CN103259479A - Method for observing left inverse state of neural network of permanent magnet synchronous motor - Google Patents
Method for observing left inverse state of neural network of permanent magnet synchronous motor Download PDFInfo
- Publication number
- CN103259479A CN103259479A CN201310205193XA CN201310205193A CN103259479A CN 103259479 A CN103259479 A CN 103259479A CN 201310205193X A CN201310205193X A CN 201310205193XA CN 201310205193 A CN201310205193 A CN 201310205193A CN 103259479 A CN103259479 A CN 103259479A
- Authority
- CN
- China
- Prior art keywords
- synchronous motor
- permagnetic synchronous
- observation
- neural network
- left inverse
- 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.)
- Granted
Links
Images
Landscapes
- Control Of Ac Motors In General (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The invention discloses a method for observing a left inverse state of a neural network of a permanent magnet synchronous motor, and belongs to the technical field of motor drive control. An observer of the left inverse state of the neural network is constructed through the method for observing the left inverse state of the neural network to achieve the state observation of a multivariable and strong-coupled time-varying nonlinear system of a permanent magnet synchronous motor system without a sensor. According to the method for observing the left inverse state of the neural network of the permanent magnet synchronous motor, the observation problem of a complex system mutually coupled by the current, the voltage and the speed of a stator is translated into a simple observation problem of a unit mapping system, therefore, high-precision observation of the rotate speed of the permanent magnet synchronous motor and the angular position of a rotor is achieved, and the interference of system noise is overcome. An observation system is made to have excellent dynamic performance, static performance and anti-interference and high-precision tracking performance.
Description
Technical field
The invention belongs to motor-driven control technology field, more precisely, the present invention relates to a kind of permagnetic synchronous motor neural net left inverse state observation method.
Background technology
As everyone knows, (it in the application in motor-driven field more and more widely for permanent magnet synchronous motor, characteristics such as the high power density that PMSM) possesses, high torque (HT) inertia ratio, high reliability because permagnetic synchronous motor.State observer can make the PMSM control system obtain PMSM system state amount rotor angle location and rotary speed information under the prerequisite that need not installation site or velocity transducer, saves manufacturing cost when improving the system reliability operation.Therefore, Chinese scholars has developed the Speedless sensor control technology in succession, has proposed different motor speeds and position detection scheme and has been widely used in PMSM control field.
In the state observer that proposes, the basic electromagnetic of utilizing motor that has concerns and the equation of motion identifying motor rotating speed and other motor status amounts at present, this method is calculated simple, is easy to realize, but relies on mathematical models, the parameter of electric machine is changed sensitivity, and performance is unstable in actual applications.Discrimination method based on expanded Kalman filtration algorithm or augmentation expanded Kalman filtration algorithm is also arranged, good at the low-speed range identification effect, yet these algorithms all are based upon on the system linearity model basis, and for non linear system, its identification result has bigger deviation.Also has the PMSM rotating speed discrimination method based on the Theory of Stability design, these class methods guarantee the asymptotic convergence of system state amount with Lyapunov equation and Popov hyperstability theory, functional when system stability, but generally need simultaneously the parameter of electric machine to be done online observation, to guarantee the identification result precision, so the model complexity height, amount of calculation is big.In addition, adopt sliding mode observer observation motor speed in addition, the motor speed when this method is applicable to the identification high speed and be easy to realize, but in the quantity of state sudden change and be subjected under the situation that external noise disturbs identification effect bad, be prone to the phenomenon of buffeting.Though this phenomenon can be improved by adding filter, in fact this is difficult to realize in actual applications.
And, for the PMSM state observer, often need when moving, motor observe system state amount real-time and accurately.Simultaneously, when practical application because the reliability of observed result is directly connected to the accuracy of Electric Machine Control, for the reliability requirement of observed result than higher.Therefore, need design in system noise and external disturbance all very under the serious situation state observer with high reliability.
A kind of nerve network reverse discrimination method at the three-phawse arc furnace electrode control system has been proposed in the existing patent " CN1794120A ".This method utilizes the arc furnace phase current as the input structure arc furnace phase voltage identification model of RBF neural net.And this identification model is reverse, as direct inverse control device model, thereby realize arc furnace three-phase current decoupling zero control.This discrimination method focuses on realizes that system decoupling control, essence of its structure identifier are in order to construct controller, but not observation system directly can not measured quantity of state.And this patent does not provide the building method of nerve network reverse identifier according to the inverse system theory.Therefore, need seek a kind of new method, according to the characteristics of PMSM control system, constructing system is the left inverse state observer of direct measured state amount not.
