CN107968613A - A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network - Google Patents

A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network Download PDF

Info

Publication number
CN107968613A
CN107968613A CN201711235087.0A CN201711235087A CN107968613A CN 107968613 A CN107968613 A CN 107968613A CN 201711235087 A CN201711235087 A CN 201711235087A CN 107968613 A CN107968613 A CN 107968613A
Authority
CN
China
Prior art keywords
bat
neural network
algorithm
fuzzy neural
bee colony
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
Application number
CN201711235087.0A
Other languages
Chinese (zh)
Inventor
乔维德
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.)
Wuxi open university
Original Assignee
Wuxi open 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 Wuxi open university filed Critical Wuxi open university
Priority to CN201711235087.0A priority Critical patent/CN107968613A/en
Publication of CN107968613A publication Critical patent/CN107968613A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
    • H02P23/00—Arrangements or methods for the control of AC motors characterised by a control method other than vector control
    • H02P23/0077—Characterised by the use of a particular software algorithm
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • 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)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Feedback Control In General (AREA)

Abstract

The present invention discloses a kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network, two kinds of algorithms of bat algorithm and artificial bee colony are blended to form bat artificial bee colony hybrid algorithm, for the structural parameters of Optimal Recursive fuzzy neural network controller, and by the controller introduce permanent magnet synchronous motor revolution speed control system.Emulation shows with experimental analysis, Recurrent Fuzzy Neural Network rotational speed governor using the present invention based on the optimization of bat artificial bee colony hybrid algorithm, it can realize the quick response of control system for permanent-magnet synchronous motor, and non-overshoot, control accuracy is high, robustness is good, strong antijamming capability, can realize accurate rotating speed control.

