A state estimation method for doubly salient permanent magnet memory motor based on hybrid correlation entropy extended Kalman filter

Abstract

In view of the shortcomings of the prior art, the present invention provides a state estimation method for a double-salient permanent magnet memory motor based on a hybrid correlation entropy extended Kalman filter, which is used to solve the problem of state estimation of the speed and angular position information of the double-salient permanent magnet memory motor. The present invention first gives the dynamic equation of the double-salient permanent magnet memory motor, introduces the kernel size performance index into the concept of the maximum hybrid correlation entropy criterion, and adjusts the kernel size to accurately estimate the difference between the actual value and the estimated value of the speed and angular position controlled by the hybrid correlation entropy, so as to more accurately describe the actual system, thereby improving the filtering performance, convergence speed and accuracy of the algorithm. This method is applied to the control system of the double-salient permanent magnet memory motor. Compared with the existing filtering algorithms, the proposed method has stronger robustness, higher accuracy and estimation ability.

Description

Technical Field

[0001] The present invention relates to the field of electric drive control equipment, and in particular to a state estimation method for a double-salient permanent magnet memory motor using a hybrid correlation entropy extended Kalman filter. Background Art

[0002] As the industrial field presents diversified requirements for motor control systems, the performance requirements are also getting higher and higher. Integrating the concept of memory motor into the design of double-salient pole motors achieves high torque density and online magnetic adjustment function of the motor. It is a new type of double-salient pole permanent magnet memory motor, which is expected to be applied to urban rail transit, high-precision CNC machine tools and electric vehicles. Due to the presence of high-coercivity and low-coercivity permanent magnets in double-salient pole permanent magnet memory motors, and the magnetic adjustment windings are placed in the inner stator, the magnetic adjustment ability, load torque and anti-demagnetization performance at high speed and high temperature are greatly challenged, which directly affects the operating performance and service life of the motor, and also limits its development in the direction of high speed and ultra-high speed.

[0003] The doubly salient permanent magnet memory motor is an important device for realizing electromechanical energy conversion. It is in an important position for energy conversion. When a doubly salient permanent magnet memory motor fails, it is of great significance to quickly identify the speed and angular position information to improve the dynamic performance and robustness of the system. The speed and angular position information of the doubly salient permanent magnet memory motor are prone to vibration and mutation with the change of working conditions, which will affect the sensor installed on the shaft and cause the motor performance to be disturbed. The Kalman filter method is a relatively mature state estimation method. After a large number of literature and patent searches, the methods for state estimation of permanent magnet motors include various Kalman filter extension methods such as extended Kalman filter, square root unscented Kalman filter, adaptive extended Kalman filter, low-order serial double extended Kalman filter, etc. In addition, there are state estimation methods such as central difference filter algorithm, particle filter, high-order sliding mode control, etc.

[0004] After searching relevant patents and literature at home and abroad, although there are many methods for various motor types, no hybrid correlation entropy extended Kalman filter method has been found to be used in double-pole permanent magnet memory motors to achieve accurate estimation of speed and angular position information. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the shortcomings of the prior art, the present invention provides a state estimation method for a double-salient permanent magnet memory motor based on a hybrid correlation entropy extended Kalman filter, which is used to solve the problem of state estimation of the speed and angular position information of the double-salient permanent magnet memory motor. The present invention first gives the dynamic equation of the double-salient permanent magnet memory motor, introduces the kernel size performance index into the concept of the maximum hybrid correlation entropy criterion, and adjusts the kernel size to accurately estimate the difference between the actual value and the estimated value of the speed and angular position controlled by the hybrid correlation entropy, so as to more accurately describe the actual system, thereby improving the filtering performance, convergence speed and accuracy of the algorithm. This method is applied to the control system of the double-salient permanent magnet memory motor. Compared with the existing filtering algorithms, the proposed method has stronger robustness, higher accuracy and estimation ability.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented by the following technical scheme: a state estimation method for a double-salient permanent magnet memory motor using a hybrid correlation entropy extended Kalman filter, comprising the following steps:

[0009] First, the dynamic equation of the doubly salient permanent magnet memory motor can be expressed as,

[0010]

[0011] In formula (1), Δp1 and Δp2 are the sum of the model uncertainty and the external disturbance of the doubly salient permanent magnet memory motor control system, respectively, x1 is the measurable state vector, and x2 is the measurable time vector. f1(t,x1) is a nonlinear vector unknown function containing motor position information and time information, G1 is a designed positive constant, and t is the time quantity.

