A model predictive controller for bearingless permanent magnet thin-film motor

Abstract

The invention discloses a bearingless permanent magnet thin-film motor model prediction controller in the field of motor control, which belongs to the field of motor control and is composed of a torque prediction control system and a suspension force prediction control system. The torque prediction control system comprises a torque subsystem prediction controller and a PI controller, a cost function calculation module, a cost function online optimization module and a first voltage-type inverter which are connected in series in sequence. The suspension force prediction control system comprises a suspension force subsystem prediction controller and a suspension force voltage reference generator, an SVPWM modulation module and a second voltage-type inverter which are connected in series in sequence. The output ends of the first and second voltage-type inverters are connected to the bearingless permanent magnet thin-film motor. Direct torque control is integrated with model prediction control to obtain better dynamic performance and robustness, and improve motor speed regulation and suspension performance. A control cycle is added as time compensation to predict motor parameters at time k+2, so as to avoid problems caused by delay.

Description

Technical Field

[0001] The invention belongs to the field of motor control, in particular to a multivariable nonlinear bearingless thin-film motor model prediction direct controller suitable for high-performance control of the bearingless thin-film motor. Background Art

[0002] The bearingless permanent magnet thin-film motor is a type of multivariable, nonlinear, and strongly coupled controlled object. It not only inherits some advantages of bearingless permanent magnet synchronous motors such as no friction, no wear, and no need for lubrication, but also the axial length of its permanent magnet rotor is much smaller than the outer diameter of the rotor, and the air gap is small, so that active suspension control of the bearingless permanent magnet thin-film motor can be achieved by controlling only two radial degrees of freedom. According to the principle of minimum magnetic resistance, passive suspension with three degrees of freedom (one axial degree of freedom and two torsional degrees of freedom) can be achieved, which greatly reduces the complexity of control and has very broad application prospects in chemical, pharmaceutical, biomedical engineering, and semiconductor manufacturing.

[0003] Predicting control variables is an important part of realizing model predictive direct control of bearingless thin-film motors. The basic idea of ​​vector control is to generate the same rotating magnetic force as the criterion, and to convert the bearingless permanent magnet thin-film motor model into a DC motor model for control. However, in actual control, the control effect is often poor due to changes in motor parameters, the dynamic response is slow, and the control method is complicated. The direct control method considers the torque winding and suspension force winding of the bearingless permanent magnet thin-film motor as two independent systems, establishes mathematical models for each system, and realizes independent control. This control method cannot maintain good suspension characteristics under load mutations, speed mutations, etc. Model predictive direct control inherits the advantages of fast dynamic response of direct control, and improves control accuracy and robustness, thereby greatly improving the performance of bearingless permanent magnet thin-film motors.

[0004] The document of Chinese patent application number CN201521037225.0 discloses a direct controller for torque and suspension force of a single-winding bearingless motor, which establishes a controller consisting of a reference value calculation part and an observation value calculation part for a single-winding bearingless motor, generates a voltage static coordinate command value in a torque and suspension force voltage reference value generator according to the flux amplitude command value, the real-time winding torque component flux amplitude and phase, and the actual feedback current, and directly controls the torque and rotor radial suspension force. The document of Chinese patent application number CN201610700872 discloses a bearingless permanent magnet synchronous motor predictive controller and construction method, which builds a predictive controller for permanent magnet synchronous motors, does not involve direct torque control, and has poor control performance. Summary of the invention

[0005] 1. Technical Purpose

[0006] The purpose of the present invention is to provide a bearingless permanent magnet thin-film motor model predictive direct controller which can not only make the bearingless permanent magnet thin-film motor have excellent dynamic performance, strong resistance to motor parameter changes and load disturbances, but also effectively improve various control performance indicators of the bearingless permanent magnet thin-film motor, such as dynamic response speed, steady-state tracking accuracy and parameter robustness.

[0007] 2. Technical solution

[0008] The technical solution adopted by the model predictive controller of a bearingless permanent magnet thin-film motor described in the present invention is: it is composed of a torque predictive control system and a suspension force predictive control system, the torque predictive control system includes a torque subsystem predictive controller and a PI controller, a cost function calculation module, a cost function online optimization module and a first voltage-type inverter connected in series in sequence, the suspension force predictive control system includes a suspension force subsystem predictive controller and a suspension force voltage reference generator, an SVPWM modulation module and a second voltage-type inverter connected in series in sequence, and the output ends of the first and second voltage-type inverters are connected to the bearingless permanent magnet thin-film motor;

[0009] The torque winding phase voltage u of the motor 1a ,u 1b , phase current i 1a ,i 1b , rotor angle θ and angular velocity ω are input into the torque subsystem prediction controller, which outputs the torque flux prediction value ψ at time k+2 s 1(k+2), torque prediction value T e (k+2), the torque winding flux phase angle λ1(k+2) and the torque winding current i in the α-β coordinate system at time k+1 1α (k+1),i 1β (k+1);