Description

A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network
Technical field
The present invention relates to permanent magnet synchronous motor control technology field, and in particular to one kind is based on bat-artificial bee colony mixing The permanent magnet synchronous motor rotational speed governor design method of algorithm optimization Recurrent Fuzzy Neural Network.
Background technology
Permanent magnet synchronous motor (abbreviation PMSM) replaces electrical excitation, no magnet exciting coil and brush, volume using high-energy permanent magnet Small with being lost, operational efficiency and reliability are high, and adaptation external environment ability is strong, its technical performance is better than permanent-magnet brushless DC electric machine And induction machine etc., obtain in the low-power applications occasion such as electric drive system and numerically-controlled machine tool, industrial robot and generally should With.But because permanent magnet synchronous motor has the characteristics that high-order, parameter time varying, multivariable, serious non-linear and strong coupling, it is difficult to Mathematical models describe its dynamic running process, and are often influenced by unfavorable factors such as load disturbances, cause interference free performance weak, Largely effect on PMSM system control performances.General PMSM governing systems use PI control methods, although PI control algolithms are simple, and And have certain control accuracy, but it still falls within Linear Control, is unable to reach the serious nonlinear system high accuracy of PMSM, fast-response is wanted Ask.At present, related scholar proposes the experimental program that PI controls are combined with artificial intelligence, design PI type Fuzzy, neuron network PI controller etc. Control method, meets the certain control requirement of system.Fuzzy control robustness is stronger, but control accuracy is not high；Neutral net is held Wrong ability and self-learning function are stronger, and but learning process is slow.For this reason, the advantages of how integrating fuzzy logic and neutral net and lack Fall into, fusion fuzzy logic, neutral net and PI control methods, for designing a kind of speed of control system for permanent-magnet synchronous motor Adjuster, realizes permanent magnet synchronous motor the control of accurate rotating speed, become the emphasis for the control of current permanent magnet synchronous motor rotating speed, One of difficult point research topic.
The content of the invention
It is an object of the invention to improve the speed responsive and control accuracy of control system for permanent-magnet synchronous motor, enhancing control The robustness and Ability of Resisting Disturbance of system, more accurately to realize the speed control of permanent magnet synchronous motor.Present invention design A kind of permanent magnet synchronous motor rotational speed governor based on bat-artificial bee colony hybrid algorithm Optimal Recursive fuzzy neural network, will Bat algorithm and artificial bee colony algorithm blend to form a kind of new bat-artificial bee colony hybrid algorithm, and Optimal Recursive obscures Speed Controller of Networks structural parameters, so as to more preferably meet the performance requirement of permanent magnet synchronous motor speed control.The present invention Technical solution be：
A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network, control system containing current inner loop and Speed outer shroud, the current regulator of current inner loop are still designed as a kind of recurrence fuzzy neural using routine PI controls, speed outer shroud Network PI controllers, i.e., form speed control by Recurrent Fuzzy Neural Network controller and pi regulator are compound.Wherein recurrence Fuzzy neural network is optimized by bat-artificial bee colony hybrid algorithm.Bat-artificial bee colony hybrid algorithm Optimal Recursive mould The step of pasting neutral net is as follows：
Step 1：Initiation parameter and bat position, that is, initialize bat population quantity D in bat algorithm, bat algorithm is most Big iterations NB, pulse (ultrasonic wave) frequency f that bat sends, maximum impulse intensity of sound S, maximum impulse frequency R0, arteries and veins Rush intensity of sound attenuation coefficient λ, pulse frequency increase coefficient δ, the position x of random initializtion bati(i=1,2 ..., D)；People Work ant colony algorithm maximum iteration limit；
Step 2：By following equation adjustment bat pulse frequency fi, renewal bat flying speed viAnd position xi, find current Optimal bat individual；
fi=fmin+(fmax-fmin)h
In formula,Respectively the location of t generations and t+1 i-th bat of generation,Respectively t generations and t+1 The flying speed of i-th bat of generation；fiFor the pulse frequency of i-th bat, fmax、fminIt is respectively the maximum of bat pulse frequency Value, minimum value；H is 0~1 section uniform random number；xbestFor current global optimum position；
Step 3：Produce random number r1If r1＞ Ri, then choose optimal solution from current population, and near optimal solution with Machine produces the new position x of a local solution, at this time batnewFor：
xnew=xold+τSt
In formula, τ be [- 1,1] scope random number, StRepresent loudness average value in the bat population same period；
Step 4：Produce random number r2If r2＜ SiAnd mean square error (fitness) J (xi) ＜ J (x0), increase according to rule RiAnd reduce Si, i.e. bat reduces the ultrasonic pulse loudness of transmitting, while increases sound pulse transmission number, i.e.,：
In formula,For sound wave pulse loudness of i-th bat in t+1 and t iteration,For t+1 iteration When i-th bat transmitting pulse frequency, λ is [0,1] value range, δ ＞ 0；
Step 5：Fitness J (x are pressed to bat individuali) assessed, find and record the optimal bat individual in current location, If meeting bat algorithm maximum iteration NB, M optimal location solution is exported, otherwise return to step 2；
Step 6：The M optimal location solution exported by step 5 reformulates artificial bee colony, and is used as artificial bee colony algorithm In initial bee source；
