Novel intelligent motor control method based on model predictive control

Abstract

The invention relates to a novel intelligent motor control method based on model predictive control. The method comprises the following steps: acquiring a two-phase rotating coordinate system current signal and a rotating angular velocity signal of a motor; measuring the noise covariance matrix to identify the stator resistance, inductance and flux linkage of the motor based on the current signal, the rotation angular velocity signal and the noise covariance matrix by adopting an online parameter identification method of extended Kalman filtering; wherein the noise covariance matrix and the measurement noise covariance matrix are set based on an improved artificial jellyfish search algorithm; a motor parameter model is obtained based on stator resistance, inductance and flux linkage of the motor, and speed and position control over the motor is achieved through model prediction control based on the motor parameter model. Compared with the prior art, the method has the advantages that the motor control convergence speed is increased, local optimum is reduced, and an accurate motor model is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent motor control, and in particular to a novel intelligent motor control method based on model predictive control. Background Art

[0002] Most devices on the market can only realize the control of various motor starting and stopping modes in terms of motor control. The speed and position control of the motor tends to be traditional control methods, with low control accuracy and poor reliability. In terms of fault diagnosis, they only diagnose overload, phase failure, three-phase imbalance, grounding and other faults by monitoring various current and voltage signals. The types of fault detection are limited and rely on the accuracy of the measured current signal. When the relevant sensor fails, the fault diagnosis will also fail.

[0003] In order to solve the parameter sensitivity problem of model predictive control in the existing control methods, the extended Kalman filter algorithm will be used for online parameter identification. The method of applying the extended Kalman filter algorithm to online parameter identification mainly focuses on how to obtain the optimal noise covariance matrix Q and the measurement noise covariance matrix R. The tuning methods of Q and R include heuristic tuning method and intelligent algorithm-based tuning method. The heuristic tuning method relies on experience, the process is complicated, and the final Q and R are not necessarily optimal; although the commonly used genetic algorithm has a good effect, it is a parameter algorithm that requires the setting of parameters such as crossover factor and mutation rate, and improper parameter setting will lead to increased calculation and decreased prediction accuracy. The artificial jellyfish search algorithm is a parameter-free intelligent optimization algorithm. The time control mechanism in the algorithm can coordinate various types of movements within the population, which is more advantageous than the genetic algorithm. However, the artificial jellyfish search algorithm has problems such as slow convergence speed and easy to fall into local optimality, resulting in inaccurate motor model. Summary of the invention

[0004] The purpose of the present invention is to provide a novel intelligent motor control method based on model predictive control in order to improve the motor control convergence speed, reduce falling into local optimum, and obtain an accurate motor model.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A novel intelligent motor control method based on model predictive control comprises the following steps:

[0007] Get the two-phase rotating coordinate system current signal I of the motor d ,I q And the angular velocity signal ω e ;

[0008] The online parameter identification method based on the extended Kalman filter is used. d ,Iq And the angular velocity signal ω e , and the noise covariance matrix Q, the measurement noise covariance matrix R identify the stator resistance R, inductance L and flux ψ of the motor f ;

[0009] The noise covariance matrix Q and the measurement noise covariance matrix R are adjusted based on the improved artificial jellyfish search algorithm;

[0010] Based on the motor's stator resistance R, inductance L and flux linkage ψ f The motor parameter model is obtained, and the speed and position control of the motor is realized by using model predictive control based on the motor parameter model.

[0011] Furthermore, the specific steps of adjusting the noise covariance matrix Q and the measurement noise covariance matrix R based on the improved artificial jellyfish search algorithm are:

[0012] (1) Determine the population size N and the maximum number of iterations MaxIt, initialize the population position within the search interval, each individual has d optimization parameters, and set the initial number of iterations;

[0013] (2) Calculate the fitness values ​​of all individuals. The fitness function calculation formula is as follows: The position X of the optimal value of the population is obtained best And the corresponding fitness value f best ; The fitness value of the individual is the mean square error between the d and q axis current measurement values ​​and the EKF estimated value. The specific formula is:

[0014]

[0015] i d (k) and i q (k) is the measured value of the dq axis current at time k, and is the EKF estimated value of the dq axis current at time k;

[0016] (3) Use the time control function value corresponding to the time control mechanism. If the time control function value is greater than the threshold, the individual follows the ocean current and calculates the individual position. Otherwise, the individual performs type A or type B movement. During the process of the individual performing type A or type B movement, if the random value of 0-1 is greater than or equal to 1-c(t), the individual performs type A movement and calculates the individual position. Otherwise, the individual performs type B movement and calculates the individual position.

