Full speed range sensorless construction method for bearingless permanent magnet synchronous machines

Abstract

The application discloses a full-speed-range speed sensorless construction method of a bearingless permanent magnet synchronous motor, a high-frequency signal injection method module and an AGA-EKF detection method module are connected in parallel and then jointly connected in series at the front end of a rotating speed switching algorithm module to form a composite method rotating speed detection module, the composite method rotating speed detection module is connected in series into a control system of the bearingless permanent magnet synchronous motor to inject a high-frequency voltage signal; the rotating speed of the motor is detected by using the pulse array high-frequency signal injection method under zero speed and low speed conditions, and the rotating speed information is detected by using an extended Kalman filter method optimized by using a self-adaptive genetic algorithm under medium and high speed conditions; the pulse array high-frequency signal injection method and the rotating speed and detection composite method based on the extended Kalman filter optimized by using the genetic algorithm are adopted to complete the speed detection of the motor running in the full-speed range of zero speed, low speed and high speed, and realize the smooth switching of the motor from low speed to high speed and the stable suspension running of the motor in the full-speed range.

Description

Technical Field

[0001] This invention belongs to the field of control of bearingless permanent magnet synchronous motors, specifically a method for constructing a bearingless permanent magnet synchronous motor without a speed sensor. Background Technology

[0002] Bearingless permanent magnet synchronous motors are a new type of motor that combines permanent magnet synchronous motor technology and magnetic bearing technology. Its stator slots contain torque windings and levitation windings with a difference of one pole pair. Stable levitation operation is achieved by controlling the current in the torque windings and the levitation windings. Bearingless permanent magnet synchronous motors not only possess the high power density and efficiency of traditional permanent magnet synchronous motors, but also have the advantages of magnetic bearings, such as no mechanical wear, no lubrication required, low noise, and long service life. They have broad application prospects in aerospace, biomedicine, and flywheel energy storage.

[0003] Accurately observing the displacement and angular position of the rotor in a bearingless permanent magnet synchronous motor (PMSM) is fundamental for achieving stable levitation operation of the rotor. Typically, rotor speed is observed using mechanical sensors such as eddy current sensors and photoelectric encoders. However, the use of mechanical sensors has several drawbacks. For example, the measurement accuracy of mechanical sensors is significantly affected by environmental factors. Furthermore, the use of mechanical sensors increases the complexity and cost of the bearingless PMSM system and hinders its development towards miniaturization, high speed, and high precision. Therefore, employing sensorless speed technology is beneficial for improving the reliability and stability of the bearingless PMSM system while reducing its cost.

[0004] Chinese Patent Publication No. CN 106130426A, entitled "A Sensorless Ultra-High-Speed ​​Permanent Magnet Synchronous Motor Speed ​​Control Method Based on EKF (Extended Kalman Array)," employs the EKF method to control the speed of a permanent magnet synchronous motor. However, the values ​​of the covariance matrices Q and R of the system noise and measurement noise are determined through experience and simulation experiments, which introduces certain errors and fails to achieve optimal performance. Furthermore, this method is only suitable for speed control under medium-to-high speed conditions. Chinese Patent Publication No. CN 106020803A, entitled "A Sensorless Speed ​​Control Method Based on Sliding Mode Observer," uses a sliding mode observer to control the speed of a permanent magnet synchronous motor. However, this method heavily relies on the mathematical model of the motor's fundamental excitation and cannot detect speed at zero or low speeds. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned deficiencies in the sensorless control of existing bearingless permanent magnet synchronous motors. It proposes a sensorless construction method covering the entire speed range. This sensorless system can accurately and rapidly detect rotor position and speed information across the entire speed range, enabling stable levitation operation of the motor without a speed sensor and improving the reliability and stability of the bearingless permanent magnet synchronous motor control system.

[0006] To achieve the above objectives, the technical solution adopted by the present invention includes:

[0007] Step 1): Construct a high-frequency signal injection module

[0008] The current i in the two-phase stationary coordinate system α-β α i β The current i in the rotating coordinate system dq is obtained by inverse Park transformation. d i q Current i d i q After passing through a BPF filter, a high-frequency current containing rotational speed information is obtained. Then use high-frequency current and modulated current signal sinω h After multiplying by t, the error function f(Δθ) is obtained through LPF. The error function f(Δθ) is then adjusted to 0 by the PI module to obtain the angle estimate under zero low-speed conditions. And the angle estimate The speed estimate is obtained by differentiation.