The nerve network reverse flexible measurement method has been applied to the process control occasion of this class of bioleaching process, for example in document " expansion of nerve network reverse flexible measurement method and the application in bioleaching process ", (publish in March, 2012 Chinese journal of scientific instrument the 33rd volume, the 3rd phase, the 661-669 page or leaf), a kind of neural net flexible measurement method has been proposed.This method can solution the never direct estimation problem of measured state.Its structure as shown in Figure 1.Fig. 1 is a kind of traditional neural net left inverse identification structure, by original system 1 include the sensor subsystem 21 component unit mapped system 2 of connecting with the soft measuring appliance 22 of traditional left inverse, system state amount is observed.Wherein traditional soft measuring appliance 22 of left inverse is made up of some pure differential links and general static neural network.Its principle is: at first suppose exist in the inside of original system 1 one with directly measured state amount x be input, with direct measured state amount z be output include sensor subsystem 21, set up the Mathematical Modeling that this includes sensor subsystem 21 by modeling algorithm then, and obtaining the inverse system of " including transducer ", this inverse system is exactly the soft measuring appliance 22 of traditional left inverse that will set up.Because traditional soft measuring appliance 22 of left inverse is made of pure differential link and traditional neural net in this method, and the introducing of pure differential link is big for the state variation amplitude, the serious system of system noise is such as control system for permanent-magnet synchronous motor, can cause the amplification noise, cause the consequence of vibration, the quantity of state that identification quickly and accurately is required.Therefore, need make adjustment to the structure of contrary soft measuring appliance, the unfavorable factor of avoiding the pure differential link to bring is constructed a left inverse state observer that is applicable to control system for permanent-magnet synchronous motor.
Summary of the invention
The objective of the invention is: at the deficiency of permagnetic synchronous motor state observation technology in the prior art, a kind of novel permagnetic synchronous motor neural net left inverse state observation method is provided, realizes the high performance control under the Permanent-magnet Synchronous-motor Speed Servo System line position-sensor-free condition.This state observation method will be followed the tracks of differentiator (tracking differentiator, TD) be applied to neural net left inverse identifier, thereby overcome system not modeling disturb dynamically, suppress parameter perturbation, improve dynamic responding speed and the steady-state tracking precision of identification effect, and ready for realizing the high-performance robust control.
Specifically, the present invention adopts following technical scheme to realize, comprises the following steps:
1) will combine with voltage source inverter under the space vector pulse width modulation mode against the Park conversion and constitute expansion inverter control section to drive permagnetic synchronous motor, connect current detecting and computing module simultaneously, constitute an integral body and form the compound object of observation of permagnetic synchronous motor;
2) set up the equivalent mathematical model that permagnetic synchronous motor includes transducer, and according to this equivalence Mathematical Modeling, at the coupling between motor speed, voltage and the stator current, analyzing on reversible basis, a permagnetic synchronous motor left side, adopt multilayer feedforward neural network to add the permagnetic synchronous motor neural net left inverse state observer that some Nonlinear Tracking differentiator constructing virtuals include transducer, introduce parameter and weights coefficient that multilayer feedforward neural network is regulated in the training of steepest decline learning algorithm;
3) will train permagnetic synchronous motor neural net left inverse state observer be connected on component unit pseudo-linear system after the compound object of observation of permagnetic synchronous motor, this moment, original system was become 1 rotating speed to observe sub-linear system and 1 position detection subsystem by equivalence;
4) unit's of utilization pseudo-linear system, motor speed and the rotor angle location of observation permagnetic synchronous motor.
Of the present invention being further characterized in that: the compound object of observation of described permagnetic synchronous motor, it is output as permagnetic synchronous motor in the following stator voltage of two-phase rotating coordinate system: d shaft voltage u
Sd, q shaft voltage u
Sq, it directly can measure state variable is the electric current of permagnetic synchronous motor under the two-phase rotating coordinate system: d shaft current i
Sd, q shaft current i
Sq, the rotational speed omega that it directly can not be measured state variable and be output as permagnetic synchronous motor
rAnd rotor angular position thetar.
Of the present invention being further characterized in that: described multilayer feedforward neural network has 5 input nodes and 2 output nodes, and described tracking differentiator is 4, and 5 input variables of multilayer feedforward neural network are
Output variable is ω
rAnd θ, wherein
Be respectively through following the tracks of the u of differentiator gained
Sd, u
Sq, i
Sd, i
SqTracking signal,
Be the i through tracking differentiator gained
SqThe tracking differential signal.
Of the present invention being further characterized in that: described multilayer feedforward neural network adopts three-layer network, ground floor is input layer, input number of nodes is 5, neuron is the input node, represent the input language variable, this layer only is used for transmitting signal and arrives one deck down, the second layer is hidden layer, the node number is 16, each node is represented a linguistic variable value, any Nonlinear Mapping between being used for realizing importing and exporting, and the 3rd layer is output layer, utilize error weights and the threshold value of each node between dragover modification level and the layer successively, export simultaneously.
