Permanent magnet motor rotor permanent magnet fault prediction and health management method based on all-parameter estimation

Abstract

The invention discloses a permanent magnet motor rotor permanent magnet fault prediction and health management method based on all-parameter estimation, and provides a permanent magnet fault prediction and health management method based on all-parameter estimation for an irreversible demagnetization phenomenon of a permanent magnet motor under an extreme working condition. Technical support is provided for safety and reliability of a permanent magnet motor driving system. In the proposed scheme, a sinusoidal current signal is injected into a straight axis to obtain multiple groups of operation states of the motor, so that a parameter estimation equation reaches a full-rank condition. Furthermore, an inductance monitor is designed based on a dynamic forgetting factor recursive least square method so as to obtain accurate rectangular axis inductance parameters on line. Secondly, acquiring resistance and permanent magnet flux linkage data of the permanent magnet motor in real time by adopting double H-infinity filters, and transmitting estimation data between the two algorithms to eliminate permanent magnet flux linkage estimation deviation caused by parameter mismatch; and finally, based on the finally obtained permanent magnet flux linkage parameters, accurate evaluation and demagnetization fault diagnosis of the residual magnetism level of the rotor permanent magnet are expected to be realized.

Description

Technical Field

[0001] The present invention belongs to the field of motor fault prediction and health management, and particularly relates to a method for permanent magnet motor rotor permanent magnet fault prediction and health management based on full parameter estimation. Background Art

[0002] Permanent magnet motors have received extensive attention in transportation fields such as rail transit, electric vehicles, and aerospace due to their advantages of wide speed regulation range, high operating efficiency, and low maintenance cost. However, when the permanent magnet electric drive system is under high load or in the acceleration process, heat sources such as switching losses generated by power electronic devices, Joule losses generated by current flowing through winding resistors, and eddy current and hysteresis losses generated by electromagnetic effects will cause the internal temperature of the motor to rise rapidly. At the same time, limited by the installation space, the system has poor heat dissipation conditions, which is extremely likely to cause thermal anomaly problems, and then lead to irreversible demagnetization faults of the permanent magnet of the permanent magnet motor rotor. In addition, irreversible demagnetization faults in a high-temperature environment will cause perturbations in magnetic flux, inductance, and resistance parameters in the motor mathematical model, further deteriorating the stability and control accuracy of the system. If such faults are not detected and processed in a timely and effective manner in the early stage, it may affect the normal operation of other components in the new energy vehicle power system, and then cause a series of chain destruction reactions, and even threaten personal safety in severe cases. Therefore, real-time management and monitoring of the health status of the rotor permanent magnet is of great significance for ensuring the safety and reliability of the system.

[0003] Model-based diagnostic methods do not require additional sensor devices and can intuitively observe the health status of permanent magnets through magnetic flux values, which have great value in occasions with strict requirements for safety and stability. However, existing research shows that in the fault prediction and health management of rotor permanent magnets, relying solely on single magnetic flux parameter estimation cannot achieve high-precision monitoring of magnet health. It is necessary to comprehensively consider the coupling relationship between various motor parameters to effectively improve the accuracy of magnetic flux estimation.

[0004] In view of the above problems, the present invention proposes a method for permanent magnet motor rotor permanent magnet fault prediction and health management based on full parameter estimation. Summary of the Invention

[0005] To overcome the deficiencies of the prior art, the present invention provides a method for permanent magnet motor rotor permanent magnet fault prediction and health management based on full parameter estimation. First, the method injects a sinusoidal perturbation current into the direct axis of the permanent magnet motor to obtain multiple sets of operating states of the motor, so as to solve the under-rank problem in the parameter estimation process and will not significantly affect the normal operating state of the motor. Further, an inductance estimator is designed in combination with the recursive least squares method with a dynamic forgetting factor to obtain accurate direct and quadrature axis inductance parameters online. Secondly, a double H ∞A filter estimator is used to obtain the resistance and permanent magnet flux data of the permanent magnet motor in real time. Finally, the calculation results between the other two algorithm estimators are transmitted to each other to eliminate the parameter mismatch caused by the demagnetization effect, and to achieve accurate estimation of the remaining magnetic field strength of the permanent magnet and demagnetization fault diagnosis.