[0010] The output end of the torque subsystem prediction controller is connected to the cost function calculation module and the suspension force subsystem prediction controller respectively, and the torque flux prediction value ψ s 1(k+2) and torque prediction value T e (k+2) as the first and second inputs of the cost function calculation module, the torque command value T e * and torque flux command value ψ s1 * As the third and fourth inputs of the cost function calculation module, the cost function calculation module outputs 8 cost function values ​​m, which are evaluated by the cost function optimization module to obtain the optimal switch state and input into the first voltage source inverter;

[0011] The torque flux prediction value ψ s 1(k+2), torque winding flux phase angle λ1(k+2), torque winding current i 1α (k+1),i 1β (k+1) is used as the first to fourth inputs of the suspension force subsystem predictive controller, and the suspension winding phase voltage u 2a ,u 2b and phase current i 2a ,i 2b As the fifth to eighth inputs of the predictive controller of the suspension force subsystem,

[0012] The predicted suspension force F in the α-β coordinate system at the output of the suspension force subsystem prediction controller at time k+2 is α (k+2),F β The synthetic air gap flux ψ at time (k+2) and k+2 m (k+2) and its phase angle μ(k+2), as well as the torque winding current i in the α-β coordinate system at time k+1 1α (k+1),i 1β (k+1);

[0013] The synthetic air gap flux ψ m (k+2) and its phase angle μ(k+2) and the torque winding current i 1α (k+1),i 1β (k+1) These four parameters serve as the first to fourth inputs of the suspension force voltage reference generator.

[0014] Motor suspension force command value F α * and the predicted suspension force F α (k+2) is the difference in suspension force comparison value △F α , suspension force command value F β * and the predicted suspension force F β (k+2) Subtract and get the suspension force comparison value △F β , Suspension force comparison value △F α ,△F β As the fifth and sixth inputs of the suspension force voltage reference generator 44, the suspension force voltage reference generator 44 obtains the suspension winding phase voltage command value u 2α * ,u 2β * And input to SVPWM modulation module 45.

[0015] The advantages of the present invention are:

[0016] 1. The present invention combines direct torque control (DTC) with model predictive control (MPC), which not only has the advantages of simple structure of direct torque control and low dependence on motor parameters, but also has the characteristics of fast torque response and high control accuracy of model predictive control, thereby obtaining better dynamic performance and robustness, and improving motor speed regulation and suspension performance.

[0017] 2. The present invention delays the phase voltage and phase current by one moment, and uses a predictive controller to predict the motor parameters at moment k+2, thus avoiding the problems caused by the delay. Its feedback correction has a strong ability to resist disturbances and overcome system uncertainties. In the rolling optimization, each moment has a local optimization performance index relative to that moment, and the optimization result is used to determine the future control effect, so that the uncertainty caused by model mismatch, distortion, disturbance, etc. can be compensated in time, thereby obtaining better dynamic control performance.

[0018] 3. The present invention adds a control cycle as time compensation in the prediction model, avoiding the large amount of calculation time of the prediction algorithm, delays such as sensor communication and model circuit filtering, and the problem that the output of the controller always lags behind the control input, resulting in a decrease in the stability of the motor.

[0019] 4. The model predictive controller of the present invention adopts finite set control set model predictive control, which considers the target optimization and switch state decision process as a whole. Compared with continuous set model predictive control, it has the advantages of wide application range and simple application. The prediction algorithm itself adopts multi-step prediction, rolling optimization and feedback control strategies, with good control performance, not overly dependent on model accuracy, and good robustness.

[0020] 5. The present invention can improve the control performance of bearingless permanent magnet thin-film motors, and is suitable for other bearingless motor control systems, as well as various types of motor control systems supported by magnetic bearings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a block diagram of the model predictive controller for the bearingless permanent magnet thin-film motor of the present invention:

[0022] Figure 2 yes Figure 1 The structural block diagram of the torque subsystem prediction controller 11;

[0023] Figure 3 yes Figure 1 The structural block diagram of the mid-suspended force subsystem predictive controller 31;