Of the present invention being further characterized in that: parameter and the weights coefficient of quoting steepest decline learning algorithm training adjusting multilayer feedforward neural network described step 2) specifically may further comprise the steps:
2-1) with { u
Sd, u
SqBe added to 2 inputs of the compound object of observation of permagnetic synchronous motor system respectively, gather the rotor velocity ω of permagnetic synchronous motor with the predetermined sampling period
rAnd rotor angular position thetar and phase current i
a, i
b, detect and computing module acquisition desired data { i through rate of current
Sd, i
SqAnd preserve;
2-2) with the data-signal { u that preserves
Sd, u
Sq, i
Sd, i
SqRespectively application tracking differentiator off-line extract and ask the single order of electric current, and then signal is done standardization processing, form the training sample set of neural net
2-3) use steepest decline learning algorithm off-line training fuzzy neural network, adjust weights and the threshold value coefficient of each node of network, neural net output mean square error precision is remained in the predetermined mean square error precision.
Of the present invention being further characterized in that: the described predetermined sampling period is 5ms.
Of the present invention being further characterized in that: described predetermined mean square error precision is 0.0005.。
Beneficial effect of the present invention is as follows: the left inverse neural net state observation method that the present invention adopts has in theory and pro forma unification, and clear physics conception is directly perceived, and using method is simple and clear.Add the tracking differentiator with neural net and construct the left inverse observation system that compound object of observation includes transducer, be completely free of traditional observation procedure for the dependence of permagnetic synchronous motor system mathematic model and parameter.Left inverse observation system and former compound object of observation include the pseudo-linear system of the compound formation of transducer, can realize the state observation of original system, are conducive to the comprehensive, simple in structure of system, the system robustness height.Among the present invention, the application of following the tracks of differentiator has broken through traditional contrary soft measuring appliance building method.The parameter of following the tracks of differentiator by reasonable adjusting is the tracing preset signal fast, and from by rational extraction differential signal the signal of noise pollution, can carry out the high accuracy state observation easily.Therefore add and follow the tracks of the range of application that differentiator can be expanded neural net left inverse discrimination method, make it be applicable to motion control field.
Description of drawings
Fig. 1 is traditional neural net left inverse identification structure schematic diagram.
The neural net left inverse state observer structural representation that Fig. 2 proposes for the present invention.
Fig. 3 is the compound object of observation structural representation of permagnetic synchronous motor system of the present invention.
Fig. 4 is identification mapped system equivalent schematic of the present invention.
Fig. 5 does not have transducer control structure schematic diagram for the permagnetic synchronous motor based on neural net left inverse state observer of the present invention.
Fig. 6 is control system for permanent-magnet synchronous motor experiment porch structural representation of the present invention.
Fig. 7 is given as ramp signal and the external loading control identification effect figure when constant for motor speed of the present invention.
Control identification effect figure when Fig. 8 is given as constant signal and external loading sudden change for motor speed of the present invention.
Embodiment
With reference to the accompanying drawings and in conjunction with example the present invention is described in further detail.
The present invention is by using neural net left inverse state observation method construct left inverse neural net state observer, realizes the control system for permanent-magnet synchronous motor under the no transducer multivariable, close coupling the time change non linear system state observation.The observation problem of the complication system that this method intercouples stator current, voltage and speed is converted into the observation problem of simple unit mapped system, thereby realized the high accuracy observation to permagnetic synchronous motor rotating speed and rotor angle location, overcome the interference of system noise, made observation system have good dynamic and static performance, anti-interference and high precision tracking performance.
The building method of permagnetic synchronous motor neural net left inverse state observer of the present invention comprises following key step:
1, forms the compound object of observation of permagnetic synchronous motor.The integral body that the expansion inverter control section that formed by the voltage source inverter under contrary Park conversion, the space vector pulse width modulation mode and permagnetic synchronous motor body are formed is carried out equivalence, making it similarly, equivalence becomes direct current machine, with current detection module, constitute an integral body and form the compound object of observation of permagnetic synchronous motor then.
2, tectonic system includes transducer.Analyze the directly observer state variable in the permagnetic synchronous motor hybrid system, seek itself and the direct mathematical relationship between measured state variable motor speed, the rotor angle location, be input with motor speed, rotor angle location, with hybrid system output and directly the observer state variable be output, make up the virtual transducer that includes.
3, the reversible Analysis of Existence in a system left side.Can obtain the Mathematical Modeling of whole permagnetic synchronous motor composite controlled object under vector control mode by analysis and derivation is the two-phase rotating coordinate system, it is the Third-Order Nonlinear Differential Equations group under the d-q coordinate system, and according to the inverse system theoretical proof this to include transducer be that partial left is reversible in its working region, and then derive the left inverse system that this includes transducer.