[0006] The technical solution adopted in the present invention is a non-invasive fault prediction and health management method for the permanent magnet rotor of a permanent magnet motor based on full parameter estimation. The specific content is as follows:

[0007] Step 1): Mathematical model of a permanent magnet motor considering demagnetization effect. The voltage and flux linkage equations of a permanent magnet motor can be expressed as

[0008]

[0009] ψ d =L d i d +ψ m

[0010] ψ q =L q i q

[0011] In the formula, u d , u q , i d , i q respectively represent the stator voltages and currents of the d-q axes of the motor, ω e is the electrical angular velocity, L d , L q , R s , ψ m are the actual measured values of the stator inductances, stator resistance and permanent magnet flux linkage of the d-q axes respectively, ψ d and ψ q are the actual measured values of the stator flux linkages of the d-q axes respectively.

[0012] According to the above formula, the voltage equation is written in the following form with the parameters to be estimated separated

[0013]

[0014] In the formula, X is the vector of parameters to be estimated, and its coefficient matrix is A.

[0015] As can be seen from the above formula, the unique solution of X cannot be directly obtained from the external measurement values of the motor. At this time, grouped identification is an effective way to solve multi-parameter estimation. At the same time, if the motor operates at i d= 0 control mode, the amplitude component of the d-axis inductance will not be identified, and the estimated value of the d-axis inductance may deviate seriously from the nominal value. At the same time, in order to meet the torque output capacity at low residual magnetism levels, the stator current is required to compensate, which leads to significant thermal effects during the demagnetization process of the permanent magnet motor, causing the overall parameters of the motor to be perturbed.

[0016] Specifically, the linear thermal model of permanent magnet flux and stator resistance is

[0017] ψ m =ψ m0 [1+ρ 1 (T-20)]

[0018] R s =R s0 [1+ρ 2 (T-20)]

[0019] In the formula, ψ m0 and R s0 are the nominal values ​​of permanent magnet flux and stator resistance, respectively, m and R s are the actual measured values ​​of permanent magnet flux and stator resistance, respectively, 1 is the thermal coefficient of permanent magnet flux, ρ 2 represents the stator winding resistivity, and T represents the motor operating temperature.

[0020] For the stator inductance, according to the finite element theory, ignoring the winding resistance voltage drop, the stator inductance can be calculated by the following formula

[0021]

[0022] Where, L s and L ef Represent the stator inductance and core length respectively, i s is the stator current, H l , B l and ΔS l They respectively represent the magnetic field intensity, remanent magnetic density and calculation unit area corresponding to the th finite element mesh division.

[0023] After the demagnetization fault, the faulty permanent magnet and the air gap magnetic field at the corresponding position will be weakened to varying degrees, resulting in uneven distribution of the air gap magnetic field. In addition, since the magnetic flux density and magnetic resistance are linearly related, the change in magnetic flux density will cause magnetic resistance fluctuations, which in turn causes the average value of inductance to change. Introducing the disturbance caused by demagnetization into the three-phase inductance matrix and ignoring the high-order harmonics and leakage inductance of the inductance, we can get

[0024]

[0025] Where, La , L b and L c respectively represent the actual values of the three-phase self-inductance after the demagnetization fault, M ab , M bc and M ca respectively represent the actual values of the three-phase mutual inductance after the demagnetization fault. L dc = L dc0 +ΔL dc and L ac = L ac0 +ΔL ac are respectively the actual values of the DC component and the AC component of the self-inductance after the demagnetization fault, L dc0 and L ac0 are respectively the nominal values of the DC component and the AC component of the self-inductance of the healthy motor, ΔL dc and ΔL ac are respectively the perturbation values of the DC component and the AC component of the self-inductance introduced after the demagnetization fault, θ e is the electrical angle of the motor.

[0026] From the above formula, it can be obtained that

[0027]

[0028] To sum up, after the demagnetization phenomenon occurs, under the influence of electromagnetic induction and thermal effects, the numerical values of the stator resistance, inductance and rotor permanent magnet flux linkage of the motor will be disturbed simultaneously. More importantly, from the above analysis, restricted by the motor mathematical model, the real-time identification degree of freedom of the full motor parameters identification is limited. That is to say, the inaccuracy of any parameter will affect the final estimation accuracy of the permanent magnet flux linkage, which may cause misdiagnosis of the demagnetization fault, thus increasing the difficulty of implementing the model-based magnet fault prediction and health management scheme.

[0029] Step 2): High-precision inductance parameter identification. In order to achieve accurate estimation of the d-q axis inductance, the present invention designs an inductance estimation method combining d-axis sine current signal injection and the least square method. In the designed method, multiple groups of operating data of the motor are obtained by injecting a sine current signal into the d axis of the motor to achieve full-rank solution in the inductance estimation process. Further, combined with the dynamic forgetting factor recursive least square method, the d-q axis inductance parameters are obtained in real time online.