[0017] (4) Based on the pelican surface flight mechanism including adaptive selection and Levy flight strategy, the individual position is updated to obtain a new individual;

[0018] (5) Calculate the fitness value of the new individual, update the position of the optimal value of the population based on the fitness value of the new individual, and update the current number of iterations;

[0019] (6) Determine whether the current number of iterations is greater than the maximum number of iterations MaxIt. If so, output the position X of the optimal value of the population best , otherwise return to (3).

[0020] Furthermore, the individual position when following the ocean current is:

[0021]

[0022] Among them, X i (t+1) represents the position of the jellyfish of the i-th individual in the t+1-th iteration; X * is the jellyfish with the best position in the current jellyfish group, β is the distribution coefficient, and rand(0,1) is a random value between 0 and 1; is the average position of all jellyfish.

[0023] Furthermore, the individual position when the individual performs type A movement is:

[0024] X i (t+1)=X i (t)+γ*rand(0,1)*(U b -L b )

[0025] Among them, U b and L b are the upper and lower bounds of the search interval; γ is the motion coefficient.

[0026] Furthermore, the individual position when the individual performs type B movement is:

[0027] X i (t+1)=X i (t)+S

[0028] Where S represents the movement step length, S = rand (0, 1) * D, and D represents the movement direction, which is:

[0029]

[0030] Among them, X j represents the position of the jth individual jellyfish, f(X i ) represents the food content at the location of the ith individual, f(X j ) represents the food content at the location of the jth individual.

[0031] Furthermore, the individual position is updated based on the pelican surface flight mechanism including adaptive selection and Levy flight strategy as follows:

[0032]

[0033] in,

[0034]

[0035] Where d is the dimension; j∈[1,d] represents the ordinal number of the dimension, t represents the tth iteration, and Z represents X i The update method of the d elements, Q represents the ratio of the remaining number of iterations to the maximum number of iterations, Levy represents the random step length of Levy flight, T is a constant of 0.2, X ij represents the jth element of the i-th individual position, Xbest ij Represents the current optimal value of the jth element in the i-th individual position.

[0036] Further, Levy specifically states:

[0037]

[0038] Among them, u and v are random variables that obey the normal distribution, η=2*rand.

[0039] Furthermore, the random variables u and v that obey the normal distribution satisfy u~N(0,σ 2 ),v~N(0,1), where σ represents the variance.

[0040] Furthermore, the variance σ is:

[0041]

[0042] Among them, Γ represents and β is the distribution coefficient.

[0043] Furthermore, the method also includes: in the process of realizing speed and position control of the motor by using model predictive control based on the motor parameter model, the motor parameter model outputs a predicted current, and a residual signal is generated after comparing the predicted current with the measured signal, and then a discrete wavelet transform is used to obtain a time-frequency domain diagram of the residual signal, and fault features are extracted through two-dimensional convolution 2D-CNN, and then the SVM multi-classifier is used to obtain the motor fault type to realize motor fault diagnosis.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present invention uses a parameter-free intelligent algorithm to more effectively adjust the Q and R matrices. At the same time, it combines the artificial jellyfish search algorithm based on the Pelican surface flight mechanism of adaptive Levy diversity to improve its convergence speed of searching for the optimal matrix and its ability to jump out of the local optimum, so that the Q and R matrices in the extended Kalman filter can obtain the optimal solution. This method improves the accuracy of online parameter identification, reduces noise sensitivity, obtains an accurate motor model, and avoids the problem of poor control effect caused by inaccurate parameters in model predictive control. . BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is the overall flow chart of the present invention;

[0047] Figure 2 It is the extended Kalman filter flow chart of the present invention;