[0012] The hybrid correlation entropy extended Kalman filter method has a high dependence on the weights of the resistance and inductance parameters in the double-salient permanent magnet memory motor control system, so generally only the inductance and resistance need to be considered as state vectors. The hybrid correlation entropy extended Kalman filter method is specifically developed as follows:

[0013] The dynamic equation of the doubly salient permanent magnet memory motor in equation (1) can be converted into a discrete equation of a state model and an observation model, which can be specifically expressed as:

[0014] x k+1 =f(x k )+w k (2)

[0015] z k =h(x k )+v k (3)

[0016] In formula (2-3), f(xk ) are the doubly salient permanent magnet memory motor system matrices, x k is the state vector at time k, which contains the speed information, z k is the measurement function of the motor when it is running at time k, h(x k ) is the motor system observation matrix, w k With v k They represent the process noise and measurement noise during the operation of the doubly salient permanent magnet memory motor at time k respectively.

[0017] The i-th kernel set σ i It is an important performance indicator of hybrid correlation entropy, which determines the random variable estimation range of the speed and angular position of the double-salient permanent magnet memory motor. It is mainly used to reduce the influence of the parameter changes of the double-salient permanent magnet memory motor on the estimation accuracy of the speed and angular position information, and reduce the convergence speed and steady-state error performance. By adjusting the i-th kernel set σ i , to achieve the degree of similarity between the actual and estimated values ​​of the speed and angular position controlled by the hybrid correlation entropy. The hybrid correlation entropy extended Kalman filter method can be used to minimize the objective function J k To reduce the i-th kernel set σ i The purpose of

[0018]

[0019] In formula (4), is the a posteriori estimate of the speed state variable, ∥·∥ 2 is the L2 norm of the vector, ∥·∥ A represents the weighted norm of the vector, m is an integer, μ is the step size, z ik is the i-th measurement function when the motor is running at time k.

[0020] The mixed correlation entropy represents the generalized function between the two random variables of speed and corner position. There are multiple sets of data in the kernel set σ that can represent the kernel size, speed, and corner position. For the convenience of calculation, let the kernel size, speed, and corner position of the i-th kernel set σ i :

[0021]

[0022] In formula (5), i = 1, 2, ..., n, n is an integer, and r ik is the cost function. After expanding the cost function using Taylor's formula, it contains the high-order statistical error e i,D (k) In the hybrid correlation entropy extended Kalman filter, fast convergence and lower steady-state error are achieved through the characteristics of high-order statistical error, which further shows that the constructed cost function is robust to control system disturbances. i(k) Impulse response to be identified, κ σ is the kernel parameter associated with the kernel size σ. Equation (5) is related to the cost function of the hybrid correlation entropy extended Kalman filter method. Hybrid correlation entropy has a significant advantage over the maximum correlation entropy of a single kernel. In order to quickly extract the speed and angular position information of the doubly salient permanent magnet memory motor and improve the anti-interference ability, the kernel size σ is used as a measured characteristic parameter, and the kernel size is re-evaluated each time the measurement is made. During the measurement process, when the speed and angular position information are subject to possible interference, the kernel size σ will be abnormal.

[0023] The speed state variable prior estimation and angular position prior estimation of the doubly salient permanent magnet memory motor in formula (6-7) require the i-th kernel set σ of formula (4-5) to be i Substitution can be obtained.

[0024] In summary, the prior estimation of the speed state variable is given first: and the prior estimate of the corner position P k∣k-1 :

[0025]

[0026]

[0027] In formula (6-7), σ1 is the first kernel set σ i , σ2 is the second kernel set σ i , Φ k and Q k are the trust levels of the rotation speed and the angular position information, P k-1 is the cross-covariance matrix, λ is the cross-covariance matrix coefficient, is Φ k The transposed matrix of the speed state variable is estimated a posteriori based on the a priori estimation formula (6-7). and the posterior estimate of the corner position P k∣k The update of the parameters is realized when the next sampling period arrives, and the parameters are updated to realize the online correction of the motor speed and angular position parameters. The update formula is as follows:

[0028]

[0029]

[0030] In formula (8-9), H i represents the i-th row of the non-negative definite matrix H, R k is a positive definite matrix, K k is the gain matrix, Expressed as a weighted center matrix, (·) T represents the transposed matrix, Szz∣k-1 is the square root of the innovation error covariance matrix, and Tria(·) represents the decomposed upper triangular matrix.