[0030] Discretize the voltage equation of the permanent magnet motor in step 1) and rewrite it in the least square form

[0031]

[0032] In the formula,

[0033]

[0034] where \(k\) is the \(k\)th sample data, \(T\) s is the control period of the system, and the superscript “^” represents the estimated value.

[0035] The discretized residual value \(\varepsilon\) k \(= [\varepsilon\) 1，k \(\varepsilon\) 2，k \) T can be expressed as

[0036]

[0037] The dynamic forgetting factor \(\gamma\) k is designed as

[0038]

[0039] where \(\alpha\) is a positive adjustable factor close to and less than 1, and \(\beta\) is a positive adjustable factor, whose value is usually greater than \(\alpha\). According to the above formula, under the combined action of \(\alpha\) and \(\beta\), the forgetting factor \(\gamma\) k can be adjusted dynamically in real time. When the residual value \(\varepsilon\) k is large, the forgetting factor \(\gamma\) k gradually approaches \(\alpha\), thereby improving the response speed in the parameter estimation process. When the residual value \(\varepsilon\) k is small, the forgetting factor \(\gamma\) k approaches 1 infinitely, to enhance the anti-noise ability in the parameter estimation process and reduce the steady-state error.

[0040] The iterative formula of the dynamic forgetting factor recursive least squares method can be expressed as follows

[0041]

[0042] where \(W\) 0 generally needs to be determined in advance. Usually, \(W\) 0 \(= LI\), \(I\) is the identity matrix, and \(L\) is a sufficiently large positive real number, with a value range between 10 4 ~10 6 .

[0043] Step 3): Design of the permanent magnet monitor. The designed dual H ∞ filter contains two interconnected and coupled filters: one filter is used to estimate the current dynamics, and the other filter is used to estimate the motor parameters.

[0044] The discrete nonlinear state space equation of the dual H ∞ filter for parameter estimation can be expressed as \(x\) k \(= d(x\) k-1 , \(p\) k-1 , \(u\) k-1 ) + w\) x,k-1

[0045] p k = p k-1 + w p,k-1

[0046] y k = Y(x k , p k ) + v k

[0047] wherein, x k = [i d，k i q，k T , u k = [u d，k u q,k T and y k = [i d,k i q，k T are the system state vector, input vector, and output vector respectively, and p k = [R s，k ψ m，k T is the motor parameter to be estimated. w x,k-1 and w p,k-1 are the process noises of the system state and parameters respectively, representing the uncertainty of the system, and v k is the measurement noise, representing the measurement error in the measurement process of the current sensor. d(x k-1 , p k-1 , u k-1 ) and Y(x k , p k ) are nonlinear functions.

[0048] The calculation steps of the dual H ∞ filter are as follows, where the superscripts "+" and "-" are the correction value and prediction value of the dual H ∞ filter.

[0049] (1) Parameter prediction

[0050]

[0051] (2) State prediction

[0052]

[0053] (3) State correction

[0054]

[0055] (4) Parameter correction

[0056] ​​​​

[0057] Double H ∞ In the filter steps (1) to (4),

[0058]

[0059] wherein,

[0060]

[0061] At the initial stage of algorithm startup, K x，k-1 and are respectively initialized as 2×2 all-zero matrices. In the above formula, Q x,k and Q p，k are respectively the covariance matrices corresponding to the system state and parameters, and R k is the covariance matrix corresponding to the measurement noise. S x，k and S p，k are respectively the weight matrices of the state vector x and the parameter p. θ x and θ p are the performance constraint boundaries related to the system model accuracy, and θ x and θ p can be set as positive numbers close to zero to improve the accuracy of the state space model. D x，p，k-1 is the Jacobian matrix of the function with respect to the state vector x, and Y x，k and Y p，k are respectively the Jacobian matrices of the output function Y(x k , p k ) with respect to the state vector x and the parameter p. P x,k and P p,k are respectively the error covariance matrices of the state and parameters. K x,k and K p，k are respectively the gain matrices of the state and parameters.