[0024] In the figure: 01. Speed ​​sensor; 02, 03. First displacement sensor; 11. Torque subsystem prediction controller; 12. First 3s / 2s conversion module; 13. First delay compensator; 14. Torque winding flux observer; 15. Torque winding flux prediction module; 16. Second delay compensator; 17. Current prediction module; 18. Torque prediction module; 21. Torque prediction control system; 22. First current sensor; 23. Cost function calculation module; 24. Cost function online optimization module; 25. First voltage source inverter; 26. PI controller; 28. Speed ​​circuit interface module; 31. Suspension force subsystem prediction controller; 32. Second 3s / 2s conversion module; 33. Third delay compensator; 34. Suspension winding flux observer; 35. Suspension winding flux prediction module; 36. Torque winding synthetic air gap flux prediction module; 37. Suspension force prediction module; 42. Second current sensor; 44. Suspension force voltage reference generator; 45. SVPWM modulation module; 46. Second voltage type inverter; 47. First PID controller; 48. Second PID controller; 49. Displacement circuit interface module. DETAILED DESCRIPTION

[0025] See also Figure 1 The bearingless permanent magnet thin-film motor model predictive controller 51 of the present invention is connected to the bearingless permanent magnet thin-film motor (BPMSM). The bearingless permanent magnet thin-film motor model predictive controller 51 is composed of a torque predictive control system 21 and a suspension force predictive control system 41. Among them, the torque predictive control system 2 includes a torque subsystem predictive controller 11 and a PI controller 26, a cost function calculation module 23, a cost function online optimization module 24 and a first voltage-type inverter 25 connected in series in sequence; the suspension force predictive control system 41 includes a suspension force subsystem predictive controller 31 and a suspension force voltage reference generator 44, a SVPWM modulation module 45, and a second voltage-type inverter 46 connected in series in sequence.

[0026] The first voltage source inverter 25 outputs the torque winding phase voltage u in the orthogonal coordinate system 1a ,u 1b ,u 1c The output end of the first voltage source inverter 25 is directly connected to the input end of the bearingless permanent magnet thin-film motor, and the torque winding phase voltage u 1a ,u 1b ,u 1c Enter the bearingless permanent magnet thin-film motor.

[0027] The output end of the first voltage source inverter 25 is also connected to the torque subsystem prediction controller 11, and the torque winding phase voltage u 1a ,u 1bThe torque winding phase current i is input into the torque subsystem prediction controller 11 as the first and second inputs of the torque subsystem prediction controller 11. The torque winding phase current i is collected from the output end of the voltage source inverter 25 by the current sensor 22. 1a ,i 1b The first current sensor 22 converts the phase current i 1a ,i 1b Input to the torque subsystem prediction controller 11, the phase current i 1a ,i 1b As the third and fourth inputs of the torque subsystem prediction controller 11. The speed sensor 01 is used to detect the rotor angle θ of the bearingless permanent magnet thin-film motor. The rotor angle θ is converted into the motor angular velocity ω by the speed circuit interface module 28. The rotor angle θ and the motor angular velocity ω are input into the torque subsystem prediction controller 11 as the fifth and sixth inputs of the torque subsystem prediction controller 11. The output of the torque subsystem prediction controller 11 is the torque flux prediction value ψ at time k+2 s 1(k+2), Torque prediction value T at time k+2 e The torque winding flux phase angle λ1(k+2) at time (k+2) and k+2, and the torque winding current i in the α-β coordinate system at time k+1 1α (k+1), i 1β (k+1) these five parameters.

[0028] The output end of the torque subsystem prediction controller 11 is connected to the cost function calculation module 23 and the suspension force subsystem prediction controller 31. The torque flux prediction value ψ output by the torque subsystem prediction controller 11 is s 1(k+2) and torque prediction value T e The two parameters (k+2) are input to the cost function calculation module 23 as the first and second inputs of the cost function calculation module 23. The torque flux prediction value ψ output by the torque subsystem prediction controller 11 s 1(k+2), torque winding flux phase angle λ1(k+2), torque winding current i 1α (k+1), i 1β The four parameters (k+1) are commonly input into the suspension force subsystem predictive controller 31 as the first to fourth inputs of the suspension force subsystem predictive controller 31 .

[0029] The motor angular velocity ω and angular velocity command value ω output by the speed circuit interface module 28 *作 The angular velocity difference △ω is input to the PI controller 26, and the PI controller 26 outputs the torque command value T e * , torque command value T e* Input into the cost function calculation module 23 as the third input of the cost function calculation module 23, and the torque flux command value ψ s1 * As the fourth input of the cost function calculation module 23, the cost function calculation module 23 outputs 8 cost function values ​​m, m=m0~m7, and the 8 cost function values ​​m0~m7 are input into the cost function optimization module 24. The cost function optimization module 24 evaluates the switch state with the best control effect, and generates the corresponding inverter switch state S by selecting the minimum value among the 8 function values ​​m0~m7. 1a , S 1b , S 1c The first voltage source inverter 25 is input to control the torque of the bearingless permanent magnet thin-film motor.