[0031]

[0032]

[0033] When the i-th measurement of the speed and angular position parameters of the doubly salient permanent magnet memory motor is subject to significant interference, the i-th kernel set σ in equation (10) is adjusted. i Then, the diagonal matrix C k Substitute the gain matrix K into equation (11): k , so that the error between the estimated and actual speed and angular position is minimized, thereby completing the online state estimation of the doubly salient permanent magnet memory motor based on the hybrid correlation entropy extended Kalman filter.

[0034] (III) Beneficial effects

[0035] The present invention provides a state estimation method for a double-salient permanent magnet memory motor using a hybrid correlation entropy extended Kalman filter, which has the following beneficial effects:

[0036] 1. For the input signal height of the speed and angular position information of the double-salient permanent magnet memory motor control system, the convergence speed of the Kalman filter method alone is slow, which affects the state estimation. The proposed hybrid correlation entropy extended Kalman filter method has fast calculation speed and strong system robustness, and can significantly improve the online state estimation capability of the speed and angular position information of the double-salient permanent magnet memory motor.

[0037] 2. The hybrid correlation entropy is combined with the extended Kalman filter algorithm to construct a new cost function. By selecting and optimizing the hybrid parameters and kernel size in the cost function as the global measurement standard, the requirements for speed and angular position information identification can be reduced in the case of pulse interference, achieving better dynamic performance and filtering accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further explained below in conjunction with specific implementation modes. It should be understood that the following specific implementation modes are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0040] As shown in the figure, the present invention provides a state estimation method for a double-salient permanent magnet memory motor using a hybrid correlation entropy extended Kalman filter, comprising the following steps:

[0041] First, the dynamic equation of the doubly salient permanent magnet memory motor can be expressed as,

[0042]

[0043] In formula (1), Δp1 and Δp2 are the sum of the model uncertainty and the external disturbance of the doubly salient permanent magnet memory motor control system, respectively, x1 is the measurable state vector, and x2 is the measurable time vector. f1(t,x1) is a nonlinear vector unknown function containing motor position information and time information, G1 is a designed positive constant, and t is the time quantity.

[0044] The hybrid correlation entropy extended Kalman filter method has a high dependence on the weights of the resistance and inductance parameters in the double-salient permanent magnet memory motor control system, so generally only the inductance and resistance need to be considered as state vectors. The hybrid correlation entropy extended Kalman filter method is specifically developed as follows:

[0045] The dynamic equation of the doubly salient permanent magnet memory motor in equation (1) can be converted into a discrete equation of a state model and an observation model, which can be specifically expressed as:

[0046] x k+1 =f(x k )+w k (2)

[0047] z k =h(x k )+v k (3)

[0048] In formula (2-3), f(x k ) is the doubly salient permanent magnet memory motor system matrix, x k is the state vector at time k, which contains the speed information, z k is the measurement function of the motor when it is running at time k, h(x k ) is the motor system observation matrix, w k With v k They represent the process noise and measurement noise during the operation of the doubly salient permanent magnet memory motor at time k respectively.

[0049] The i-th kernel set σ i It is an important performance indicator of hybrid correlation entropy, which determines the random variable estimation range of the speed and angular position of the double-salient permanent magnet memory motor. It is mainly used to reduce the influence of the parameter changes of the double-salient permanent magnet memory motor on the estimation accuracy of the speed and angular position information, and reduce the convergence speed and steady-state error performance. By adjusting the i-th kernel set σ i , to achieve the degree of similarity between the actual and estimated values ​​of the speed and angular position controlled by the hybrid correlation entropy. The hybrid correlation entropy extended Kalman filter method can be used to minimize the objective function J k To reduce the i-th kernel set σi The purpose of

[0050]

[0051] In formula (4), is the a posteriori estimate of the speed state variable, ∥·∥ 2 is the L2 norm of the vector, ∥·∥ A represents the weighted norm of the vector, m is an integer, μ is the step size, z ik is the i-th measurement function when the motor is running at time k.