Step 7：The primary iteration times N C=1 of artificial bee colony algorithm is made, leads bee continuous when searching for nectar source in bee colony Update current location xij, then the nectar source fitness of search is evaluated, i.e.,：
Wherein, Fiti、fiI-th of nectar source fitness and adaptive value are represented respectively
Step 8：The new nectar source of bee searching will be led, if the latter is less than the former, to be adopted compared with the fitness of green molasses source With new nectar source position substitution green molasses source position, otherwise constant and NC+1；
Step 9：Calculate the probable value P of each nectar source position, bee is followed with reference to P in bee colonyiSelection leads what bee searched New nectar source, and calculate its fitness value；
Step 10：Compare the new nectar source for following bee to select and green molasses source fitness value, if the former is more than the latter, with new honey Source position substitution green molasses source；Otherwise constant and NC+1；
Step 11：When iterations NC exceedes artificial bee colony algorithm cycle-index maximum limit, preserve and export group Optimal nectar source individual, is used as Recurrent Fuzzy Neural Network optimized parameter a in bodyij、bij、rij、ωkOptimal initial value.
Step 12：The global optimum's individual exported after bat-artificial bee colony hybrid algorithm optimization is substituted into recurrence to obscure Parameters of Neural Network Structure, utilizes BP algorithm training network by input sample, finally makes mean square error (the target letter of network Number) MSE values minimum, so as to export the optimum structure parameter a of Recurrent Fuzzy Neural Networkij、bij、rij、ωk；MSE is defined as：
In formula, n is sample number, YK, pFor training sample P k-th of output node reality output；QK, pFor the corresponding phase Hope output, 1/ (MSE+1) is defined as the fitness function in bat-artificial bee colony algorithm here.
The beneficial effects of the invention are as follows：
Study is optimized to Recurrent Fuzzy Neural Network controller parameter using bat-artificial bee colony hybrid algorithm, can Asked with overcoming the convergence of low speed present in traditional BP algorithm Optimal Recursive fuzzy neural network parameter, being easily absorbed in local minimum etc. Topic.Bat algorithm search speed early period is fast, can search global optimum region rapidly, and artificial bee colony algorithm is with very strong Ability of searching optimum, but the speed of search early period feasible solution is slower, as long as searching feasible solution, the search capability of the algorithm It will improve quickly.The present invention is by relatively and combining both advantage and defect, effective integration bat algorithm and artificial bee colony Algorithm, forms bat-artificial bee colony hybrid algorithm, and it is excellent to be used for Recurrent Fuzzy Neural Network speed control structural parameters first Change.Emulation shows with result of the test, and using control strategy proposed by the present invention, the response of permanent magnet synchronous motor system is fast, non-overshoot, Control accuracy is high, robustness and its strong antijamming capability, can realize accurate rotating speed control.
Brief description of the drawings
Fig. 1 is control system for permanent-magnet synchronous motor structure principle chart
Fig. 2 is Recurrent Fuzzy Neural Network controller (RFNN) structure chart
Fig. 3 is bat-artificial bee colony hybrid algorithm Optimal Recursive fuzzy neural network flow chart
Fig. 4 is the PMSM rate curves under PI and RFNN controllers act on respectively
Fig. 5 is experimental principle schematic diagram
Fig. 6 is the PMSM rotating speed response empirical curves based on RFNN controllers
Fig. 7 is the PMSM torque response empirical curves based on RFNN controllers
Embodiment
With reference to the accompanying drawings and embodiments to technical solution of the present invention respectively from control system for permanent-magnet synchronous motor structure, pass Return fuzzy neural network controller design, bat-artificial bee colony hybrid algorithm Optimal Recursive fuzzy neural networks controller parameter, Emulation is further elaborated with 4 aspects such as experimental verifications, and embodiment is as follows：
1. control system for permanent-magnet synchronous motor structure
Control system for permanent-magnet synchronous motor structural principle is as shown in Figure 1.Control system contains current inner loop and speed outer shroud, electricity Bad current regulator can generally meet that system control requires still using routine PI controls in stream.Speed outer shroud is by original normal Rule pi regulator is designed to a kind of Recurrent Fuzzy Neural Network PI controllers, i.e., by Recurrent Fuzzy Neural Network controller and PI tune Save the compound composition speed control of device.The speed control is suitable for controlling in PI controllers and RFNN automatically under different service conditions Switched between device processed.Parameter in PI controllers is designed according to conventional Tuning, once PMSM system set-points become suddenly Change, generating state or structure interference cause Parameters variation cause system oscillation (i.e. ∑ | ei|=| ∑ ei|) or overshoot (i.e. e= 0, and de/dt ≠ 0) during phenomenon, switch S can jump to RFNN controller operating statuses automatically.Intelligent coordinated device is responsible in line traffic control The performance indicators such as the automatic switchover of system switch S, the different error domains of conventional pi regulator and RFNN controller and operation switching The relevant knowledge and fuzzy rule of condition are stored in intelligent coordinated device.
Recurrent Fuzzy Neural Network 2. (RFNN) controller design