[0009] Step 2): Construct the AGA-EKF detection module

[0010] Choose state variable x = [i α i β [ω2, θ2] T and output variable y = [i α i β ] T Take the covariance matrix Q = Diag(Q 11 Q 22 Q 33 Q 44 ) and R = Diag(R 11 R 22 The optimal values ​​of covariance matrices Q and R are obtained by selecting the covariance matrices Q and R using the AGA method; Q 11 Q 22 Q 33 Q 44 i α i βErrors in the variation of ω2 and θ2, R 11 R 22 i α i β The variation error; ω2 and θ2 are the actual rotational speed and actual angle of the rotor under medium and high speed conditions, respectively; T is the matrix transpose.

[0011] Based on the optimal Q and R values, the rotor speed estimate under medium- and high-speed operating conditions is obtained using the extended Kalman algorithm. and angle estimates

[0012] Step 3): Construct the speed switching algorithm module

[0013] Determine the speed switching weighting coefficient, and calculate the rotor speed detection value based on the weighting coefficient. and angle detection value

[0014] Step 4): The high-frequency signal injection method module and the AGA-EKF detection method module are connected in parallel and then in series at the front end of the speed switching algorithm module to form a composite method speed detection module.

[0015] Step 5): Connect the speed detection module of the composite method into the control system of the bearingless permanent magnet synchronous motor and inject a high-frequency voltage signal to achieve sensorless control.

[0016] Furthermore, the aforementioned high-frequency current Average inductance L = (L dh +L qh ) / 2, half-differential inductance ΔL=(L qh -L dh ) / 2, L dh L qh U represents the high-frequency inductance component of the torque winding on the dq axis. mh ω is the signal amplitude. h Let t be the frequency of the injected high-frequency voltage signal, and t be the time. Estimate the angle error. θ1 is the actual angle.

[0017] Furthermore, the extended Kalman algorithm includes state prediction estimation, state estimate correction, covariance estimation, error covariance matrix update, and Kalman gain calculation, continuously updating and iterating to obtain the rotational speed estimate. and angle estimates

[0018] The beneficial effects of this invention are:

[0019] 1. This invention uses a pulse array high-frequency signal injection method to detect the motor speed under zero-speed and low-speed conditions, and uses an extended Kalman filter method optimized by an adaptive genetic algorithm to detect the speed information under medium- and high-speed conditions. Compared with traditional high-frequency signal injection methods and extended Kalman algorithms, the rotor speed information detection accuracy is higher, it does not require a lot of training or the collection of many variables, the process is simple and easier to implement, and the extended Kalman filter has high accuracy in identifying angle and speed signals, which can meet the requirements of stable levitation operation of bearingless permanent magnet synchronous motors without speed sensors.

[0020] 2. This invention employs a pulse array high-frequency signal injection method and a combined method of speed detection based on extended Kalman filtering optimized by a genetic algorithm. This method can detect the speed of the motor across the entire speed range, including zero speed, low speed, and high speed, enabling smooth switching of the motor from low speed to high speed. It also achieves stable levitation operation of the bearingless permanent magnet synchronous motor across the entire speed range. Furthermore, it avoids the installation and maintenance of mechanical sensors, simplifies the control system of the bearingless permanent magnet synchronous motor, reduces control costs, and facilitates its development towards high speed and high precision. Attached Figure Description

[0021] Figure 1 This is a block diagram of the composite method speed detection module 1;

[0022] Figure 2 This is a coordinate system diagram of the motor;

[0023] Figure 3 yes Figure 1 Block diagram of module 2 for medium- and high-frequency signal injection method;

[0024] Figure 4 yes Figure 1 Flowchart of the adaptive genetic algorithm optimization in module 3 of the AGA-EKF detection method;

[0025] Figure 5 This is a schematic diagram of the linear weighted algorithm in the speed switching algorithm module 4;

[0026] Figure 6 Based on Figure 1 The diagram shows the block diagram of the bearingless permanent magnet synchronous motor speed detection module 1 of the composite method for the operation control system without speed sensor.