[0030] The second current sensor 22 is used to detect the suspension force phase current i of the bearingless permanent magnet thin-film motor. 2a ,i 2b , suspension force phase current i 2a ,i 2b Input to the suspension force subsystem prediction controller 31 as the fifth and sixth inputs of the suspension force subsystem prediction controller 31, the suspension winding phase voltage u 2a ,u 2b The seventh and eighth inputs of the suspension force subsystem prediction controller 31. The output of the suspension force subsystem prediction controller 31 is the predicted suspension force F at time k+2 in the α-β coordinate system. α (k+2),F β (k+2), the synthetic air gap flux ψ at time k+2 m (k+2) and its phase angle μ(k+2), as well as the torque winding current i in the α-β coordinate system at time k+1 1α (k+1),i 1β The suspension force subsystem prediction controller 31 outputs the synthetic air gap flux ψ at time k+2. m (k+2) and its phase angle μ(k+2) and the torque winding current i in the α-β coordinate system at time k+1 1α (k+1),i 1β The four parameters (k+1) are commonly input into the levitation force voltage reference generator 44 as the first to fourth inputs of the levitation force voltage reference generator 44 .

[0031] The x-direction displacement sensor 02 and the y-direction displacement sensor 03 are used to detect the radial displacement of the rotor of the bearingless permanent magnet thin-film motor model, and the real-time radial displacement x and y of the rotor on the x-axis and y-axis are output through the displacement circuit interface module 49. The real-time radial displacement values ​​x and y are compared with the corresponding rotor displacement command value x.* ,y * Compare them respectively to obtain displacement difference Δx and Δy, Δx is input into the corresponding first PID control 47, and the first PID control 47 obtains the suspension force command value F α * , Δy is input into the corresponding second PID control 48, and the second PID control 48 obtains the suspension force command value F β * .

[0032] Suspension force command value F α * and the predicted suspension force F at time k+2 output by the suspension force subsystem prediction controller 31 α (k+2) is subtracted to get the comparative value of suspension force △F α Suspension force command value F β * and the predicted suspension force F at time k+2 output by the suspension force subsystem prediction controller 31 β (k+2) is subtracted to get the comparative value of suspension force △F β The suspension force comparison value △F α ,△F β The levitation force voltage reference generator 44 is used as the fifth and sixth inputs of the levitation force voltage reference generator 44. The levitation force voltage reference generator 44 obtains the levitation winding phase voltage command value u 2α * ,u 2β * , and the suspension winding phase voltage command value u 2α * ,u 2β * The SVPWM modulation module 45 outputs a switch state signal S by combining the basic voltage vectors in one switching cycle so that the average value is equal to the voltage vector command value. 2a , S 2b , S 2c , drives the second voltage source inverter 46, and the second voltage source inverter 46 outputs the suspension winding phase voltage u 2a ,u 2b ,u 2c , the suspension force predictive control is realized for the bearingless permanent magnet thin-film motor model.

[0033] See also Figure 2 The torque subsystem prediction controller 11 shown in the figure has six inputs: torque winding phase voltage u 1a ,u 1b , phase current i 1a ,i 1b, rotor angle θ and motor angular velocity ω, the torque subsystem prediction controller 11 includes a first 3s / 2s transformation module 12, a first delay compensator 13, a torque winding flux observer 14 and a torque winding flux prediction module 15 which are connected in series in sequence, and also includes a second delay compensator 16, a current prediction module 17 and a torque prediction module 18 which are connected in series in sequence.

[0034] Torque winding phase voltage u 1a ,u 1b , phase current i 1a ,i 1b The first 3s / 2s transformation module 12 transforms the torque winding phase voltage and phase current u in the orthogonal coordinate system into the first 3s / 2s transformation module 12 and the second delay compensator 16. 1a ,u 1b ,i 1a ,i 1b After transformation, the torque winding phase voltage u at time k in the α-β coordinate system is obtained 1α (k),u 1β (k) and phase current i 1α (k),i 1β (k), in order to compensate for the delays of prediction algorithm calculation, sensor communication and model circuit filtering, a first delay compensator 13 is added before the input prediction module. The torque winding phase voltage u at time k 1α (k),u 1β (k) and phase current i 1α (k),i 1β (k) is input to the first delay compensator 13 to obtain the torque winding phase voltage u at time k+1 1α (k+1),u 1β (k+1) and phase current i 1α (k+1),i 1β (k+1), the torque winding phase current i 1α (k+1),i 1β (k+1) is directly input into the suspension force voltage reference generator 44, while the torque winding phase voltage u 1α (k+1),u 1β (k+1) and phase current i 1α (k+1),i 1β (k+1) are input into the torque winding flux observer 14, which estimates the torque winding stator flux and its phase angle, and obtains the torque winding stator flux ψ at the delayed k+1 time according to formula (1): s1α (k+1),ψ s1β (k+1):

[0035]

[0036] In formula (1), t is the running time, R s is the stator resistance of the torque winding.