[0048] Figure 3 A flow chart of adjusting the noise matrix for the artificial jellyfish search algorithm of the present invention;

[0049] Figure 4 Predict control flow graph for the model;

[0050] Figure 5 A schematic diagram of the simplified components of a single functional unit. DETAILED DESCRIPTION

[0051] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0052] Aiming at the problem that the existing intelligent motor control center has low control accuracy for the speed and position of the motor, the fault diagnosis method depends on the sensor and the diagnosed fault types are limited, the present invention adopts the improved extended Kalman filter and the improved artificial jellyfish algorithm to identify the motor model parameters online, and then uses the model predictive control to realize the precise control of the motor speed and position; at the same time, the accurate motor model is used to predict the current signal, and the residual is obtained after comparison with the measured signal, and then the discrete wavelet transform, 2D-CNN, SVM and other methods are used to more accurately diagnose more types of motor faults. Each intelligent motor control unit of the present invention uses the current sensor and the position sensor to sample and obtain the three-phase current and rotation angle signal of the motor operation, and obtains the current signal and angular velocity signal of the two-phase rotating coordinate system through current transformation and derivative transformation, and then uses the extended Kalman filter method to identify the motor parameters online, and at the same time uses the improved artificial jellyfish search algorithm to adjust the noise covariance matrix and the measurement noise covariance matrix of the extended Kalman filter, and after obtaining the accurate motor parameter model, the model predictive control is used to realize the speed and position control of the motor. At the same time, the current signal is predicted according to the motor model, and the residual signal is obtained after comparison with the measured signal. The time-frequency domain diagram is obtained by discrete wavelet transform, and the fault features are extracted using a two-dimensional convolutional neural network (2D-CNN). Then, the fault type is determined by combining it with the motor parameter changes using a support vector machine (SVM) multi-classifier. Each intelligent motor control unit is integrated into a control cabinet to form an intelligent motor control center.

[0053] The overall algorithm block diagram of each intelligent motor control unit is as follows: Figure 1 As shown in the figure, it mainly consists of three parts: online parameter identification based on extended Kalman filter, model predictive control and fault diagnosis.

[0054] correspond Figure 1 The novel intelligent motor control method based on model predictive control proposed by the present invention comprises the following steps:

[0055] Get the two-phase rotating coordinate system current signal I of the motor d ,I q And the angular velocity signal ω e ;

[0056] The online parameter identification method based on the extended Kalman filter is used. d ,I q And the angular velocity signal ω e , and the noise covariance matrix Q, the measurement noise covariance matrix R identify the stator resistance R, inductance L and flux ψ of the motor f ;

[0057] The noise covariance matrix Q and the measurement noise covariance matrix R are adjusted based on the improved artificial jellyfish search algorithm;

[0058] Based on the motor's stator resistance R, inductance L and flux linkage ψ f The motor parameter model is obtained, and the speed and position control of the motor is realized by using model predictive control based on the motor parameter model.

[0059] The online parameter identification method based on extended Kalman filter uses the two-phase rotating coordinate system current signal I obtained by acquisition and transformation. d ,I q And the angular velocity signal ω e And the noise covariance matrix Q, the measurement noise covariance matrix R identify the stator resistance R, inductance L and flux ψ of the motor f , where the artificial jellyfish search algorithm adjusts the noise covariance matrix Q and the measurement noise covariance matrix R by comparing the current value in the motor model with the measured current value.

[0060] The flowchart of extended Kalman filter for online parameter identification is as follows: Figure 2 The input value and noise of the system are calculated through the state equation to obtain the system state vector, and then the measurement value of the system is obtained through the measurement noise and observation equation; then the measurement value is compared with the predicted value obtained through the prediction equation, and the estimated value is obtained by using the correction equation, which is further substituted into the prediction equation to obtain the predicted value at the next moment.

[0061] Extended Kalman filter algorithm:

[0062] The state equation and observation equation of the discrete nonlinear system of the extended Kalman filter are:

[0063]

[0064] In formula (1-1), x k and x k-1 is the state vector of the system, u k-1 is the input vector of the system; z k is the measurement vector of the system; w k-1 、v k are Gaussian noises with zero mean, and are system and measurement noises, respectively.