4, the neural net left inverse state observer of the compound object of observation of structure.Employing have study and performances such as Function approximation capabilities, predictive ability all more excellent multilayer feedforward neural network add 4 tracking differentiators and come constructing neural network left inverse state observer, realize the state observation of former compound object of observation.Wherein multilayer feedforward neural network be one have 5 the input nodes, the three-layer network of 2 output nodes, by weights and the threshold value coefficient that genetic algorithm and steepest decline learning algorithm combine and determine and adjust each node of network, realize the left inverse state observation function of compound object of observation.
Fig. 2 has provided the new neural network left inverse state observer structural representation that the present invention proposes.As shown in Figure 2, the new neural network left inverse state observer 3 that the present invention proposes is constituted with multilayer feedforward neural network 32 by following the tracks of differentiator group 31.U wherein
1~u
pBe system's input variable, z
1~z
L-mBe the direct measured state amount of system, x
1~x
mBe not direct measured state amount,
Be the tracking signal through each variable of TD gained,
Be the tracking differential signal through each variable of TD gained.
Concrete enforcement of the present invention is divided into following a few step:
1, the compound object of observation of structure permagnetic synchronous motor system.
As shown in Figure 3, the compound object of observation 4 of permagnetic synchronous motor system of the present invention is connected and composed by expansion inverter control section 41, current detecting and computing module 42 and permagnetic synchronous motor 43.Wherein, expansion inverter control section 41 is formed the permagnetic synchronous motor 43 of connecting thereafter by contrary Park conversion and the voltage source inverter under the space vector pulse width modulation mode.Current detecting and computing module 42 are connected in parallel between expansion inverter control section 41 and the permagnetic synchronous motor 43.
The compound object of observation 4 of permagnetic synchronous motor system be input as the stator voltage of permagnetic synchronous motor 43 under (d-q coordinate system) under the two-phase rotating coordinate system: d shaft voltage u
Sd, q shaft voltage u
Sq, be designated as u=[u
1, u
2]
T=[u
Sd, u
Sq]
TWhat current detecting and computing module 42 was used for obtaining the compound object of observation 4 of permagnetic synchronous motor system directly can measure state variable, i.e. the electric current of permagnetic synchronous motor 43 under the d-q coordinate system: d shaft current i
Sd, q shaft current i
Sq, be designated as x=[x
1, x
2]
T=[i
Sd, i
Sq]
TThe compound object of observation 4 of permagnetic synchronous motor system directly can not measure the rotational speed omega that state variable and system are output as permagnetic synchronous motor 43
rAnd rotor angular position thetar, be designated as y=[y
1, y
2]
T=[θ, ω
r]
Tu
Sd, u
SqAfter contrary Park conversion, obtain the voltage u under the two-phase rest frame (alpha-beta coordinate system)
S α, u
S β, obtain voltage u under the three phase static coordinate system (ABC coordinate system) by inverter through space vector pulse width modulation again
SA, u
SB, u
SC, and input to permagnetic synchronous motor 43.
2, set up the equivalent mathematical model that permagnetic synchronous motor includes transducer.
At first set up the Mathematical Modeling of compound object of observation, operation principle based on permagnetic synchronous motor, set up the Mathematical Modeling of permagnetic synchronous motor 43, this Mathematical Modeling is the three rank differential equation group under the d-q coordinate system that obtain after coordinate transform (Clarke conversion, Park conversion).Secondly select wherein current equation and rotating speed equation engineering to include sensor model according to the state identification demand, according to the inverse system theory, through deriving as can be known, the inverse system of the subsystem that this second order differential equation fabric becomes exists, and a left side is reversible.
3, include the equivalent mathematical model of transducer according to permagnetic synchronous motor, adopt feedforward neural network to add and follow the tracks of differentiator structure permagnetic synchronous motor neural net left inverse state observer.
Fig. 4 has provided two unit mapped systems (is that 61, one of rotor velocity mapped systems are rotating speed mapped systems 62) that the permagnetic synchronous motor neural net left inverse state observer 5 identification mapped system 6 that compound object of observation 4 series connection are combined into the permagnetic synchronous motor system and equivalence thereof become.As shown in Figure 4, permagnetic synchronous motor neural net left inverse state observer 5 is made of feedforward neural network 51 and 4 adjustable tracking differentiators of parameter with 5 input nodes, 2 output nodes.u
Sd, u
SqBe system's input variable, i
Sd, i
SqBe the direct measured state amount of system, ω
rReaching θ is direct measured state amount,
Be the tracking signal through each variable of TD gained,
Be the i through the TD gained
SqFollow the tracks of differential signal.