[0062] Step 4): Demagnetization fault diagnosis. In the present invention, first, a sinusoidal current signal is injected into the d-axis to obtain multiple groups of operating states of the permanent magnet motor, and the electrical angular velocity, d-q axis voltage and current signals during the operation of the permanent magnet motor are collected. Secondly, the collected signal data is input into the permanent magnet monitor designed by the present invention, and the d-q axis inductance estimation result obtained by the recursive least squares method with a mutual transfer dynamic forgetting factor and the resistance and flux linkage estimation results of the double H ∞ filter are mutually transferred to eliminate the flux linkage estimation deviation caused by parameter mismatch after the demagnetization fault, so as to obtain the final permanent magnet flux linkage estimation result. Finally, combining the final flux linkage estimation result, the fault diagnosis is completed by setting a reasonable threshold.

[0063] Compared with the existing technical solutions, the beneficial effects brought by the technical solution of the present invention are as follows:

[0064] In view of the irreversible demagnetization phenomenon of permanent magnet motors under extreme working conditions, the present invention proposes a method for permanent magnet motor rotor permanent magnet fault prediction and health management based on full parameter estimation, and obtains the following conclusions:

[0065] (1) By combining signal injection and recursive least squares method with a dynamic forgetting factor, on the basis of solving the inherent underdetermined rank problem of the voltage equation, accurate estimation of the d-q axis inductance is achieved;

[0066] (2) Adopt a dual-H ∞ filter to estimate the resistance and flux linkage parameters in real time, and make the estimation results between it and the recursive least squares method with a dynamic forgetting factor be transmitted to each other, eliminating the problem of the decrease in the estimation accuracy of the average residual magnetic field strength of the permanent magnet caused by parameter mismatch. Description of the Drawings

[0067] Figure 1 is the structure diagram of the permanent magnet monitor of the present invention;

[0068] Figure 2 is the flowchart of the demagnetization fault diagnosis of the present invention;

[0069] Figure 3 is the control block diagram of the present invention; Detailed Embodiment

[0070] The following makes a detailed description of a method for permanent magnet motor rotor permanent magnet fault prediction and health management based on full parameter estimation of the present invention in combination with embodiments and drawings.

[0071] The method of the present invention includes the following steps:

[0072] Step 1): The mathematical model of the permanent magnet motor considering the demagnetization effect. The voltage and flux linkage equations of the permanent magnet motor can be expressed as

[0073]

[0074] ψ d = L d i d + ψ m

[0075] ψ q = L q i q

[0076] In the formula, u d , u q , i d , i q respectively represent the d-q axis stator voltages and currents of the motor, ωe is the electrical angular velocity, L d , L q , R s , ψ m are the actually measured values of the d-q axis stator inductance, stator resistance, and permanent magnet flux linkage respectively. ψ d and ψ q are the actually measured values of the d-q axis stator flux linkage respectively.

[0077] According to Equation (1), the voltage equation is written in the following form with the parameters to be estimated separated

[0078]

[0079] where X is the vector of parameters to be estimated, and its coefficient matrix is A.

[0080] It can be seen from the above equation that the unique solution of X cannot be directly obtained from the external measurement values of the motor. At this time, grouped identification is an effective way to solve multi-parameter estimation. At the same time, if the motor operates in the i d = 0 control mode, the amplitude component of the d-axis inductance cannot be identified. At this time, the estimated value of the d-axis inductance may deviate significantly from the nominal value. At the same time, to meet the torque output ability under the low residual magnetism level, the stator current needs to be compensated, resulting in a significant thermal effect during the demagnetization process of the permanent magnet motor, causing the overall parameters of the motor to perturb.

[0081] Specifically, the linear thermal models of the permanent magnet flux linkage and the stator resistance are

[0082] ψ m = ψ m0 [1 + ρ 1 (T - 20)] (3)

[0083] R s = R s0 [1 + ρ 2 (T - 20)] (4)

[0084] where ψ m0 and R s0 are the nominal values of the permanent magnet flux linkage and the stator resistance respectively, ρ 1 is the thermal coefficient of the permanent magnet flux linkage, ρ 2 represents the resistivity of the stator winding, and T represents the operating temperature of the motor.

[0085] For the stator inductance, according to the finite element theory, ignoring the winding resistance voltage drop, the stator inductance can be calculated by the following formula

[0086]

[0087] where L s and Lef represent the stator inductance and the core length respectively, and i s is the stator current, and H l , B l and ΔS l represent the magnetic field intensity, the remanent magnetic density, and the calculation cell area corresponding to the j-th finite element mesh division respectively.