[0065] The extended Kalman filter is generally divided into two stages: prediction and correction.

[0066] In the prediction stage, the state vector and error covariance matrix are initialized, and the state prediction vector at the k-1 sampling time is calculated according to the discrete state space equation of the system without considering the system process noise. and the estimated error prediction covariance matrix P k|k-1 The formula is as follows:

[0067]

[0068] In formula (1-2), is the state transfer matrix; Q and R are w k-1 、v k The covariance matrix of .

[0069] In the correction phase, the actual value of the output measurement is used as the feedback, and the state variables and the error prediction covariance matrix P are further corrected. k|k-1 , thus obtaining the best estimate of time k The formula is as follows:

[0070]

[0071] In formula (1-3), is the state observation matrix; I is the unit matrix. Based on the known state prediction vector, the linear minimum mean square error is used as the basic criterion. and the actual measured value y k , combined with the state estimation error prediction covariance matrix P k|k-1 Calculate the Kalman gain K k , and finally get the best estimate

[0072] The process of the artificial jellyfish search algorithm to adjust the noise covariance matrix Q and the measurement noise covariance matrix R is as follows Figure 3 The current, voltage and speed measurement values ​​can be used to obtain the estimated value of the current after the extended Kalman filter, and the fitness value is calculated by the fitness function with the current measurement value, which is substituted into the artificial jellyfish search algorithm to continuously optimize the noise matrix and feed back to the calculation process of the extended Kalman filter.

[0073] The artificial jellyfish algorithm is a meta-inspiration algorithm developed by studying the behavior of jellyfish in the ocean. Its core formula is as follows:

[0074] X i (t+1)=X i (t)+rand(0,1)*(X*-β*rand(0,1)*u)#(2-1)

[0075] Formula (2-1) describes the movement of jellyfish following the ocean currents. i (t+1) is the position of the jellyfish at the t+1th iteration; X * is the jellyfish with the best position in the current jellyfish group; β is the distribution coefficient, which is 3; rand(0,1) is a random value between 0 and 1; u is the average position of all jellyfish.

[0076] X i (t+1)=X i(t)+γ*rand(0,1)*(U b -L b )#(2-2)

[0077] Formula (2-2) describes the A-type movement of jellyfish within the population. b and L b are the upper and lower bounds of the search interval; γ is the motion coefficient, and its value is 0.1.

[0078] S=X i (t+1)-X i (t)#(2-3)

[0079] S=rand(0,1)*D#(2-4)

[0080]

[0081] In formula (2-5), f is the objective function value at position X.

[0082] After simplification, we can get:

[0083] X i (t+1)=X i (t)+S#(2-6)

[0084] Formulas (2-3)-(2-6) describe the type B movement of jellyfish within a population.

[0085] The AJS algorithm uses a time control mechanism to regulate the above three movements. The formula is as follows:

[0086]

[0087] In formula (2-7): t is the current number of iterations, and MaxIt is the maximum number of iterations.

[0088] As the number of iterations of the artificial jellyfish algorithm increases, the quality of the population will continue to decline, which will in turn lead to a decrease in the global search ability of the algorithm. This paper introduces the Pelican surface flight mechanism with adaptive Levy diversity for improvement.

[0089] The pelican optimization algorithm simulates the hunting process of pelicans in nature. The pelican's surface flight mechanism is to spread its wings on the water surface after arriving near the prey and catch the prey. The calculation formula is:

[0090]

[0091] In formula (2-8), T is a constant with a value of 0.2; t is the current number of iterations; MaxIt is the maximum number of iterations; rand is a random number between 0 and 1; is the search radius of the Pelican. In the early stage, the radius is large and the search area is also large. In the later stage, the radius gradually decreases and the search area also gradually shrinks, which helps the algorithm better converge to the global optimal solution.

