2009-12-30 2012-07-11 南京信息职业技术学院 Method and device for inversely controlling fuzzy compensation of radial position of bearing-free synchronous reluctance motor

Also Published As

Publication number Publication date
CN101917150A (en) 2010-12-15

Similar Documents

Publication Publication Date Title
CN101917150B (en) Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof
CN106026822B (en) The inertia on-line identification method and system of servo motor driving system
CN103259479B (en) A kind of permanent magnet synchronous electric machine neural network left inverse state observation method
CN106788052B (en) Asynchronous machine command filtering error compensation fuzzy control method based on observer
CN102611380B (en) Online identification method for parameters of double-fed motor
CN106849814A (en) Leapfroged Fuzzy Neural PID linear synchronous generator control method based on fruit bat
CN105375848B (en) A kind of permanent magnet synchronous motor Adaptive Identification control method and its control system
CN105186958B (en) The five mutually fault-tolerant magneto internal model control methods based on Neural Network Inverse System
CN104734595A (en) Identification method for rotary inertia of permanent magnet synchronous motor based on model reference self-adaption
CN104953913A (en) Networked AC (alternating current) motor LS-SVM (least squares support vector machine) generalized inverse decoupling control method based on active-disturbance rejection
Yan et al. Torque estimation and control of PMSM based on deep learning
CN110212575A (en) The small signal stability modal analysis method of double-fed blower one machine infinity bus system
CN102769424A (en) Support vector machine inverse-based bearingless asynchronous motor control system
Rodrigues et al. Predictive incremental vector control for DFIG with weighted-dynamic objective constraint-handling method-PSO weighting matrices design
CN116738636A (en) Multi-machine equivalent method of doubly-fed wind power station considering impedance characteristics and synchronization mechanism
Song et al. Identification of PMSM based on EKF and elman neural network
CN102790580A (en) Constructing method for decoupling controller of support vector machine inverter of bearingless asynchronous motor
CN104022701B (en) Mould method for control speed in a kind of permanent magnetic linear synchronous motor Newton method
Xie et al. Optimization research of FOC based on PSO of induction motors
CN101630940B (en) Fuzzy neural network inverse robust controller of induction motor speed regulation system and construction method
CN111293693A (en) Doubly-fed wind turbine converter control parameter identification method based on extended Kalman filtering
Barra et al. Sensorless nonlinear control of wind energy systems with doubly fed induction generator
Wang et al. Lion swarm optimisation‐based tuning method for generalised predictive fractional‐order PI to control the speed of brushless direct current motor
CN102790582B (en) Manufacturing method of radial fuzzy neural network generalized inverse controller of bearingless asynchronous motor
Zhang et al. Wave heave compensation based on an optimized backstepping control method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20211220

Address after: 212200 No. 1, pilot Road, Yangzhong high tech Industrial Development Zone, Zhenjiang City, Jiangsu Province

Patentee after: DAHANG YOUNENG ELECTRICAL CO.,LTD.

Address before: 212013 No. 301, Xuefu Road, Zhenjiang, Jiangsu

Patentee before: JIANGSU University

TR01 Transfer of patent right