[0037] Torque winding stator flux ψ s1α (k+1),ψ s1β (k+1), torque winding phase voltage u 1α (k+1),u 1β (k+1) and the torque winding phase current i output by the current prediction module 17 1α (k+2),i 1β (k+2) are input together into the torque winding flux prediction module 15, and the torque winding flux prediction module 15 obtains the predicted torque winding stator flux ψ at time k+2 according to formula (2): s1 (k+2) and phase angle λ1(k+2), torque winding stator flux ψ s1 (k+2) and phase angle λ1(k+2) as the output of the torque subsystem predictive controller 11:

[0038]

[0039] Where Ts is a sampling period.

[0040] At the same time, the phase voltage and phase current u 1a ,u 1b ,i 1a ,i 1b After the second delay compensator 16, the torque winding phase voltage u predicted at time k+1 is obtained. 1a (k+1),u 1b (k+1) and phase current i 1α (k+1),i 1β (k+1). The torque winding phase voltage u 1a (k+1),u 1b (k+1), phase current i 1α (k+1),i 1β (k+1) and the rotor angle θ and the motor angular velocity ω are used as the input of the current prediction module 17. The current prediction module 17 first performs a Clark coordinate transformation on the phase current and phase voltage in the three-phase stationary coordinate system, and then performs a Park coordinate transformation on the phase current and phase voltage in the two-phase stationary coordinate system, as shown in the following formula (3):

[0041]

[0042] In formula (3), θ is the rotor angle, i 1a (k+1), i 1b (k+1),u 1a (k+1),u1b (k+1) are the torque winding phase current and phase voltage of the three-phase stationary coordinate system at time k+1, i 1d (k+1), i 1q (k+1),u 1d (k+1),u 1q (k+1) are the phase current and phase voltage of the torque winding in the dq axis coordinate system at time k+1 respectively.

[0043] Then, according to formula (4), the torque winding phase current i of the dq axis at time k+2 is obtained: 1d (k+2), i 1q The torque winding phase current i predicted in the α-β two-phase stationary coordinate system at time (k+2) and k+2 1α (k+2),i 1β (k+2):

[0044]

[0045] L d , L q is the winding inductance in the dq axis coordinate system, ψ f is the permanent magnet flux, and ω is the motor angular velocity.

[0046] The torque winding phase current i 1α (k+2),i 1β (k+2) and the torque winding stator flux ψ s1 (k+2) and the phase angle λ1(k+2) are used as the input of the torque prediction module 18. The torque prediction module 18 calculates the predicted torque T at time k+2 according to formula (5): e (k+2), as the output of the torque subsystem prediction controller 11:

[0047]

[0048] Where n p is the number of torque winding pole pairs.

[0049] Therefore, the torque winding subsystem prediction controller 11 provides the necessary flux prediction and torque prediction information.

[0050] See also Figure 3 The suspension force subsystem prediction controller 31 shown includes a second 3s / 2s transformation module 32, a third delay compensator 33, a suspension winding flux observer 34 and a suspension winding flux prediction module 35 which are connected in series in sequence, and also includes a torque winding composite air gap flux prediction module 36 and a suspension force prediction module 37 which are connected in series in sequence.

[0051] The phase voltage u of the suspension winding in the orthogonal coordinate system 2a,u 2b , phase current i 2a ,i 2b Input 3s / 2s conversion module 32, after conversion, output phase voltage u at time k 2α (k),u 2β (k) and phase current i 2α (k),i 2β (k), phase voltage u 2α (k),u 2β (k) and phase current i 2α (k),i 2β (k) The suspension winding phase voltage u at time k+1 is obtained through the third delay compensator 33. 2α (k+1),u 2β (k) and phase current i 2α (k),i 2β (k). Suspension winding phase voltage u 2α (k+1),u 2β (k) and phase current i 2α (k),i 2β The signal (k) is input into the suspension winding flux observer 34, and the suspension winding flux observer 34 obtains the suspension winding flux ψ at time k+1 according to formula (6): s2 (k+1) and phase angle λ2(k+1):

[0052]

[0053] In the formula, i 2α (k+1), i 2β (k+1),u 2α (k+1),u 2β (k+1) is the phase current and phase voltage of the suspension winding in the α-β axis coordinate system at time k+1, R s2 is the stator resistance of the suspension winding, ψ s2α (k+1),ψ s2β (k+1) is the magnetic flux amplitude of the suspension winding in the α-β axis coordinate system at time k+1, ψ s2 (k+1) and λ2(k+1) are the suspension winding flux amplitude and suspension winding flux angle at time k+1 respectively.