[0092] Adding the Levy flight mechanism to the pelican surface flight mechanism can improve the quality of the population. The small step tracking feature of Levy flight can enhance the algorithm's local search ability, enabling the algorithm to obtain a more accurate optimal solution and improve the quality of the population. In addition, the long-distance jump feature of Levy flight disturbs the stable path, helps the algorithm jump out of the local optimal solution, and further increases the diversity of the population. The formula is as follows:

[0093]

[0094] u~N(0,σ 2 ),v~N(0,1)#(2-10)

[0095] η=2*rand#(2-11)

[0096]

[0097] Formulas (2-9)-(2-12) calculate Levi's random step length for flight.

[0098] Adding Levi's flight strategy to the Pelican's surface flight mechanism, the formula is as follows:

[0099]

[0100] In order to further enhance the algorithm's ability to escape from local optimality, an adaptive selection link is added. The formula is as follows:

[0101]

[0102] In formulas (2-9)-(2-16), u and v are random variables that obey the normal distribution; d is the dimension; j∈[1,d]; and t is the current number of iterations.

[0103] The process of improving the artificial jellyfish search algorithm is as follows:

[0104] 1. Algorithm initialization. Determine the population size N and the maximum number of iterations MaxIt, and initialize the population position within the search interval. Each individual in the population has d optimization parameters, that is, the maximum dimension of the variable, and set the initial number of iterations t = 1.

[0105] 2. Calculate the fitness values ​​of all individuals and obtain the position X of the optimal value of the population best And the fitness value f best .

[0106] 3. Calculate c(t) in the time control mechanism according to formula (2-7). If the value of c(t) is greater than or equal to 0.5, the individual follows the ocean current and calculates the individual position according to formula (2-1); otherwise, it performs type A or type B movement. If the random value of 0-1 is greater than or equal to 1-c(t), the individual performs type A movement and calculates the position according to formula (2-2). Otherwise, it performs type B movement and calculates the position according to formula (2-6).

[0107] 4. Update the population position of the introduced mechanism according to formulas (2-9)-(2-16).

[0108] 5. X new Perform area detection and correction. Calculate X new The fitness function value of the jellyfish is updated to update the individual position and X best .

[0109] Determine whether the current number of iterations is greater than the maximum number of iterations MaxIt. If it is less than or equal to it, return to continue execution; otherwise, output the optimal value.

[0110] The flowchart of model predictive control is as follows Figure 4 The inner loop obtains the three-phase control quantity of the inverter according to the current prediction model and the value optimization function, and the outer loop still uses the proportional integral control method to improve the accuracy and stability of the control system.

[0111] The fault diagnosis method is to predict the current based on an accurate motor model, generate a residual signal after comparing it with the measured signal, and then use discrete wavelet transform to obtain the time-frequency domain diagram of the residual signal, extract the fault features through 2D-CNN, and finally use SVM multi-classifier to obtain the motor fault type and realize motor fault diagnosis.

[0112] The advantages of the present invention compared with the prior art are:

[0113] (1) High-precision motor parameter model. The extended Kalman filter is used for online parameter identification, and the artificial jellyfish search algorithm is used to adjust the noise matrix. This can avoid the problem of poor model prediction control effect caused by motor parameter changes due to temperature rise and magnetic field saturation in industrial environments, and improve the accuracy of the motor parameter model.

[0114] (2) Achieve precise control of motor speed and position. Model predictive control is performed after using the method of online motor parameter identification, which effectively avoids the parameter sensitivity problem of model predictive control and can achieve higher precision control of motor speed and position.

[0115] (3) A wider range of faults can be diagnosed. The current error signal and motor parameter changes are used to detect motor faults through the support vector machine multi-classifier method. Various faults can be diagnosed, including turn-to-turn short circuit, rotor bar breakage, phase failure, air gap eccentricity, voltage anomaly, bearing failure, and looseness failure.

[0116] (4) Better fault diagnosis effect. Applying methods such as model-based, discrete wavelet transform, two-dimensional convolutional neural network, and support vector machine to fault diagnosis, and implementing fault diagnosis based on current prediction signals and measured signals can combine the advantages of various methods to improve the accuracy of fault diagnosis.