[0027] In the diagram: 1. Composite method speed detection module; 2. High-frequency signal injection method module; 3. AGA-EKF detection method module; 4. Speed ​​switching algorithm module; 5. PI module; 6. Park inverse converter module; 7. 2 / 3 converter module; 8. CRPWM (current regulation inverter) module; 9. BPMSM (bearingless permanent magnet synchronous motor) module; 10. Clark module; 11. Suspension force control module. Detailed Implementation

[0028] See Figure 1 Module 2 for constructing pulse array high-frequency signal injection method.

[0029] In the dq-axis coordinate system, the mathematical model of the voltage component of the torque part of a bearingless permanent magnet synchronous motor under high-frequency signal excitation can be simplified as follows:

[0030]

[0031] Converted to the differential form of the current, it is:

[0032]

[0033] In the formula: u dh u qh and i dh i qh These are the components of the high-frequency voltage and current of the torque winding on the d and q axes, respectively; L dh L qh This represents the high-frequency inductance component of the torque winding on the dq axis.

[0034] like Figure 2 The coordinate system diagram shown is dq, which is the actual rotating coordinate system. This is an estimated rotating coordinate system, where α-β is the actual two-phase stationary coordinate system. θ1 is the actual angle. It represents the estimated angle, and Δθ represents the error in the estimated angle.

[0035] Estimating the rotating coordinate system The coordinate transformation matrix to the actual rotating coordinate system dq satisfies:

[0036]

[0037] And matrix (3) satisfies:

[0038]

[0039] T is the matrix transpose.

[0040] Then by Figure 1 The estimated angular velocity error can be obtained as follows:

[0041]

[0042] In estimating the rotating coordinate system Signal injection and response signal extraction are performed, i.e., injecting a high-frequency pulse array signal, and obtaining the components i of the high-frequency current of the torque winding on the d and q axes according to coordinate transformation theory. dh i qh and high-frequency voltage component u dh uqh :

[0043]

[0044]

[0045] Combining equation (6) with the left-hand side of equation (2), we simplify the differential term to obtain:

[0046]

[0047] In the formula, ω e The actual rotor speed in the zero low-speed range.

[0048] Since the motor operates at zero low speed, equation (8) will ignore the rotor angular velocity ω. e Item can be obtained:

[0049]

[0050] Substituting equations (7) and (9) into equation (2), and combining them with equation (4), we can simplify equation (2) to obtain the relationship between high-frequency voltage and current:

[0051]

[0052] The value of the d-axis injected pulse array high-frequency voltage signal is:

[0053]

[0054] In the formula: For the injected high-frequency voltage signal, U mh ω is the signal amplitude. h Let t be the frequency of the injected high-frequency signal and t be the time.

[0055] Combining equations (10) and (11), the high-frequency current can be obtained as follows:

[0056]

[0057] Where, average inductance L = (L dh +L qh ) / 2, half-differential inductance ΔL=(L qh -L dh ) / 2.

[0058] From equation (12), it can be seen that the estimated q-axis current signal includes the rotor position estimation error, and the q-axis current is:

[0059]

[0060] The structure of module 2 for high-frequency signal injection is as follows: Figure 3As shown, the current i in the two-phase stationary coordinate system α-β α i β The current i in the rotating coordinate system dq is obtained through the inverse Park transform. d i q Current i d i q After further filtering by a BPF (bandpass filter), a high-frequency current containing rotational speed information is obtained. Then, the high-frequency current containing rotational speed information is... and modulated current signal sinω h After multiplication by t, the signal passes through an LPF (low-pass filter) to remove low-frequency signals.

[0061] The error function f(Δθ) is obtained:

[0062]

[0063] In the formula: The error function f(△θ) is adjusted by the PI module. By adjusting the PI module, equation (14) can be made zero, thus estimating the angle error. This allows the estimated angular position to be equal to the actual angular position, i.e., the estimated rotor angle under zero-speed conditions. Equal to the actual angle θ1, the estimated angle value can be obtained. And output the estimated value of the angle. Angle estimates By performing differentiation, the estimated rotor speed under zero low-speed conditions can be obtained. Achieve self-detection of rotor speed under zero-speed and low-speed conditions.