[0054] Suspension force winding flux ψ s2 The two signals (k+1) and the phase angle λ2(k+1) are input to the suspension winding flux prediction module 35. The suspension winding flux prediction module 35 obtains the suspension winding flux ψ at time k+2 according to formula (7): s2 (k+2) and phase angle λ2(k+2):

[0055]

[0056] ψ s2α (k+2),ψ s2β (k+2) is the magnetic flux amplitude of the suspension winding in the α-β axis coordinate system at time k+2.

[0057] Suspension force winding flux ψ s2 (k+2) and phase angle λ2(k+2) are input to the suspension force prediction module 37 as the first and second inputs of the suspension force prediction module 37. At the same time, the torque winding flux ψ at time k+2 obtained from the torque prediction control s1 (k+2) and phase angle λ1(k+2) and torque winding phase current i at time k+1 1α (k+1),i 1β (k+1), and the torque winding combined air gap flux prediction module 36 is input together. The torque winding combined air gap flux prediction module 36 calculates and outputs the torque winding combined air gap flux ψ at time k+2 according to formula (8) m (k+2) and phase angle μ(k+2);

[0058]

[0059] Among them, L 1l is the leakage inductance, ψ mα (k+2),ψ mβ (k+2),ψ m (k+2) and μ(k+2) are the torque winding air gap flux amplitude, synthetic air gap flux amplitude and phase of the α-β axis at time k+2 respectively.

[0060] Torque winding combined air gap flux ψ m (k+2) and phase angle μ(k+2) are input into the suspension force prediction module 37 as the third and fourth inputs of the suspension force prediction module 37. At the same time, the torque winding forms the air gap flux ψ m (k+2) and the phase angle μ(k+2) are directly used as the two outputs of the suspension force subsystem predictive controller 31.

[0061] Suspension force prediction module 37 pairs of suspension force winding flux ψ s2 (k+2) and phase angle λ2(k+2), torque windings form air gap flux ψ m (k+2) and phase angle μ(k+2), and the predicted suspension force F at time k+2 is calculated according to formula (9): α (k+2),F β (k+2):

[0062]

[0063] In the formula, π is the circumference of a circle, P1 and P2 are the equivalent pole pairs of the torque winding and the suspension winding respectively; l is the effective length of the motor rotor core; r is the rotor radius; W1 and W2 are the effective turns of each phase of the torque winding and the suspension winding in series respectively; μ0 is the air magnetic permeability; ψ m (k+2) and μ2(k+2) are the torque windings and air gap magnetic flux density at time k+2. Chain Amplitude and Phase ; F α (k+2), F β (k+2) is the predicted levitation force at time k+2.

[0064] The suspension force subsystem prediction controller 31 provides the necessary suspension force and synthetic air gap flux prediction information for the bearingless permanent magnet thin-film motor model.

[0065] See also Figure 1 , in one sampling period, the torque flux prediction value ψ corresponding to the 8 voltage vectors s 1(k+2), torque prediction value T e (k+2) and torque command value T e * , torque flux command value ψ s1 * The input is sent to the cost function calculation module 23, and the cost function calculation module 23 calculates eight cost function values ​​m=m0-m7 according to the following formula (10):

[0066]

[0067] The part before the square root symbol is the square of the relative error of the predicted torque, and the part after it is the square of the relative error of the stator flux. Using the relative error method, the cost function is converted into a dimensionless quantity. The smaller the cost function value m, the better the control strategy.

[0068] The eight cost function values ​​m=m0-m7 are input into the cost function online optimization module 24, and the cost function online optimization module 24 selects the minimum function value and generates the corresponding inverter switch state S 1a , S 1b , S 1c , input into the flat voltage inverter 25 to complete the torque control of the bearingless permanent magnet thin-film motor.

[0069] The suspension force voltage reference generator 44 calculates according to formula (11):

[0070]

[0071] and is the voltage command value for controlling the suspension force under two-phase stationary coordinates, ΔF α is the suspension force command value and the predicted value F α The comparison difference of (k+2); ΔFβ is the suspension force command value and the predicted value F β The comparison difference of (k+2).

[0072] In this way, dual control of the torque winding and the suspension force winding of the bearingless permanent magnet thin-film motor is achieved.