(recurrence obscures by speed control application routine PI (proportional integration) and RFNN in control system for permanent-magnet synchronous motor Neutral net) Compound Control Strategy.Consider that dynamic mapping mistake can not be presented with static feedforward neural network in common fuzzy logic Journey and identification dynamic characteristic, therefore present invention introduces a kind of Recurrent Fuzzy Neural Network RFNN, exactly design a kind of recurrence link simultaneously Applied to the blurring layer of common fuzzy neural network, by the neuron operation in the recurrence link, feed back and preserve in time Information.So the output valve of Recurrent Fuzzy Neural Network depends on the present input, in the past input and output in the past of network, from And the recursive structure of network part or the network overall situation is established, overcome the serious Nonlinear Mapping occurred in PMSM system operations Problem.For the theory structure of Recurrent Fuzzy Neural Network (RFNN) controller as shown in Fig. 2, being made of altogether 4 parts, i.e., the 1st layer defeated Enter layer, the 2nd layer of blurring layer, the 3rd layer of fuzzy rule layer, the 4th layer of output layer.
The input quantity X of RFNN input layers1、X2Need to convert into the value between -1 to 1.Input layer interior joint output valve is：
In formula, i=1,2x1=e, x2=ec
The input variable of RFNN blurring layers is derived from the output of input layer, makees blurring calculation process to it on request.Respectively The Fuzzy Linguistic Variable of input variable is expressed as FB (negative big), FS (negative small), ZE (zero), PS (just small), PB (honest).By formula Calculate the membership function of each input fuzzy variable, membership function is chosen for typical Gaussian bases here.This layer has 2 × 5 Output node, each node output rule such as following formula：
In formula (2),I=1,2；J=1,2 ..., 5；aijFor the center of Gaussian bases, bijFor gaussian basis Function widths.
In 10 input nodes for considering RFNN blurrings layer proposed by the present invention, the recurrence ring of identical function is all devised Section structure, the input for being then blurred each node of layer are：
Here rijThe connection weight of recurrence link each unit is represented,Represent the output of blurring layer last moment Amount.
In the fuzzy rule Rotating fields of RFNN, each Fuzzy Linguistic Variable AND operation operating function is mainly realized.The layer Output and input and calculated respectively by formula (4), formula (5)：
In formula (4), k1=k2=1,2 ..., 5；K=k1k2=1,2 ..., 25
The output layer of RFNN is the final tache of network, because this layer input is defeated after computing for upper strata fuzzy rule layer Go out value, this layer of all input quantities are subjected to de-fuzzy operation and data normalized for this.The input quantity of this layer and output Value is calculated by formula (6) and formula (7)：
In formula (6), ωkRepresent fuzzy rule layer with exporting the connection weight of interlayer.The recurrence designed more than obscures In neutral net, structural parameters aij、bij、rij、ωkIt is required for after adjusting and optimizing repeatedly, can just obtains satisfied RFNN knots Structure.
3. bat-artificial bee colony hybrid algorithm optimization RFNN controller parameters
Adjustable parameter aij、bij、rij、ωkDeng there is very big influence on system performance, if still calculated using traditional BP Method optimizes study to the parameter of the Recurrent Fuzzy Neural Network, is often restrained there are low speed, is easily absorbed in local minimum etc. Bottleneck problem.Bat algorithm search speed early period is fast, can rapidly search for and lock global optimum region to be asked, but the later stage Convergence rate is slack-off during optimizing, and searching precision reduces.And artificial bee colony algorithm has very strong ability of searching optimum, but early period searches The speed of rope feasible solution is slower, as long as searching feasible solution, the search capability of the algorithm will improve quickly.For two kinds Bat algorithm (abbreviation BA) and artificial bee colony algorithm (abbreviation ABC) are blended shape by advantage existing for algorithm and deficiency, the present invention Into bat-artificial bee colony hybrid algorithm (i.e. BA-ABC), on-line study is carried out to Recurrent Fuzzy Neural Network, BP algorithm can be overcome Deficiency, improve Recurrent Fuzzy Neural Network generalization ability and quick global convergence etc..Controlled for PMSM rotational speed regulations and be System, the RFNN controllers that speed regulator is designed by the present invention substitute, and the input variable of RFNN controllers is respectively set as system Given rotating speed nrefWith the speed error e and its error rate ec of the actually detected rotating speed n of motor.Using BA-ABC algorithms pair RFNN is trained, in real time adjustment and on-line optimization aij、bij、rij、ωkDeng structural parameters, to strengthen RFNN rotational speed governors Control ability.The flow of bat-artificial bee colony hybrid algorithm optimization RFNN is as shown in figure 3, detailed process is as follows：
Step 1：Initiation parameter and bat position, that is, initialize bat population quantity D in bat algorithm, bat algorithm is most Big iterations NB, pulse (ultrasonic wave) frequency f that bat sends, maximum impulse intensity of sound S, maximum impulse frequency R0, arteries and veins Rush intensity of sound attenuation coefficient λ, pulse frequency increase coefficient δ, the position x of random initializtion bati(i=1,2 ..., D)；People Work ant colony algorithm maximum iteration limit；
Step 2：By following equation adjustment bat pulse frequency fi, renewal bat flying speed viAnd position xi, find current Optimal bat individual；
fi=fmin+(fmax-fmin)h (8)
In formula,Respectively the location of t generations and t+1 i-th bat of generation,Respectively t generations and t+1 The flying speed of i-th bat of generation；fiFor the pulse frequency of i-th bat, fmax、fminIt is respectively the maximum of bat pulse frequency Value, minimum value；H is 0~1 section uniform random number；xbestFor current global optimum position；
Step 3：Produce random number r1If r1＞ R1, then choose optimal solution from current population, and near optimal solution with Machine produces the new position x of a local solution, at this time batnewFor：
xnew=xold+τSt (11)
In formula, τ be [- 1,1] scope random number, StRepresent loudness average value in the bat population same period；
Step 4：Produce random number r2If r2＜ SiAnd mean square error (fitness) J (xi) ＜ J (x0), increase according to rule RiAnd reduce Si, i.e. bat reduces the ultrasonic pulse loudness of transmitting, while increases sound pulse transmission number, i.e.,：