[0117] The artificial jellyfish algorithm is improved to some extent and combined with the extended Kalman filter. The advantages are:

[0118] 1. In view of the parameter sensitivity problem in model predictive control, the extended Kalman filter algorithm will be used for online parameter identification. The method of applying the extended Kalman filter algorithm to online parameter identification mainly focuses on how to obtain the optimal noise covariance matrix Q and the measurement noise covariance matrix R. The tuning methods of Q and R include heuristic tuning method and intelligent algorithm-based tuning method. The heuristic tuning method relies on experience, the process is complicated, and the final Q and R are not necessarily optimal; although the commonly used genetic algorithm has a good effect, it is a parameter algorithm that requires the setting of parameters such as crossover factor and mutation rate, and improper parameter setting will lead to increased calculation and decreased prediction accuracy. Therefore, the use of non-parameter intelligent algorithm can more effectively tune the Q and R matrices. The artificial jellyfish search algorithm is a non-parameter intelligent optimization algorithm. The time control mechanism in the algorithm can coordinate various movements within the population, which is more advantageous than the genetic algorithm.

[0119] 2. However, the artificial jellyfish search algorithm has problems such as slow convergence speed and easy to fall into local optimality. Therefore, the artificial jellyfish search algorithm is improved by using the Pelican surface flight mechanism based on adaptive Levy diversity to improve its convergence speed of searching for the optimal matrix and its ability to jump out of the local optimality, so that the Q and R matrices in the extended Kalman filter can obtain the optimal solution. This method improves the accuracy of online parameter identification, reduces noise sensitivity, obtains an accurate motor model, and avoids the problem of poor control effect caused by inaccurate parameters in model predictive control.

[0120] The main difference between an intelligent motor control center and a traditional motor control center is the use of microprocessors, network communications and protection devices. Therefore, during the implementation process, current transformers, high-performance DSP (digital signal processing) controllers, human-machine interfaces, fault protection devices and other equipment are required. These devices are integrated into a pull-out functional unit. Multiple functional units can be installed on a drawer cabinet to form an intelligent motor control center.

[0121] In a single withdrawable functional unit, the current transformer collects the three-phase current of the motor and transmits it to the DSP controller through a high-precision ADC (analog-to-digital converter). The controller identifies motor parameters, performs model predictive control, fault diagnosis methods and other digital processing, and then outputs the control quantity to the inverter to achieve control of the motor. When a fault occurs, appropriate control measures are taken in a timely manner. At the same time, the DSP needs to communicate with the human-machine interface to exchange real-time voltage, current, power, control output and possible fault types, and display corresponding information when the user queries. It uses indicator lights to display key start, stop and fault information. Users can directly control the start and stop of the motor through control buttons, and can also set relevant parameters for various protections, such as the alarm value, trip value, trip delay time and other parameters of the three-phase unbalanced protection. The brief components of a single functional unit are as follows: Figure 5 shown.

[0122] After multiple functional units are installed in a drawer cabinet, the information of each unit needs to be exchanged with the upper management system through the communication interface. The upper management system can realize point-to-point control of each unit through the computer interface, and can also check possible problems in real time through monitoring information, which constitutes a DCS (distributed control system).

[0123] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A novel intelligent motor control method based on model predictive control, characterized in that: The method comprises the following steps: Get the two-phase rotating coordinate system current signal I of the motor d ,I q And the angular velocity signal ω e ; The online parameter identification method based on the extended Kalman filter is used. d ,I q And the angular velocity signal ω e , and the noise covariance matrix Q, the measurement noise covariance matrix R identify the stator resistance R, inductance L and flux ψ of the motor f ; The noise covariance matrix Q and the measurement noise covariance matrix R are adjusted based on the improved artificial jellyfish search algorithm; Based on the motor's stator resistance R, inductance L and flux linkage ψ f The motor parameter model is obtained, and the speed and position control of the motor is realized by using model predictive control based on the motor parameter model.