[0064] See Figure 1 A detection module based on the extended Kalman filter (EKF) optimized by the adaptive genetic algorithm (AGA) is constructed.

[0065] First, establish the state equations for the rotating part of the bearingless permanent magnet synchronous motor in the stationary coordinate system:

[0066]

[0067] In the formula: i α i β α and β are the current components of the torque winding on the α and β axes, respectively; p1 is the number of pole pairs of the motor; Ψ is the rotor flux linkage of the motor; r1 is the stator resistance; L1 is the inductance coefficient of the stator winding; ω2 is the actual rotor speed under medium and high speed conditions; θ2 is the actual rotor angle under medium and high speed conditions.

[0068] According to equation (15), select the state variable x = [i α i β [ω2, θ2]T ,but Control variable u = [u α u β ] T Output variable y = [i α i β ] T ,in Let x denote the first-order differential of x, and T denote the transpose of the matrix.

[0069] The state equations for the bearingless permanent magnet synchronous motor are as follows:

[0070]

[0071] In the formula: Ψ f This indicates a permanent magnet flux linkage.

[0072] Equation (16) can be transformed into:

[0073]

[0074] In the formula: Let represent the first differential with respect to y, where

[0075]

[0076] The approximate discrete equation is:

[0077]

[0078] In the formula: I is the identity matrix; t is the sampling interval; x k Let x be the system state variable at time k; k-1 y is the system state state at time k-1; k For the system output variable at time k; u k-1 Let be the control variables of the system at time k-1.

[0079] From equation (18)x k y k Taking the partial derivative with respect to x, we obtain the Jacobian matrix and the transfer matrix, and the discretized transfer Jacobian matrix of the system.

[0080] The ratio matrix is:

[0081]

[0082] The transfer matrix is:

[0083]

[0084] The optimal values ​​of covariance matrices Q and R are obtained by using an adaptive genetic algorithm (AGA). The process of AGA to optimize covariance matrices Q and R is as follows: Figure 4 As shown, the specific optimization process is as follows: Select the state variable and output variable as x = [i α i β [ω2, θ2] T y = [i α i β ] T Let the Q and R matrices be Q = Diag(Q) 11 Q 22 Q 33 Q 44 ), R = Diag(R 11 R 22 ), where Q 11 In the state variable x, i β The variation error, Q 22 In the state variable x, i β The variation error, Q 33 Q represents the error in the change of ω2 in the state variable x. 44 R represents the error in the change of θ2 in the state variable x. 11 Indicates i in the output variable y α The variation error, R 22 In the state variable y, i β The variation error. Take its six elements to form ε=(Q 11 Q 22 Q 33 Q 44 R 11 R 22 The population is initialized, and its value range and precision are determined. First, the initial population parameters are binary encoded to obtain chromosome 1. A crossover operation is performed on chromosome 1 to obtain sub-chromosome 2. A mutation operation is performed on sub-chromosome 2 to obtain sub-chromosome 3. Chromosomes 1, 2, and 3 are decoded and their fitness calculated. The results are then fed into a roulette wheel selection model to obtain a new set of chromosomes. A termination condition is determined; if the condition is not met, the fitness value of the population is calculated. Based on the fitness value, an adaptive algorithm is used to adaptively adjust the crossover probability P. c Probability of mutation P m The value of the chromosome is updated, and if the condition is met, the optimal solution is obtained and the global optimal solution is output. The adaptive algorithm formula is:

[0085]

[0086]

[0087] In the formula: f max It is the maximum fitness value of the population, f avg f' is the average fitness value of the population, b1 and b3 are the crossover probability parameters, and b2 and b4 are the mutation probability parameters. These parameters are selected according to the optimization process. By repeating the process of selection, crossover, and mutation, the optimal chromosome that satisfies the convergence condition of the entire population can be obtained, that is, the Q and R noise matrices that satisfy the optimal filtering condition can be obtained.