[0052] The mixed correlation entropy represents the generalized function between the two random variables of speed and corner position. There are multiple sets of data in the kernel set σ that can represent the kernel size, speed, and corner position. For the convenience of calculation, let the kernel size, speed, and corner position of the i-th kernel set σ i :

[0053]

[0054] In formula (5), i = 1, 2, ..., n, n is an integer, and r ik is the cost function. After expanding the cost function using Taylor's formula, it contains the high-order statistical error e i,D (k) In the hybrid correlation entropy extended Kalman filter, fast convergence and lower steady-state error are achieved through the characteristics of high-order statistical error, which further shows that the constructed cost function is robust to control system disturbances. i (k) Impulse response to be identified, κ σ is the kernel parameter associated with the kernel size σ. Equation (5) is related to the cost function of the hybrid correlation entropy extended Kalman filter method. Hybrid correlation entropy has a significant advantage over the maximum correlation entropy of a single kernel. In order to quickly extract the speed and angular position information of the doubly salient permanent magnet memory motor and improve the anti-interference ability, the kernel size σ is used as a measured characteristic parameter, and the kernel size is re-evaluated each time the measurement is made. During the measurement process, when the speed and angular position information are subject to possible interference, the kernel size σ will be abnormal.

[0055] The speed state variable prior estimation and angular position prior estimation of the doubly salient permanent magnet memory motor in formula (6-7) require the i-th kernel set σ of formula (4-5) to be i Substitution can be obtained.

[0056] In summary, the prior estimation of the speed state variable is given first: and the prior estimate of the corner position P k∣k-1 :

[0057]

[0058]

[0059] In formula (6-7), σ1 is the first kernel set σ i , σ2 is the second kernel set σ i , Φ k and Q k are the trust levels of the rotation speed and the angular position information, P k-1 is the cross-covariance matrix, λ is the cross-covariance matrix coefficient, is Φ k The transposed matrix of the speed state variable is estimated a posteriori based on the a priori estimation formula (6-7). and the posterior estimate of the corner position P k∣k The update of the parameters is realized when the next sampling period arrives, and the parameters are updated to realize the online correction of the motor speed and angular position parameters. The update formula is as follows:

[0060]

[0061]

[0062] In formula (8-9), H i represents the i-th row of the non-negative definite matrix H, R k is a positive definite matrix, K k is the gain matrix, Expressed as a weighted center matrix, (·) T represents the transposed matrix, S zz∣k-1 is the square root of the innovation error covariance matrix, and Tria(·) represents the decomposed upper triangular matrix.

[0063]

[0064]

[0065] When the i-th measurement of the speed and angular position parameters of the doubly salient permanent magnet memory motor is subject to significant interference, the i-th kernel set σ in equation (10) is adjusted. i Then, the diagonal matrix C k Substitute the gain matrix K into equation (11): k , so that the error between the estimated and actual speed and angular position is minimized, thereby completing the online state estimation of the doubly salient permanent magnet memory motor based on the hybrid correlation entropy extended Kalman filter.

[0066] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode, but also include technical schemes composed of any combination of the above-mentioned technical features. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also regarded as the protection scope of the present invention.