In formula,For sound wave pulse loudness of i-th bat in t+1 and t iteration,For t+1 iteration When i-th bat transmitting pulse frequency, λ is [0,1] value range, δ ＞ 0；
Step 5：Fitness J (x are pressed to bat individuali) assessed, find and record the optimal bat individual in current location, If meeting bat algorithm maximum iteration NB, M optimal location solution is exported, otherwise return to step 2；
Step 6：The M optimal location solution exported by step 5 reformulates artificial bee colony, and is used as artificial bee colony algorithm In initial bee source；
Step 7：Make the primary iteration times N C=1 of artificial bee colony algorithm, led in bee colony bee when searching for nectar source according to (14) formula constantly updates current location xij, then the nectar source fitness of search is evaluated by (15) formula：
Wherein, Fiti、fiI-th of nectar source fitness and adaptive value are represented respectively
Step 8：The new nectar source of bee searching will be led, if the latter is less than the former, to be adopted compared with the fitness of green molasses source With new nectar source position substitution green molasses source position, otherwise constant and NC+1；
Step 9：By the probable value P of each nectar source position of (16) formula calculating, bee is followed with reference to P in bee colonyiSelection leads bee The new nectar source searched, and calculate its fitness value；
Step 10：Compare the new nectar source for following bee to select and green molasses source fitness value, if the former is more than the latter, with new honey Source position substitution green molasses source；Otherwise constant and NC+1；
Step 11：When iterations NC exceedes artificial bee colony algorithm cycle-index maximum limit, preserve and export group Optimal nectar source individual, is used as Recurrent Fuzzy Neural Network optimized parameter a in bodyij、bij、rij、ωkOptimal initial value.
Step 12：The global optimum's individual exported after bat-artificial bee colony hybrid algorithm optimization is substituted into recurrence to obscure Parameters of Neural Network Structure, utilizes BP algorithm training network by input sample, finally makes mean square error (the target letter of network Number) MSE values minimum, so as to export the optimum structure parameter a of Recurrent Fuzzy Neural Networkij、bij、rij、ωk；MSE is defined as：
In formula, n is sample number, YK, pFor training sample p k-th of output node reality output；QK, pFor the corresponding phase Hope output, 1/ (MSE+1) is defined as the fitness function in bat-artificial bee colony algorithm here.
4. simulation analysis and experimental verification
(1) simulation analysis
The simulation model of control system for permanent-magnet synchronous motor is established using Matlab/Simulink and is analyzed.Emulation Chosen with permanent magnet synchronous motor parameter as follows：Rated power 500W, rated speed ne=1500r/min, stator phase resistance Rs= 4.475 Ω, stator phase winding self-induction L=0.025H, mutual inductance M=-0.0075H, rotary inertia J=0.00187kgm2.According to Multiple simulated experiment repeatedly, bat-artificial bee colony algorithm initiation parameter that the present invention chooses are：Bat population quantity is 150, Maximum iteration is 200, and bat pulse frequency f is [0,1], maximum impulse loudness S=0.45, maximum impulse frequency R0= 0.75, pulse loudness attenuation coefficient λ=0.85, pulse frequency increase coefficient δ=0.20；ABC algorithm iteration number limiting values are Limit=180；The error precision ε of network training is 10-4.Conventional PI control device and Recurrent Fuzzy Neural Network (RFNN) control Device is used separately as the speed control of PMSM systems.Current regulator still uses pi regulator (K in control systemp=3, Ki= 37).In given rotating speed nrefSystem operation under the conditions of=1500r/min, and 10Nm loads of uprushing as t=0.6s, rotating speed become Change curve such as Fig. 4.1. represent the rate curve during effect of conventional PI control device, 2. for RFNN controllers when rate curve, it is 2. bent Line is superior to 1. curve in response speed, overshoot and antijamming capability etc., shows through bat-artificial bee colony algorithm optimization RFNN controllers can obtain more excellent index and more preferable control effect.
(2) experimental result
Optimize the control performance of RFNN controllers for further verification bat-artificial bee colony algorithm, establish permanent magnet synchronous electric Machine controls experimental provision and carries out verification analysis, and experimental principle schematic diagram is as shown in Figure 5.TI companies DSP cores are applied in system control Piece TMS320F28035, system data acquisition, signal processing are handled by the dsp chip with functions such as controls and completed.System control The peripheral circuit of system includes analog line driver, inverter circuit, current sample detection circuit, oscillograph, PC machine, CAN communication electricity Road etc..Given speed instruction is still 1500r/min, and impact 10Nm was loaded at 0.7 second.Fig. 6 is rotating speed n response curves, Fig. 7 is torque T response curves, and as seen from the figure, command value of the motor speed when reaching stable state only than given rotating speed is declined slightly, It is about very small in 1447r/min, error；The fluctuation of speed influences little when shock load is disturbed.The torque change curve of Fig. 7 compared with To be smooth, although load undergos mutation torque change procedure also than shallower, and the overshoot produced is smaller.Experimental result Show, the RFNN controllers fast response time through bat-artificial bee colony algorithm optimization, stable state accuracy is high, Ability of Resisting Disturbance is strong, tool There is higher robustness.
The above, is only presently preferred embodiments of the present invention, the technical spirit scope of every foundation the present patent application patent It is interior to any modification made for any of the above embodiments, equivalent change and modification, all belong to the covering scope of technical solution of the present invention.