Feedforward neural network 51 adopts three-layer network.Ground floor is input layer, and input number of nodes is 5, and neuron represents the input language variable for the input node, and this layer only is used for transmitting signal and arrives one deck down.Second layer hidden layer, node number are 16, and each node is represented a linguistic variable value, any Nonlinear Mapping between being used for realizing importing and exporting.The 3rd layer is that output layer utilizes error weights and the threshold value of each node between dragover modification level and the layer successively, exports simultaneously.So feedforward neural network 51 is formed permagnetic synchronous motor neural net left inverse state observer 5 with 4 tracking differentiators, the output of feedforward neural network 51 is exactly the output of permagnetic synchronous motor neural net left inverse state observer 5.
4, carry out the adjustment of feedforward neural network parameter and weight coefficient value.
Utilize steepest decline learning algorithm, the study of feedforward neural network 51 be 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 the compound object of observation 4 of permagnetic synchronous motor system respectively, gather the rotor velocity ω of permagnetic synchronous motor 43 with the sampling period of 5ms
rAnd rotor angular position thetar and phase current i
a, i
b, detect and computing module 42 acquisition desired data { i through rate of current
Sd, i
SqAnd preserve.Concrete, rotor velocity ω
rAnd rotor angular position thetar can use photoelectric encoder to obtain, and rate of current detects with computing module 42 earlier with phase current i
a, i
bBe converted to the current i under the two-phase rest frame (alpha-beta coordinate system)
S α, i
S β, be converted into i again
Sd, i
Sq
2. with the data-signal { u that preserves
Sd, u
Sq, i
Sd, i
SqRespectively application tracking differentiator off-line extract and ask the single order of electric current, and then signal is done standardization processing, thereby forms the training sample set of neural net
3. use steepest decline learning algorithm off-line training fuzzy neural network, adjust weights and the threshold value coefficient of each node of network, neural net output mean square error precision is remained in 0.0005.
5, constitute the identification mapped system.
With train permagnetic synchronous motor neural net left inverse state observer 5 be connected on component unit pseudo-linear system after the compound object of observation of permagnetic synchronous motor system 4, this moment, former compound object of observation equivalence became 1 rotating speed to observe sub-linear system and 1 position detection subsystem.Shown in the right figure of Fig. 4, two subsystem inputs that equivalence becomes are respectively θ, ω
r, corresponding output also is θ, ω
rRealized the purpose that complication system output reappears.
6, motor speed and the rotor angle location of observation permagnetic synchronous motor.
As shown in Figure 5, permagnetic synchronous motor neural net left inverse state observer 5 utilizes and follows the tracks of the signal { u that differentiator calculates system acquisition
Sd, u
Sq, i
Sd, i
SqRefining and obtain its differential, feedforward neural network 51 identifying motors that recycling trains can not be surveyed quantity of state θ, ω
r, and be fed back in the control system, realize the closed-loop control of system.
Fig. 6 has provided the situation of the control system for permanent-magnet synchronous motor experiment porch of the present invention's use.This experiment porch is made up of permagnetic synchronous motor body 43, dSPACE real-time emulation system 81, special intelligent power model (IPM) 82 and industrial computer 83.Wherein, the dSPACE real-time emulation system be by a cover of German dSPACE company exploitation based on the control system of MATLAB/Simulink in exploitation and the test environment implemented under the environment.As experiment porch, realized the complete seamless link with MATLAB/Simulink/RTW based on the dSPACE real-time emulation system.The dSPACE real-time emulation system has real-time, reliability height, advantage such as 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 that code generates automatically/downloads and test/kit debugged.Utilize its powerful software and hardware experiments platform can realize the high accuracy state observation of control system for permanent-magnet synchronous motor, thereby further realize accurately control, the control algolithm of finishing 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.
Particularly, the dSPACE real-time emulation system is used for realizing the compound object of observation 4 of control permagnetic synchronous motor system, and its subsidiary module comprises analog input ADC module, simulation output DAC module, incremental encoder interface, PWM output, input part, Hall element, magnetic brake unit, industry control display module and DSP subsystem.Special intelligent power model (IPM) is Mitsubishi PS12036 special intelligent power model ASPIM, is used for realizing inverter.Controlled permagnetic synchronous motor software environment comprises that mainly real-time code generates download software RTI and Comprehensive Experiment and test environment software ControlDesk; Controlled permagnetic synchronous motor model is 1FT6072.
The experiment control program of design downloads to control board by host computer; send the experiment enabling signal by the Comprehensive Experiment interface; the control system independent operating; 6 road pwm control signals of control board output are to the Intelligent Power Module drive motors; test section collection electric current, voltage, guard signal feed back to control board; utilize dSPACE to realize function output motor speed and the rotor angle location of permagnetic synchronous motor neural net left inverse state observer 5; and feed back in the control loop, but off-line or online modification parameter control motor are to reach high accuracy stable operation.