Claims

1. A state estimation method for a doubly salient permanent magnet memory motor based on a hybrid correlation entropy extended Kalman filter, characterized in that: include: Firstly, the dynamic equation of the doubly salient permanent magnet memory motor is given. Then, the kernel size performance index is introduced into the maximum hybrid correlation entropy criterion. By adjusting the kernel size, the difference between the actual and estimated values ​​of the speed and angular position controlled by the hybrid correlation entropy is accurately estimated, and the actual doubly salient permanent magnet memory motor system is accurately described. The dynamic equation of the double-salient permanent magnet memory motor is expressed as: Among them, Δp1 and Δp2 are the sum of the model uncertainty and the external disturbance of the doubly salient permanent magnet memory motor control system, x1 is the measurable state vector, x2 is the measurable time vector; f1(t,x1) is a nonlinear vector unknown function containing motor position information and time information, and t is the time quantity; Only inductance and resistance are considered as state vectors; the method of hybrid correlation entropy extended Kalman filter is as follows: Convert equation (1) into the discrete equations of the state model and observation model, which can be expressed as follows: x k+1 =f(x k )+w k (2) z k =h(x k )+v k (3) Among them, f(x k ) is the doubly salient permanent magnet memory motor system matrix, x k is the state vector at time k, which contains the speed information, z k is the measurement function of the motor when it is running at time k, h(x k ) is the motor system observation matrix, w k With v k They represent the process noise and measurement noise during the operation of the doubly salient permanent magnet memory motor at time k respectively; The i-th kernel set σ i It is an important performance indicator of hybrid correlation entropy, which determines the random variable estimation range of the speed and angular position of the double-salient permanent magnet memory motor. It is used to reduce the influence of the parameter changes of the double-salient permanent magnet memory motor on the estimation accuracy of the speed and angular position information, and reduce the convergence speed and steady-state error performance. By adjusting the i-th kernel set σ i , to achieve the degree of similarity between the actual value and the estimated value of the speed and angular position controlled by the hybrid correlation entropy; by minimizing the objective function J of the hybrid correlation entropy extended Kalman filter method k To reduce the i-th kernel set σ i The purpose of in, is the a posteriori estimate of the speed state variable, ∥·∥ 2 is the L2 norm of the vector, ∥·∥ A represents the weighted norm of the vector, m is an integer, μ is the step size, z ik is the ith measurement function of the motor when it is running at time k, is the prior estimate of the speed state variable.

2. The state estimation method of a doubly salient permanent magnet memory motor using a hybrid correlation entropy extended Kalman filter according to claim 1 is characterized in that: Mixed correlation entropy represents a generalized function between two random variables, rotation speed and angular position; the kernel set σ has multiple sets of data that can represent the kernel size, rotation speed, and angular position. Let the kernel size, rotation speed, and angular position of the i-th kernel set σ i : Where i = 1, 2, ..., n, n is an integer, r ik is the cost function. After expanding the cost function using Taylor's formula, it contains the high-order statistical error e i,D (k) In the hybrid correlation entropy extended Kalman filter method, fast convergence and lower steady-state error are achieved through the characteristics of high-order statistical error, which further shows that the constructed cost function is robust to control system disturbances; u i (k) Impulse response to be identified, κ σ is a kernel parameter associated with the kernel size; Formula (5) is related to the cost function of the hybrid correlation entropy extended Kalman filter method; the kernel size is used as a measured characteristic parameter, and the kernel size is re-evaluated for each measurement; during the measurement process, when the rotation speed and angular position information are subject to possible interference, the kernel size will be abnormal.

3. The state estimation method of a doubly salient permanent magnet memory motor using a hybrid correlation entropy extended Kalman filter according to claim 2 is characterized in that: The speed state variable prior estimation and angular position prior estimation of the doubly salient permanent magnet memory motor in formulas (6) and (7) are obtained by substituting the i-th kernel set σ of formulas (4) and (5) into i Substitute it into the result, which is: First, a priori estimate of the speed state variable is given and the prior estimate of the corner position P k∣k-1 : Among them, σ1 is the first kernel set σ i , σ2 is the second kernel set σ i , Φ k and Q k are the trust levels of the rotation speed and the angular position information, P k-1 is the cross-covariance matrix, λ is the cross-covariance matrix coefficient, is Φ k The transposed matrix of the speed state variable is estimated a posteriori based on the prior estimation formulas (6) and (7). and the posterior estimate of the corner position P k∣k The update of the parameters is realized when the next sampling period arrives, and the parameters are updated to realize the online correction of the motor speed and angular position parameters. The update formula is as follows: Among them, H i represents the i-th row of the non-negative definite matrix H, R k is a positive definite matrix, K k is the gain matrix, Expressed as a weighted center matrix, (·) T represents the transposed matrix, S zz∣k-1 is the square root of the information error covariance matrix, Tria(·) represents the decomposed upper triangular matrix; When the i-th measurement of the speed and angular position parameters of the doubly salient permanent magnet memory motor is subject to significant interference, the i-th kernel set σ in equation (10) is adjusted. i Then, the diagonal matrix C k Substitute the gain matrix K into equation (11): k , so that the error between the estimated and actual speed and angular position is minimized, thereby completing the online state estimation of the doubly salient permanent magnet memory motor based on the hybrid correlation entropy extended Kalman filter.

Images