[0088] After the demagnetization fault, the fault permanent magnet and the air-gap magnetic field at the corresponding position will be weakened to varying degrees, resulting in uneven distribution of the air-gap magnetic field. In addition, since the magnetic flux density and the magnetic resistance are linearly related, the change in the magnetic flux density will cause fluctuations in the magnetic resistance, and then cause a change in the average value of the inductance. Introduce the disturbance caused by demagnetization into the three-phase inductance matrix, and ignore the higher harmonics and leakage inductance of the inductance to obtain

[0089]

[0090] In the formula, L a , L b and L c represent the actual values of the three-phase self-inductances after the demagnetization fault respectively, and M ab , M bc and M ca represent the actual values of the three-phase mutual inductances after the demagnetization fault respectively. L dc = L dc0 + ΔL dc and L ac = L ac0 + ΔL ac are the actual values of the DC component and the AC component of the self-inductance after the demagnetization fault respectively, L dc0 and L ac0 are the nominal values of the DC component and the AC component of the self-inductance of the healthy motor respectively, and ΔL dc and ΔL ac are the disturbance values of the DC component and the AC component of the self-inductance introduced after the demagnetization fault respectively, and θ e is the electrical angle of the motor.

[0091] It can be obtained from Equation (6) that

[0092]

[0093] To sum up, after the demagnetization phenomenon occurs, under the influence of electromagnetic induction and thermal effects, the numerical values of the stator resistance, inductance, and rotor permanent magnet flux linkage of the motor will be disturbed simultaneously. More importantly, from the above analysis, restricted by the mathematical model of the motor, the real-time identification degree of freedom of the full-parameter identification of the motor is limited. That is to say, the inaccuracy of any parameter will affect the final estimation accuracy of the permanent magnet flux linkage, and may cause misdiagnosis of the demagnetization fault, thus increasing the difficulty of implementing the model-based magnet fault prediction and health management scheme.

[0094] Step 2: High-precision inductance parameter identification. To achieve accurate estimation of the d-q axis inductance, the present invention designs an inductance estimation method combining d-axis sinusoidal current signal injection with the least squares method. In the designed method, multiple sets of operating data of the motor are obtained by injecting a sinusoidal current signal into the d-axis of the motor to achieve a full-rank solution in the inductance estimation process. Further, in combination with the recursive least squares method with a dynamic forgetting factor, the d-q axis inductance parameters are obtained in real-time online.

[0095] Discretize the voltage equation of the permanent magnet motor in Step 1 and rewrite it in the least squares form

[0096]

[0097] where

[0098]

[0099] where k is the kth sample data, T s is the control period of the system, and the superscript "^" represents the estimated value.

[0100] The discretized residual value ε k =[ε 1,k ε 2，k T can be expressed as

[0101]

[0102] Design the dynamic forgetting factor γ k as

[0103]

[0104] where α is a positive adjustable factor close to and less than 1, and β is a positive adjustable factor, whose value is usually greater than α. According to the above formula, under the combined action of α and β, the forgetting factor γ k can be adjusted dynamically in real-time. When the residual value ε k is large, the forgetting factor γ k gradually approaches α, thereby improving the response speed in the parameter estimation process. When the residual value ε k is small, the forgetting factor γ k approaches 1 infinitely to enhance the anti-noise ability in the parameter estimation process and reduce the steady-state error.

[0105] The iterative formula of the recursive least squares method with a dynamic forgetting factor can be expressed as follows

[0106]

[0107] where W 0 ​Generally, it needs to be determined in advance. Usually, W is defined as 0 = LI, where I is the identity matrix and L is a sufficiently large positive real number, with a value range between 10 4 ~10 6 .

[0108] Step 3): Design of the permanent magnet monitor. The designed dual-H ∞ filter contains two interconnected and coupled filters: one filter is used to estimate the current dynamics, and the other filter is used to estimate the motor parameters. The specific structure of the permanent magnet monitor is as shown in Figure 1 .

[0109] The discrete nonlinear state-space equation of the dual-H ∞ filter for parameter estimation can be expressed as

[0110]

[0111] where x k = [i d，k i q，k T , u k = [u d，k u q，k T and y k = [i d，k i q，k T are the system state vector, input vector, and output vector respectively, p k = [R s，k ψ m，k T is the motor parameter to be estimated. w x,k-1 and ω p，k-1 are the process noises of the system state and parameters respectively, representing the uncertainty of the system, and v k is the measurement noise, representing the measurement error in the measurement process of the current sensor. d(x k-1 , p k-1 , u k-1 ) and Y(x k , p k ) are nonlinear functions.

[0112] The calculation steps of the dual-H ∞ filter are as follows, where the superscripts "+" and "-" are the correction value and prediction value of the dual-H ∞ filter.