[0088] Given the optimal covariance matrices Q and R, the extended Kalman algorithm is used to obtain the rotor speed estimate under high-speed conditions. and angle estimates The specific implementation process of the Extended Kalman Algorithm is as follows:

[0089] State prediction estimation:

[0090]

[0091] State estimate correction:

[0092]

[0093] Covariance estimation:

[0094] P k / k-1 =P k-1 +(F k-1 P k-1 +F T k-1 P k-1 )·t+Q (23)

[0095] Error covariance matrix update:

[0096] P k =P k / k-1 -K k HP k / k-1 (twenty four)

[0097] Kalman gain calculation:

[0098]

[0099] In the formula: k / k-1 represents the state transition from time k-1 to time k; the subscripts k and k-1 both represent that time. These are estimates of the state variables; Let f be the state quantity at time k, and the nonlinear function f represents the relationship between the (k-1)th order state and the kth order state; B k-1 Let u represent the system state variables of B at time k-1. k-1 Let K be the control variable of the system at time k-1.k It is the gain matrix; H k The transfer matrix at time k; P is the transfer matrix at the state transition time. k P k / k-1 P k-1 These are the error covariance matrices at the corresponding times; F k-1 It is the gradient matrix at time k-1, x k-1 y is the system state state at time k-1; k u is the system output variable at time k; k-1 Let be the control variables of the system at time k-1.

[0100] Under medium- and high-speed operating conditions, the current i ɑ i β and voltage u α u β The input is fed into the AGA-EKF (Adaptive Genetic Algorithm-Optimized Extended Kalman Filter) detection module 3, and the state variable x = [i] is selected according to equation (15). α i β [ω2, θ2] T ,but That is, x3 = ω2, x4 = θ2. For the optimal Q and R values, according to equations (21) to (25), through the process of state prediction estimation, state estimate correction, covariance estimation, error covariance matrix update, and Kalman gain calculation, ω2 and θ2 are continuously updated and iterated, which are the rotor speed estimates under medium and high speed conditions. and angle estimates

[0101] See Figure 1 Module 4 is constructed to implement the speed switching algorithm.

[0102] Speed ​​switching algorithm module 4 uses the rotor speed estimate under medium and high speed conditions. and angle estimates Rotor speed estimation under zero low-speed conditions and angle estimates As input, the rotor speed detection value and angle detection value This is the output.

[0103] A speed switching algorithm is set up, combining the pulse array high-frequency signal injection method and the extended Kalman filter method optimized by the adaptive genetic algorithm to achieve smooth switching of the motor from low speed to high speed. First, the following weighting coefficient ξ1 is determined:

[0104]

[0105] ω is the actual speed of the motor, that is, when the actual speed of the motor ω≤ωL When the actual angular velocity of the motor ω≥ω, the weighting coefficient is 1; H When ω = 0, the weighting coefficient is 0; when ω = 0, the weighting coefficient is 0. L ≤ω≤ω H When, the weighting coefficient is ω L It switches the lower limit of the rotational speed, ω H It switches the maximum speed limit.

[0106] Then, calculate the rotor speed detection value output by the speed switching algorithm module 4 according to the following formula. and angle detection value The following equations must be satisfied:

[0107]

[0108] like Figure 1 As shown, the high-frequency signal injection method module 2 and the AGA-EKF (Adaptive Genetic Algorithm-Optimized Extended Kalman Filter) detection module 3 are connected in parallel and then in series at the front end of the speed switching algorithm module 4 to form the composite method speed detection module 1. The input of the composite method speed detection module 1 is the current i. ɑ i β and voltage u α u β The output is the rotor speed detection value. and angle detection value

[0109] When the composite method speed detection module 1 is working, under the motor's zero-low speed condition, the motor current i ɑ i β The current i is input to the high-frequency signal injection module 2. ɑ i β The current i is obtained after Park inverse transformation. d i q The high-frequency current signal containing rotational speed information is then obtained after BPF filtering. Then combine it with the modulation signal sinω h The product of t and t is processed by an LPF to remove low-frequency signals and then adjusted to obtain an angle estimate. Then, the estimated rotor speed is obtained by differentiation.