Fig. 7 has provided the embodiment of the invention and has been given as ramp signal and the external loading control identification effect figure when constant at motor speed.Fig. 8 has provided the control identification effect figure of the embodiment of the invention when motor speed is given as constant signal and external loading sudden change.NNLI among the figure is based on general neural net left inverse identification result, and TD-NNLI is based on the neural net left inverse identification result of following the tracks of differentiator.Fig. 7 and Fig. 8 all show the present invention state variable polluted by system noise and the situation of load disturbance under, can effectively suppress system noise to the influence of the dynamic property of rotating speed identification, show more superior interference free performance simultaneously, for the precision that improves system's closed-loop control and the robustness of system provide assurance.
Though the present invention is with preferred embodiment openly as above, embodiment be not limit of the present invention.Without departing from the spirit and scope of the invention, any equivalence of doing changes or retouching, belongs to the present invention's protection range equally.Therefore protection scope of the present invention should be standard with the application's the content that claim was defined.
Claims (7)
1. a permagnetic synchronous motor neural net left inverse state observation method is characterized in that, may further comprise the steps:
1) will combine with voltage source inverter under the space vector pulse width modulation mode against the Park conversion and constitute expansion inverter control section to drive permagnetic synchronous motor, connect current detecting and computing module simultaneously, constitute an integral body and form the compound object of observation of permagnetic synchronous motor;
2) set up the equivalent mathematical model that permagnetic synchronous motor includes transducer, and according to this equivalence Mathematical Modeling, at the coupling between motor speed, voltage and the stator current, analyzing on reversible basis, a permagnetic synchronous motor left side, adopt multilayer feedforward neural network to add the permagnetic synchronous motor neural net left inverse state observer that some Nonlinear Tracking differentiator constructing virtuals include transducer, introduce parameter and weights coefficient that multilayer feedforward neural network is regulated in the training of steepest decline learning algorithm;
3) will train permagnetic synchronous motor neural net left inverse state observer be connected on component unit pseudo-linear system after the compound object of observation of permagnetic synchronous motor, this moment, original system was become 1 rotating speed to observe sub-linear system and 1 position detection subsystem by equivalence;
4) unit's of utilization pseudo-linear system, motor speed and the rotor angle location of observation permagnetic synchronous motor.
2. permagnetic synchronous motor neural net left inverse state observation method according to claim 1 is characterized in that, the compound object of observation of described permagnetic synchronous motor, and it is output as permagnetic synchronous motor in the following stator voltage of two-phase rotating coordinate system: d shaft voltage u
Sd, q shaft voltage u
Sq, it directly can measure state variable is the electric current of permagnetic synchronous motor under the two-phase rotating coordinate system: d shaft current i
Sd, q shaft current i
Sq, the rotational speed omega that it directly can not be measured state variable and be output as permagnetic synchronous motor
rAnd rotor angular position thetar.
3. permagnetic synchronous motor neural net left inverse state observation method according to claim 2, it is characterized in that, described multilayer feedforward neural network has 5 input nodes and 2 output nodes, and described tracking differentiator is 4, and 5 input variables of multilayer feedforward neural network are
Output variable is ω
rAnd θ, wherein
Be respectively through following the tracks of the u of differentiator gained
Sd, u
Sq, i
Sd, i
SqTracking signal,
Be the i through tracking differentiator gained
SqThe tracking differential signal.
4. permagnetic synchronous motor neural net left inverse state observation method according to claim 3, it is characterized in that, described multilayer feedforward neural network adopts three-layer network, ground floor is input layer, input number of nodes is 5, neuron is the input node, represent the input language variable, this layer only is used for transmitting signal and arrives one deck down, and the second layer is hidden layer, the node number is 16, each node is represented a linguistic variable value, any Nonlinear Mapping between being used for realizing importing and exporting, and the 3rd layer is output layer, utilize error weights and the threshold value of each node between dragover modification level and the layer successively, export simultaneously.
5. permagnetic synchronous motor neural net left inverse state observation method according to claim 3, it is characterized in that, described step 2) parameter and the weights coefficient of quoting steepest decline learning algorithm training adjusting multilayer feedforward neural network in specifically may further comprise the steps:
2-1) with { u
Sd, u
SqBe added to 2 inputs of the compound object of observation of permagnetic synchronous motor system respectively, gather the rotor velocity ω of permagnetic synchronous motor with the predetermined sampling period
rAnd rotor angular position thetar and phase current i
a, i
b, detect and computing module acquisition desired data { i through rate of current
Sd, i
SqAnd preserve;
2-2) with the data-signal { u that preserves
Sd, u
Sq, i
Sd, i
SqRespectively application tracking differentiator off-line extract and ask the single order of electric current, and then signal is done standardization processing, form the training sample set of neural net
2-3) use steepest decline learning algorithm off-line training fuzzy neural network, adjust weights and the threshold value coefficient of each node of network, neural net output mean square error precision is remained in the predetermined mean square error precision.