Claims

1. A bearingless permanent magnet thin-film motor model predictive controller, comprising a torque predictive control system (21) and a suspension force predictive control system (41), characterized in that: The torque prediction control system (21) comprises a torque subsystem prediction controller (11) and a PI controller (26), a cost function calculation module (23), a cost function online optimization module (24) and a first voltage-type inverter (25) connected in series in sequence; the suspension force prediction control system (41) comprises a suspension force subsystem prediction controller (31) and a suspension force voltage reference generator (44), an SVPWM modulation module (45) and a second voltage-type inverter (46) connected in series in sequence; the output ends of the first and second voltage-type inverters (25, 46) are connected to a bearingless permanent magnet thin-film motor; The torque winding phase voltage u of the motor 1a ,u 1b , phase current i 1a ,i 1b , rotor angle θ and angular velocity ω are input into a torque subsystem prediction controller (11), and the torque subsystem prediction controller (11) outputs a torque flux prediction value ψ at time k+2 s1 (k+2), torque prediction value T e (k+2), the torque winding flux phase angle λ1(k+2) and the torque winding current i in the α-β coordinate system at time k+1 1α (k+1),i 1β (k+1); The output end of the torque subsystem prediction controller (11) is respectively connected to the cost function calculation module (23) and the suspension force subsystem prediction controller (31), and the torque flux linkage prediction value ψ s1 (k+2) and torque prediction value T e (k+2) as the first and second inputs of the cost function calculation module (23), the torque command value T e * and torque flux command value ψ s1 * As the third and fourth inputs of the cost function calculation module (23), the cost function calculation module (23) outputs 8 cost function values ​​m, which are evaluated by the cost function online optimization module (24) to obtain the optimal switch state and input it into the first voltage source inverter (25); The torque flux prediction value ψ s1 (k+2), torque winding flux phase angle λ1(k+2), torque winding current i 1α (k+1),i 1β (k+1) as the first to fourth inputs of the suspension force subsystem prediction controller (31), the suspension winding phase voltage u 2a ,u 2b and phase current i 2a ,i 2b As the fifth to eighth inputs of the suspension force subsystem predictive controller (31), The suspension force subsystem prediction controller (31) outputs the predicted suspension force F in the α-β coordinate system at time k+2. α (k+2),F β The synthetic air gap flux ψ at time (k+2) and k+2 m (k+2) and its phase angle μ(k+2), as well as the torque winding current i in the α-β coordinate system at time k+1 1α (k+1),i 1β (k+1); The synthetic air gap flux ψ m (k+2) and its phase angle μ(k+2) and the torque winding current i 1α (k+1),i 1β (k+1) These four parameters serve as the first to fourth inputs of the suspension force voltage reference generator (44); Motor suspension force command value F α * and the predicted suspension force F α (k+2) is the difference in suspension force comparison value △F α , suspension force command value F β * and the predicted suspension force F β (k+2) Subtract and get the suspension force comparison value △F β , Suspension force comparison value △F α ,△F β As the fifth and sixth inputs of the suspension force voltage reference generator (44), the suspension force voltage reference generator (44) obtains the suspension winding phase voltage command value u 2α * ,u 2β * And input to SVPWM modulation module (45); The torque subsystem prediction controller (11) comprises a first 3s / 2s conversion module (12), a first delay compensator (13), a torque winding flux observer (14) and a torque winding flux prediction module (15) which are connected in series in sequence, and further comprises a second delay compensator (16), a current prediction module (17) and a torque prediction module (18) which are connected in series in sequence; the torque winding phase voltage and phase current u 1a ,u 1b ,i 1a ,i 1b The torque winding phase voltage u at time k in the α-β coordinate system is obtained by transformation through the first 3s / 2s transformation module (12). 1α (k),u 1β (k) and phase current i 1α (k),i 1β (k), and then the torque winding phase voltage u at time k+1 is obtained through the first delay compensator (13) 1α (k+1),u 1β (k+1) and phase current i 1α (k+1),i 1β (k+1), torque winding phase current i 1α (k+1),i 1β (k+1) is directly input into the suspension force voltage reference generator (44), and the torque winding phase voltage u 1α (k+1),u 1β (k+1) and phase current i 1α (k+1),i 1β (k+1) are input to the torque winding flux observer (14), and the torque winding flux observer (14) obtains the torque winding stator flux ψ at time k+1 s1α (k+1),ψ s1β (k+1); torque winding stator flux ψ s1α (k+1),ψ s1β (k+1), torque winding phase voltage u 1α (k+1),u 1β (k+1) and the torque winding phase current i output by the current prediction module (17) 1α (k+2),i 1β (k+2) are input into the torque winding flux prediction module (15), and the torque winding flux prediction module (15) obtains the torque winding stator flux ψ at the time k+2 s1 (k+2) and phase angle λ1(k+2); torque winding phase voltage, phase current u 1a ,u 1b ,i 1a ,i 1b The torque winding phase voltage u predicted at time k+1 is obtained through the second delay compensator (16). 1a (k+1),u 1b (k+1) and phase current i 1α (k+1),i 1β (k+1), torque winding phase voltage u 1a (k+1),u 1b (k+1), phase current i 1α (k+1),i 1β (k+1) and the rotor angle θ and the motor angular velocity ω are used as inputs of a current prediction module (17), and the current prediction module (17) outputs the torque winding phase current i predicted in the α-β two-phase stationary coordinate system 1α (k+2),i 1β (k+2), the torque winding phase current i 1α (k+2),i 1β (k+2) and the torque winding stator flux ψ s1 (k+2) and the phase angle λ1(k+2) are used as the input of the torque prediction module (18), and the torque prediction module (18) obtains the predicted torque T at the time k+2. e (k+2); The suspension force subsystem prediction controller (31) comprises a second 3s / 2s conversion module (32), a third delay compensator (33), a suspension winding flux observer (34) and a suspension winding flux prediction module (35) which are connected in series in sequence, and also comprises a torque winding composite air gap flux prediction module (36) and a suspension force prediction module (37) which are connected in series in sequence; the suspension force winding phase voltage u 2a ,u 2b , phase current i 2a ,i 2b After transformation by the second 3s / 2s transformation module (32), the phase voltage u at time k is obtained. 2α (k),u 2β (k) and phase current i 2α (k),i 2β (k), and then the suspension winding phase voltage u at time k+1 is obtained through the third delay compensator (33) 2α (k+1),u 2β (k) and phase current i 2α (k),i 2β (k), the suspension winding flux linkage ψ at time k+1 is obtained through the suspension winding flux linkage observer (34) s2 (k+1) and phase angle λ2(k+1), and then the suspension winding flux prediction module (35) is used to obtain the suspension winding flux ψ at time k+2 s2 (k+2) and phase angle λ2(k+2), suspension winding flux ψ s2 (k+2) and phase angle λ2(k+2) as the first and second inputs of the suspension force prediction module (37); the torque winding flux ψ s1 (k+2) and phase angle λ1(k+2) and torque winding phase current i 1α (k+1),i 1β (k+1) are input into a torque winding composite air gap flux prediction module (36), and the torque winding composite air gap flux prediction module (36) calculates the torque winding composite air gap flux ψ m (k+2) and phase angle μ(k+2), and serve as the third and fourth inputs of the suspension force prediction module (37). The suspension force prediction module (37) obtains the predicted suspension force F at time k+2 α (k+2),F β (k+2).