Claims (2)

1. a kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network, it is characterised in that control system contains Current inner loop and speed outer shroud, the current regulator of current inner loop are designed as a kind of recurrence using routine PI controls, speed outer shroud Fuzzy neural network PI controllers, the Recurrent Fuzzy Neural Network PI controllers by Recurrent Fuzzy Neural Network controller and The compound composition of pi regulator.
2. Recurrent Fuzzy Neural Network controller according to claim 1, it is characterised in that using bat-artificial bee colony Hybrid algorithm optimizes Recurrent Fuzzy Neural Network, a in adjustment in real time and on-line optimization Recurrent Fuzzy Neural Networkij、 bij、rij、ωkDeng structural parameters, further, the bat-artificial bee colony hybrid algorithm, searches with reference to the early period of bat algorithm It is fast to seek speed, and can search for rapidly to global optimum region, and artificial bee colony algorithm ability of searching optimum is strong but searches early period The features such as rope feasible solution speed is slower, a kind of bat-artificial bee colony of generation is blended by bat algorithm and artificial bee colony algorithm Hybrid algorithm, artificial bee colony is reformulated by the M optimal location solution exported by bat algorithm optimization, and is used as artificial bee colony Initial bee source in algorithm, then exports global optimum's individual after artificial bee colony algorithm computing optimization, substitutes into recurrence and obscures Parameters of Neural Network Structure, utilizes BP algorithm training network by input sample, finally makes mean square error (the target letter of network Number) MSE values minimum, export the optimum structure parameter a of Recurrent Fuzzy Neural Networkij、bij、rij、ωk。
CN201711235087.0A 2017-11-20 2017-11-20 A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network Pending CN107968613A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711235087.0A CN107968613A (en) 2017-11-20 2017-11-20 A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711235087.0A CN107968613A (en) 2017-11-20 2017-11-20 A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network