[0110] Under medium- and high-speed operating conditions, the current i ɑ i β and voltage u ɑ u β The inputs are fed into the AGA-EKF (Adaptive Genetic Algorithm-Optimized Extended Kalman Filter) detection module 3, and the state variable x = [i] is selected according to equation (15). α iβ [ω2, θ2] T ,but That is, x1 = i ɑ x2 = i β x3 = ω2, x4 = θ2. According to equations (19) to (21), through processes such as state prediction estimation, state estimate correction, covariance estimation, error covariance matrix update, and Kalman gain calculation, the rotor speed estimate is continuously updated and iterated to obtain the rotor speed estimate. and angle estimates

[0111] Detected by high frequency signal injection method and obtained by AGA-EKF testing The input is fed into the speed switching algorithm module 4. After weighted algorithm, the rotor speed detected by the composite speed estimation method is calculated by equations (26) and (27). and angle detection value When the estimated actual motor speed is lower than the lower switching limit, the high-frequency signal injection method is used; when the estimated actual motor speed is higher than the upper switching limit, the method switches to AGA-EKF. If the speed estimate is in the transition range, a weighted algorithm is used to estimate the speed.

[0112] Will Figure 1 The speed detection module 1 based on the composite control method shown is connected in series with the control system of the bearingless permanent magnet synchronous motor to achieve sensorless control of the bearingless permanent magnet synchronous motor.

[0113] like Figure 6 The bearingless permanent magnet synchronous motor speed sensorless control system based on the composite method shown consists of a conventional torque control section, a levitation force control section, and the composite method speed detection module 1 of this invention.

[0114] In the torque control section, the high-frequency signal U mh cos(ω h The t) is converted by the dq / abc converter module 10 and applied to the BPMSM (bearingless permanent magnet synchronous motor), and the three-phase torque current i is output by the CRPWM (current regulation inverter) module 5. 1a i 1b i 1c The torque winding module for controlling a bearingless permanent magnet synchronous motor, with three-phase torque control current i 1a i 1b i 1c Transformed into the α-β axis coordinate system by Clark transformation module 8 ɑ i β Meanwhile, the three-phase torque voltage u output by CRPWM (Current Regulated Inverter) module 5...a u b u c After being transformed by Clark transformation module 7, u is in the α-β axis coordinate system. ɑ u β Voltage u ɑ u β and current i ɑ i β The input is fed into the composite method speed detection module 1, which detects the speed of the bearingless permanent magnet synchronous motor and obtains the speed detection value. Speed ​​detection value The speed difference is obtained by subtracting the reference speed ω* from the speed difference, and then inputting the speed difference into PI module 2 to obtain the reference torque current i. 1q *, Set the reference torque current i 1d * is 0. i 1d *、i 1q *Transformed into the α-β axis coordinate system by Park inverse transformation module 3. ɑ *、i β *, and then converted into current i through the 2 / 3 conversion module 4. 1a *、i 1b *、i 1c *, the current i 1a *、i 1b *、i 1c *Converted to three-phase torque current i via CRPWM (Current Regulated Inverter) module 5. 1a i 1b i 1c This completes the speed observation of the torque component of the bearingless permanent magnet synchronous motor.

[0115] In the levitation force control section, displacement sensors are used to detect the actual displacements x and y of the bearingless permanent magnet synchronous motor. The difference between the actual displacements x and y and the set reference displacements x* and y* is calculated, and the resulting displacement difference is input to the levitation force control module 9, with the output being a three-phase torque current i. 2a i 2b i 2c Displacement observation of the suspension control section of the bearingless permanent magnet synchronous motor was completed.

[0116] The present invention can be realized based on the above description.