6. permagnetic synchronous motor neural net left inverse state observation method according to claim 5 is characterized in that the described predetermined sampling period is 5ms.
7. permagnetic synchronous motor neural net left inverse state observation method according to claim 5 is characterized in that described predetermined mean square error precision is 0.0005.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310205193.XA CN103259479B (en) | 2013-05-28 | 2013-05-28 | A kind of permanent magnet synchronous electric machine neural network left inverse state observation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310205193.XA CN103259479B (en) | 2013-05-28 | 2013-05-28 | A kind of permanent magnet synchronous electric machine neural network left inverse state observation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103259479A true CN103259479A (en) | 2013-08-21 |
CN103259479B CN103259479B (en) | 2016-08-10 |
Family
ID=48963225
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310205193.XA Active CN103259479B (en) | 2013-05-28 | 2013-05-28 | A kind of permanent magnet synchronous electric machine neural network left inverse state observation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103259479B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105337547A (en) * | 2014-06-27 | 2016-02-17 | 上海电机学院 | Rotation speed identification method of asynchronous motor speed sensor-free vector control system |
CN105846727A (en) * | 2016-03-18 | 2016-08-10 | 浙江工业大学 | Adjacent coupling type multi-motor speed tracking and synchronous control method based on fuzzy disturbance self-resistance and self-adaptive sliding mode |
CN107547024A (en) * | 2017-10-10 | 2018-01-05 | 江苏大学 | A kind of bearing-free permanent magnet synchronous motor Speedless sensor |
CN107681941A (en) * | 2017-10-10 | 2018-02-09 | 江苏大学 | A kind of building method of bearing-free permanent magnet synchronous motor without radial displacement transducer |
CN108646571A (en) * | 2018-07-12 | 2018-10-12 | 北京航空航天大学 | A kind of gyro frame servo system high precision position discrimination 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 |
CN110855169A (en) * | 2019-12-06 | 2020-02-28 | 龙岩学院 | Single-phase inverter model prediction control method without voltage sensor |
CN111525863A (en) * | 2020-06-02 | 2020-08-11 | 李敬 | Motor speed regulating device and control method thereof |
CN113396369A (en) * | 2019-02-07 | 2021-09-14 | 株式会社电装 | Abnormality detection device |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109617463B (en) * | 2018-12-20 | 2021-04-30 | 东南大学溧阳研究院 | Permanent magnet synchronous motor low-speed section rotor position observer based on BP neural network |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101227160A (en) * | 2007-11-30 | 2008-07-23 | 江苏大学 | Neural network generalized inverse permanent magnetism synchronous machine decoupling controller structure method without bearing |
CN101369132A (en) * | 2008-07-11 | 2009-02-18 | 天津大学 | Permanent magnet spherical motor mechanical decoupling control method based on neural network identifier |
CN101917150A (en) * | 2010-06-24 | 2010-12-15 | 江苏大学 | Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof |
CN102497153A (en) * | 2011-12-12 | 2012-06-13 | 东北大学 | Constant-power-angle self-adaptive control method of permanent magnet synchronous motor |
CN102522945A (en) * | 2012-01-10 | 2012-06-27 | 江苏大学 | Polyphase motor fault-tolerant control method and system based on multi-neural-network inverse model |
-
2013
- 2013-05-28 CN CN201310205193.XA patent/CN103259479B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101227160A (en) * | 2007-11-30 | 2008-07-23 | 江苏大学 | Neural network generalized inverse permanent magnetism synchronous machine decoupling controller structure method without bearing |
CN101369132A (en) * | 2008-07-11 | 2009-02-18 | 天津大学 | Permanent magnet spherical motor mechanical decoupling control method based on neural network identifier |
CN101917150A (en) * | 2010-06-24 | 2010-12-15 | 江苏大学 | Robust controller of permanent magnet synchronous motor based on fuzzy-neural network generalized inverse and construction method thereof |
CN102497153A (en) * | 2011-12-12 | 2012-06-13 | 东北大学 | Constant-power-angle self-adaptive control method of permanent magnet synchronous motor |
CN102522945A (en) * | 2012-01-10 | 2012-06-27 | 江苏大学 | Polyphase motor fault-tolerant control method and system based on multi-neural-network inverse model |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105337547A (en) * | 2014-06-27 | 2016-02-17 | 上海电机学院 | Rotation speed identification method of asynchronous motor speed sensor-free vector control system |
CN105846727B (en) * | 2016-03-18 | 2018-09-18 | 浙江工业大学 | A kind of more motor speed tracking of adjacent coupled type based on fuzzy active disturbance rejection and adaptive sliding mode and synchronisation control means |