Publications (1)

Publication Number Publication Date
CN107968613A true CN107968613A (en) 2018-04-27

Family

ID=61999200

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711235087.0A Pending CN107968613A (en) 2017-11-20 2017-11-20 A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network

Country Status (1)

Country Link
CN (1) CN107968613A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108649817A (en) * 2018-06-15 2018-10-12 湖北德普电气股份有限公司 A kind of self-adaptive initial pulse design method based on Technics of Power Electronic Conversion device
CN108762067A (en) * 2018-04-28 2018-11-06 南京理工大学 A kind of the networking Synchronizing Control Devices and acquisition methods of memristor neural network
CN109276811A (en) * 2018-10-09 2019-01-29 郑州轻工业学院 The setting device and setting method in zero magnetic field point position
CN110082717A (en) * 2019-04-30 2019-08-02 上海海事大学 A kind of underwater wireless sensor node positioning method
CN110266228A (en) * 2019-07-05 2019-09-20 长安大学 Surface permanent magnetic Synchronous Machine Models forecast Control Algorithm based on BP neural network
CN110824921A (en) * 2019-11-13 2020-02-21 华中科技大学 AC servo speed regulation system control parameter self-tuning method based on bat algorithm
TWI715425B (en) * 2018-10-12 2021-01-01 日商三菱電機股份有限公司 Positioning control device and positioning method
CN112713830A (en) * 2020-12-21 2021-04-27 华中科技大学 Permanent magnet synchronous motor speed regulation system and multi-target optimal state feedback control method