Claims

1. A method for constructing a bearingless permanent magnet synchronous motor with a full speed range without a speed sensor, characterized in that: include: Step 1): Constructing a high-frequency signal injection module The current in the two-phase stationary coordinate system α-β The current in the rotating coordinate system dq is obtained by inverse Park transformation. Current After passing through a BPF filter, a high-frequency current containing rotational speed information is obtained. Then, the high-frequency current and modulated current signal After multiplication, the error function is obtained through LPF. The error function is adjusted using a PI module. Set it to 0 to obtain the angle estimate under zero low-speed conditions. And the angle estimate The speed estimate is obtained by differentiation. ; Step 2): Constructing the AGA-EKF detection module Selecting state variables and output variables Take the covariance matrix and The optimal values ​​of covariance matrices Q and R are obtained by selecting them using the AGA method. They are respectively The variation error, They are respectively The variation error; These are the actual rotor speed and actual angle under medium- and high-speed operating conditions, respectively, and T is the matrix transpose. The AGA method described above selects the covariance matrices Q and R as follows: The population is initialized to determine the range of values ​​and population precision. The initial population parameters are binary encoded to obtain chromosome 1. A crossover operation is performed on chromosome 1 to obtain sub-chromosome 2. A mutation operation is performed on sub-chromosome 2 to obtain sub-chromosome 3. Chromosomes 1, 2, and 3 are decoded and their fitness values ​​are calculated. The results are output to the roulette wheel selection model to obtain a new set of chromosomes. The termination condition is determined. If the condition is not met, the fitness value of the population is calculated. Based on the fitness value, the crossover probability and mutation probability are adaptively changed and the chromosomes are updated using an adaptive algorithm. If the condition is met, the optimal solutions Q and R are obtained. Based on the optimal Q and R values, the rotor speed estimate under medium- and high-speed operating conditions is obtained using the extended Kalman algorithm. and angle estimates ; The extended Kalman algorithm includes state prediction estimation, state estimate correction, covariance estimation, error covariance matrix update, and Kalman gain calculation, continuously updating and iterating to obtain the rotational speed estimate. and angle estimates ; The state prediction estimate is: The state estimate correction is: The covariance estimate is: , The error covariance matrix update is: The Kalman gain is calculated as follows: k / k-1 represents the state transition from time k-1 to time k; the subscripts k and k-1 both represent that time. These are estimates of the state variables; These are the state variables at the corresponding time points, and the nonlinear function f represents the relationship between the (k-1)th order state and the kth order state. This represents the system state variables of B at time k-1. , It is the inductance coefficient of the stator winding. Let K be the control variable of the system at time k-1. k It is the gain matrix; H k The transfer matrix at time k; P is the transfer matrix at the state transition time. k , , These are the error covariance matrices at the corresponding times; F k-1 It is the gradient matrix at time k-1. Let k be the system state at time k-1; Let k be the system output variable at time k; These are the control variables of the system at time k-1; Step 3): Construct the speed switching algorithm module Determine the speed switching weighting coefficient, and calculate the rotor speed detection value based on the weighting coefficient. and angle detection value ; Step 4): The high-frequency signal injection method module and the AGA-EKF detection method module are connected in parallel and then connected in series at the front end of the speed switching algorithm module to form a composite method speed detection module. Step 5): Connect the speed detection module of the composite method into the control system of the bearingless permanent magnet synchronous motor and inject a high-frequency voltage signal to achieve sensorless control.

2. The method for constructing a bearingless permanent magnet synchronous motor with a full speed range without a speed sensor according to claim 1, characterized in that: The high-frequency current Average inductance Half-differential inductor , This represents the high-frequency inductance component of the torque winding on the dq axis. The signal amplitude, The frequency of the injected high-frequency voltage signal, For time, estimate the angle error. , That's from a practical perspective.

3. The method for constructing a bearingless permanent magnet synchronous motor with a full speed range without a speed sensor according to claim 2, characterized in that: The error function , .

4. The method for constructing a bearingless permanent magnet synchronous motor with a full speed range without a speed sensor according to claim 1, characterized in that: The variable crossover probability Probability of mutation f max It is the maximum fitness value of the population, f avg It is the average fitness value of the population, f ， The goal is to find the larger fitness value among the crossover individuals. b1 and b3 are crossover probability parameters, and b2 and b4 are mutation probability parameters. The parameters are selected based on the optimization process. By repeatedly selecting, crossing over, and mutating, the optimal chromosome that satisfies the convergence condition of the entire population can be obtained, which is to say, the Q and R noise matrices that satisfy the optimal filtering condition can be obtained.

5. The method for constructing a bearingless permanent magnet synchronous motor with a full speed range without a speed sensor according to claim 1, characterized in that: The weighting coefficients .

6. The method for constructing a bearingless permanent magnet synchronous motor with a full speed range without a speed sensor according to claim 5, characterized in that: From the formula and Calculate the rotor speed detection value and angle detection value .

Images