2. A novel intelligent motor control method based on model predictive control according to claim 1, characterized in that: The specific steps of adjusting the noise covariance matrix Q and the measurement noise covariance matrix R based on the improved artificial jellyfish search algorithm are: (1) Determine the population size N and the maximum number of iterations Maxlt, initialize the population position within the search interval, each individual has d optimization parameters, and set the initial number of iterations; (2) Calculate the fitness values ​​of all individuals. The fitness function calculation formula is as follows: The position X of the optimal value of the population is obtained best And the corresponding fitness value f best ; The fitness value of the individual is the mean square error between the d and q axis current measurement values ​​and the EKF estimated value. The specific formula is: i d (k) and i q (k) is the measured value of the dq axis current at time k, and is the EKF estimated value of the dq axis current at time k; (3) Use the time control function value corresponding to the time control mechanism. If the time control function value is greater than the threshold, the individual follows the ocean current and calculates the individual position. Otherwise, the individual performs type A or type B movement. During the process of the individual performing type A or type B movement, if the random value of 0-1 is greater than or equal to 1-c(t), the individual performs type A movement and calculates the individual position. Otherwise, the individual performs type B movement and calculates the individual position. (4) Based on the pelican surface flight mechanism including adaptive selection and Levy flight strategy, the individual position is updated to obtain a new individual; (5) Calculate the fitness value of the new individual, update the position of the optimal value of the population based on the fitness value of the new individual, and update the current number of iterations; (6) Determine whether the current number of iterations is greater than the maximum number of iterations Maxlt. If so, output the position X of the optimal value of the population best , otherwise return to (3).

3. A novel intelligent motor control method based on model predictive control according to claim 2, characterized in that: The individual position when the individual follows the ocean current is: Among them, X i (t+1) represents the position of the jellyfish of the i-th individual in the t+1-th iteration; X * is the jellyfish with the best position in the current jellyfish group, β is the distribution coefficient, and rand(0,1) is a random value between 0 and 1; is the average position of all jellyfish.

4. A novel intelligent motor control method based on model predictive control according to claim 3, characterized in that: The individual position when the individual performs type A movement is: X i (t+1)=X i (t)+γ*rand(0,1)*(U b -L b ) Among them, U b and L b are the upper and lower bounds of the search interval; γ is the motion coefficient.

5. A novel intelligent motor control method based on model predictive control according to claim 4, characterized in that: The individual position when the individual performs type B movement is: X i (t+1)=X i (t)+S Where S represents the movement step length, S=rand(0,1)*D, and D represents the movement direction, which is: Among them, X j represents the position of the jth individual jellyfish, f(X i ) represents the food content at the location of the ith individual, f(X j ) represents the food content at the location of the jth individual.

6. A novel intelligent motor control method based on model predictive control according to claim 5, characterized in that: The pelican surface flight mechanism based on adaptive selection and Levy flight strategy updates individual positions as follows: in, Where d is the dimension; j∈[1, d] represents the ordinal number of the dimension, t represents the tth iteration, and Z represents X i The update method of the d elements, Q represents the ratio of the remaining number of iterations to the maximum number of iterations, Levy represents the random step length of Levy flight, T is a constant of 0.2, X ij represents the jth element of the i-th individual position, Xbest ij Represents the current optimal value of the i-th element in the i-th individual position.

7. A novel intelligent motor control method based on model predictive control according to claim 6, characterized in that: Levy specifically: Among them, u and v are random variables that obey the normal distribution, η=2*rand.

8. The novel intelligent motor control method based on model predictive control according to claim 7 is characterized in that: The random variables u and v that obey the normal distribution satisfy u~N(0,σ 2 ), v~N(0,1), where σ represents the variance.

9. The novel intelligent motor control method based on model predictive control according to claim 8, characterized in that: The variance σ is: Among them, Γ represents and β is the distribution coefficient.

10. The novel intelligent motor control method based on model predictive control according to claim 1, characterized in that: The method also includes: in the process of realizing speed and position control of the motor by using model predictive control based on the motor parameter model, the motor parameter model outputs a predicted current, and a residual signal is generated after the predicted current is compared with the measured signal, and then a time-frequency domain diagram of the residual signal is obtained by discrete wavelet transform, fault features are extracted by two-dimensional convolution 2D-CNN, and then the motor fault type is obtained by using SVM multi-classifier to realize motor fault diagnosis.

Images