CN105846727A (en) * | 2016-03-18 | 2016-08-10 | 浙江工业大学 | Adjacent coupling type multi-motor speed tracking and synchronous control method based on fuzzy disturbance self-resistance and self-adaptive sliding mode |
CN107547024B (en) * | 2017-10-10 | 2020-03-31 | 江苏大学 | No speed sensor of no bearing PMSM |
CN107681941A (en) * | 2017-10-10 | 2018-02-09 | 江苏大学 | A kind of building method of bearing-free permanent magnet synchronous motor without radial displacement transducer |
CN107681941B (en) * | 2017-10-10 | 2020-03-31 | 江苏大学 | Method for constructing radial displacement-free sensor of bearingless permanent magnet synchronous motor |
CN107547024A (en) * | 2017-10-10 | 2018-01-05 | 江苏大学 | A kind of bearing-free permanent magnet synchronous motor Speedless sensor |
CN108646571A (en) * | 2018-07-12 | 2018-10-12 | 北京航空航天大学 | A kind of gyro frame servo system high precision position discrimination method |
CN108646571B (en) * | 2018-07-12 | 2020-10-30 | 北京航空航天大学 | High-precision position identification method for gyro frame servo system |
CN113396369A (en) * | 2019-02-07 | 2021-09-14 | 株式会社电装 | Abnormality detection device |
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 |
CN110855169A (en) * | 2019-12-06 | 2020-02-28 | 龙岩学院 | Single-phase inverter model prediction control method without voltage sensor |
CN111525863A (en) * | 2020-06-02 | 2020-08-11 | 李敬 | Motor speed regulating device and control method thereof |
CN111525863B (en) * | 2020-06-02 | 2021-12-07 | 江苏朗晟激光科技有限公司 | Motor speed regulating device and control method thereof |
Also Published As
Publication number | Publication date |
---|---|
CN103259479B (en) | 2016-08-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103259479A (en) | Method for observing left inverse state of neural network of permanent magnet synchronous motor | |
CN104639003B (en) | A kind of method for identification of rotational inertia of AC servo | |
CN106014877B (en) | Wind generator system multi-fault Diagnosis and faults-tolerant control | |
CN106533303A (en) | Permanent magnet brushless DC motor driver control method | |
CN102208891A (en) | Method for controlling PMSM (permanent magnet synchronous motor) servo system based on friction and disturbance compensation | |
CN102662323B (en) | Adoptive sliding mode control method and adoptive sliding mode control system of wind power generation variable-pitch actuator | |
CN103532459A (en) | Linear servo motor control method for numerically-controlled machine tool driving | |
CN102769424B (en) | Based on the induction-type bearingless motor control system that support vector machine is inverse | |
CN104460518A (en) | Direct-drive XY platform profile control device and method based on fuzzy disturbance compensation | |
CN101630936B (en) | Neural network inverse controller of brushless DC motor and construction method thereof | |
CN102510253B (en) | Sensorless controller of bearingless synchronous reluctance motor and method for controlling sensorless controller | |
CN105186958B (en) | The five mutually fault-tolerant magneto internal model control methods based on Neural Network Inverse System | |
CN102130647B (en) | Speed-sensor-less construction method for detecting rotating speed of bearing-less asynchronous motor | |
CN105932926B (en) | Permanent magnet synchronous motor Speedless sensor based on reduced dimension observer is counter to push away control method | |
CN105915145A (en) | Device and method of controlling permanent magnet linear synchronous motor | |
CN111865167A (en) | PMSLM (permanent magnet synchronous motor) position-sensorless vector control method based on sliding-mode observer | |
CN111835251B (en) | Permanent magnet synchronous motor high-performance control method based on speed-free sensing | |
CN109143869B (en) | Synchronous control system of recursive wavelet neural network compensator of H-shaped platform | |
CN103337997B (en) | A kind of Over Electric Motor with PMSM vector control system and method | |
Barra et al. | Sensorless nonlinear control of wind energy systems with doubly fed induction generator | |
CN111510041B (en) | Running state evaluation method and system of permanent magnet synchronous motor | |
CN205754101U (en) | Permanent-magnet Synchronous-motor Speed Servo System based on modelling | |
Xiang et al. | Cooperative Control of Multiple PM Actuators in PMLSM with Segmented Windings | |
Quang et al. | Observer-based integral sliding mode control for sensorless PMSM drives using FPGA | |
Li et al. | The design of brushless DC motor servo system based on wavelet ANN |
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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211229 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: Zhenjiang City, Jiangsu Province, 212013 Jingkou District Road No. 301 Patentee before: JIANGSU University |