2. A bearingless permanent magnet thin-film motor model predictive controller according to claim 1, characterized in that: The torque winding flux prediction module (15) is based on the formula Get the predicted torque winding stator flux ψ at time k+2 s1 (k+2) and phase angle λ1(k+2), Ts is a sampling period.

3. A bearingless permanent magnet thin-film motor model predictive controller according to claim 2, characterized in that: The torque prediction module (18) is based on the formula Calculate the predicted torque T at time k+2 e (k+2), n p is the number of torque winding pole pairs.

4. A bearingless permanent magnet thin-film motor model predictive controller according to claim 1, characterized in that: The suspension winding flux observer (34) is based on the formula Get the suspension winding flux ψ at time k+1 s2 (k+1) and phase angle λ2(k+1), R s2 is the stator resistance of the suspension winding.

5. A bearingless permanent magnet thin-film motor model predictive controller according to claim 4, characterized in that: The suspension winding flux prediction module (35) is based on the formula Get the suspension winding flux ψ at time k+2 s2 (k+2) and phase angle λ2(k+2), ψ s2α (k+2),ψ s2β (k+2) is the magnetic flux amplitude of the suspension winding in the α-β axis coordinate system at time k+2.

6. A bearingless permanent magnet thin-film motor model predictive controller according to claim 5, characterized in that: The torque winding composite air gap flux prediction module (36) calculates and outputs the torque winding composite air gap flux ψ at time k+2 according to the formula m (k+2) and phase angle μ(k+2); L 1l is the leakage inductance, ψ mα (k+2),ψ mβ (k+2),ψ m (k+2) and μ(k+2) are the torque winding air gap flux amplitude, synthetic air gap flux amplitude and phase of the α-β axis at time k+2 respectively.

7. A bearingless permanent magnet thin-film motor model predictive controller according to claim 6, characterized in that: The suspension force prediction module (37) is based on the formula Calculate the predicted suspension force F at time k+2 α (k+2),F β (k+2), π is the circumference of a circle, P1 and P2 are the equivalent pole pairs of the torque winding and the suspension winding respectively; l is the effective length of the motor rotor core; r is the rotor radius; W1 and W2 are the effective turns of each phase of the torque winding and the suspension winding in series respectively; μ0 is the air magnetic permeability; ψ m (k+2) and μ2(k+2) are the torque winding composite air gap flux amplitude and phase at time k+2 respectively; F α (k+2), F β (k+2) is the predicted levitation force at time k+2.

8. The bearingless permanent magnet thin-film motor model predictive controller according to claim 1 is characterized by: The suspension force voltage reference generator (44) is based on the formula: Calculate the voltage command value of the suspension force

Images