[0113] (1) Parameter prediction

[0114]

[0115] (2) State prediction​​​​

[0116]

[0117] (3) State correction

[0118]

[0119] (4) Parameter correction

[0120]

[0121] Double H ∞ In the filter steps (1) to (4),

[0122]

[0123] wherein,

[0124]

[0125]

[0126] In the initial stage of algorithm startup, K x，k-1 and are respectively initialized as 2×2 all-zero matrices. In the above formula, Q x，k and Q p，k are respectively the covariance matrices corresponding to the system state and parameters, and R k is the covariance matrix corresponding to the measurement noise. S x,k and S p，k are respectively the weight matrices of the state vector x and the parameter p. θ x and θ p are the performance constraint boundaries related to the system model accuracy. θ x and θ p can be set as positive numbers close to zero to improve the accuracy of the state space model. D x，p，k-1 is the Jacobian matrix of the function with respect to the state vector x, and Y x，k and Y p，k are respectively the Jacobian matrices of the output function Y(x k , p k ) with respect to the state vector x and the parameter p. P x,k and P p，k are respectively the error covariance matrices of the state and parameters. K x，k and K p，k are respectively the gain matrices of the state and parameters.

[0127] Step 4): Demagnetization fault diagnosis. The non-invasive fault prediction and health management method for the permanent magnet rotor of a permanent magnet motor based on full parameter estimation proposed by the present invention is implemented in 3 steps. Its control block diagram and diagnostic process flow chart are respectively as Figure 2 and 3 shown. Figure 2 Among them, and are respectively the reference values of the d-q axis current and voltage, and are respectively the reference values of the α-β axis current i α and i β , the α-β axis voltage u α and u β . i a , i b and i c are the three-phase stator currents. dq, αβ and abc are the two-phase rotating coordinate system, two-phase stationary coordinate system and three-phase stationary coordinate system respectively. n ref is the reference value of the rotational speed n, and PI is a proportional-integral controller.

[0128] In the proposed scheme, first, a sinusoidal current signal I inj and f inj with amplitudes and frequencies of I inj sin(2πf inj ) are injected into the d axis to obtain multiple sets of operating states of the permanent magnet motor, and the electrical angular velocity, d-q axis voltage and current signals during the operation of the permanent magnet motor are collected. Secondly, the collected signal data is input into the permanent magnet monitor designed by the present invention, and the d-q axis inductance estimation result obtained by the recursive least squares method with a mutual transfer dynamic forgetting factor and the resistance and flux linkage estimation results ∞ of the double H filter are transferred to each other to eliminate the flux linkage estimation deviation caused by parameter mismatch after the demagnetization fault, so as to obtain the final permanent magnet flux linkage estimation result ψ m,est . Finally, combined with the final flux linkage estimation result, the fault diagnosis is completed by setting a reasonable threshold. Specifically, if the error Г between the estimated value ψ m，est of the permanent magnet flux linkage and the actual measured value ψ m exceeds the set threshold ψ m，thr , it is determined that a demagnetization fault has occurred. Otherwise, the motor is considered to be in a healthy state.

[0129] As can be seen from the above description, the present invention proposes a fault prediction and health management method for the permanent magnet rotor of a permanent magnet motor based on full parameter estimation, which effectively improves the accuracy of demagnetization fault diagnosis.

[0130] The description of the functions and working processes of the present invention in combination with the accompanying drawings of the specification is only one of the preferred embodiments, but the present invention is not limited to the specific functions and working processes described above. The above specific embodiments are merely illustrative rather than restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims, and all of these are within the protection scope of the present invention.

Claims

1. A method for predicting permanent magnet faults and managing health of permanent magnet motor rotors based on full parameter estimation, characterized in that: The following steps are involved: 1) A magnet health management approach based on full parameter identification is proposed, that is, the motor status is monitored in real time through various key parameters of the motor, and the demagnetization fault and its development trend are identified in time based on the final estimated flux linkage value. 2) A fine-grained fault prediction and health management scheme for permanent magnets based on integrated full parameter estimation is designed. This scheme combines direct-axis sinusoidal current injection and dynamic forgetting factor recursive least squares method to online identify the direct-axis inductance parameters, and constructs a double H based on the principle of minimizing and maximizing the estimation error. ∞ The filter is used as the core of flux linkage identification and enables the estimation results of the two algorithms to be mutually transferred to eliminate the parameter mismatch effect caused by demagnetization fault.

Images