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103095204A (en) * 2013-01-09 2013-05-08 重庆交通大学 Control system and control method of anti-interference compensation of servo motor
CN104406593A (en) * 2014-12-03 2015-03-11 广西民族大学 Method for determining optimal route of airway of unmanned aerial vehicle
CN106384153A (en) * 2016-09-18 2017-02-08 河海大学常州校区 WSAN actuator task distribution method based on BA-BPNN data fusion
CN107085763A (en) * 2017-03-31 2017-08-22 无锡开放大学 A kind of driving motor for electric automobile system performance evaluation method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103095204A (en) * 2013-01-09 2013-05-08 重庆交通大学 Control system and control method of anti-interference compensation of servo motor
CN104406593A (en) * 2014-12-03 2015-03-11 广西民族大学 Method for determining optimal route of airway of unmanned aerial vehicle
CN106384153A (en) * 2016-09-18 2017-02-08 河海大学常州校区 WSAN actuator task distribution method based on BA-BPNN data fusion
CN107085763A (en) * 2017-03-31 2017-08-22 无锡开放大学 A kind of driving motor for electric automobile system performance evaluation method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
乔维德: "基于蚁群模糊神经网络的无刷直流电机仿真控制研究", 《电气传动自动化》 *
高姗 等: "改进人工蜂群算法求解信道分配问题", 《计算机与应用化学》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108762067A (en) * 2018-04-28 2018-11-06 南京理工大学 A kind of the networking Synchronizing Control Devices and acquisition methods of memristor neural network
CN108649817A (en) * 2018-06-15 2018-10-12 湖北德普电气股份有限公司 A kind of self-adaptive initial pulse design method based on Technics of Power Electronic Conversion device
CN109276811A (en) * 2018-10-09 2019-01-29 郑州轻工业学院 The setting device and setting method in zero magnetic field point position
CN109276811B (en) * 2018-10-09 2022-08-05 郑州轻工业学院 Setting device and setting method for zero magnetic field point position
TWI715425B (en) * 2018-10-12 2021-01-01 日商三菱電機股份有限公司 Positioning control device and positioning method
CN110082717A (en) * 2019-04-30 2019-08-02 上海海事大学 A kind of underwater wireless sensor node positioning method
CN110266228A (en) * 2019-07-05 2019-09-20 长安大学 Surface permanent magnetic Synchronous Machine Models forecast Control Algorithm based on BP neural network
CN110266228B (en) * 2019-07-05 2021-03-23 长安大学 Surface permanent magnet synchronous motor model prediction control method based on BP neural network
CN110824921A (en) * 2019-11-13 2020-02-21 华中科技大学 AC servo speed regulation system control parameter self-tuning method based on bat algorithm
CN112713830A (en) * 2020-12-21 2021-04-27 华中科技大学 Permanent magnet synchronous motor speed regulation system and multi-target optimal state feedback control method

Similar Documents

Publication Publication Date Title
CN107968613A (en) A kind of permanent magnet synchronous motor rotational speed governor based on Recurrent Fuzzy Neural Network
Chedid et al. Adaptive fuzzy control for wind-diesel weak power systems
CN104331543A (en) Fault diagnostic expert system for marine electrical propulsion system and establishing method thereof
CN105281615A (en) Method for optimizing brushless DC motor fuzzy controller based on improved particle swarm algorithm
CN111812968B (en) Fuzzy neural network PID controller-based valve position cascade control method
CN106357184A (en) Temperature compensation method for output torque of permanent magnet synchronous motor for vehicle based on neural network
CN108365784A (en) Based on the control method for brushless direct current motor for improving PSO-BP neural networks
CN103197596B (en) A kind of digital control processing parameters self-adaptive fuzzy control rule optimization method
CN104639001A (en) Servo motor control method integrating sliding mode control and fractional order neural network control
Aissaoui et al. Application of fuzzy logic in control of electrical machines
Moussavi et al. Speed control of permanent magnet DC motor by using combination of adaptive controller and fuzzy controller
Yadav et al. Comperative performance analysis of PMSM drive using MPSO and ACO techniques
Wang et al. Multiscenarios parameter optimization method for active disturbance rejection control of PMSM based on deep reinforcement learning
Sakunthala et al. A study on fuzzy controller and neuro-fuzzy controller for speed control of PMSM motor
CN110138293A (en) A kind of BP neural network optimization chaotic control method based on double-fed wind power system
Malhotra et al. Design of embedded hybrid fuzzy-GA control strategy for speed control of DC motor: a servo control case study
CN103499920B (en) Control parameter optimization method and system through vector time series prediction and expert fuzzy transformation ratio
CN110488600B (en) lQR optimized brushless DC motor speed regulation neural network PID controller
Zhang et al. An adaptive fuzzy PID control of hydro-turbine governor
CN110262220A (en) One kind welding pulse current pid parameter control method based on ant group algorithm MIG
Cai et al. Identification and control of PMSM using adaptive BP-PID neural network
Ali et al. Parameter identification using ANFIS for magnetically saturated induction motor
Gupta et al. Position control of servo motor using sliding mode fuzzy controller
Jacomini et al. Implementation of a Neuro-Fuzzy direct torque and reactive power control for DFIM
Ankarao et al. Dynamic Performance Analysis of Reactive Power and Improved Rotor Flux Based MRAS for Induction Motor Drives Employing PI and Fuzzy Controller

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20180427