Parameter adaptive measurement and control method of high-speed amorphous alloy motor sliding mode observer

Abstract

The application relates to the technical field of motor driving control, and discloses a parameter self-adaptive measurement and control method for a high-speed amorphous alloy motor sliding mode observer, which comprises the following steps: first, collecting stator voltage signals and stator current signals when the high-speed amorphous alloy motor is running, and extracting high-frequency characteristic signals representing material transient response from the stator current signals; subsequently, constructing a dynamic parameter tensor of motor resistance parameters and multi-dimensional inductance parameters, and establishing an amorphous alloy magnetic saturation tensor flow state equation according to the dynamic parameter tensor. By collecting the stator current signals when the motor is running and extracting the high-frequency characteristic signals from the inherent pulse width modulation natural switching harmonics of the inverter, the high-frequency magnetization state of the material can be accurately obtained while ensuring smooth operation of the motor, and by constructing the dynamic parameter tensor of the motor resistance parameters and the multi-dimensional inductance parameters, the mismatch risk of the sliding mode observer caused by the traditional model can be eliminated from the root.

Description

Technical Field

[0001] This invention relates to the field of motor drive control technology, specifically to an adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor. Background Technology

[0002] Amorphous alloy motors have broad application prospects in high-speed electric drives and high-power-density equipment due to their extremely low eddy current losses and excellent high-frequency magnetization characteristics in the stator core. In order to improve the reliability of the drive system, reduce weight and reduce hardware costs, sensorless control technology based on sliding mode observer (SMO) is widely used in the closed-loop control of high-speed motors. The sliding mode observer mainly estimates the extended back electromotive force in real time by constructing a mathematical model containing the internal physical parameters of the motor, and then uses phase-locked loop and other technologies to analyze the rotor position and speed.

[0003] However, during actual operation, the stator resistance of the motor will drift with the accumulation of temperature, and the quadrature-axis inductance will also change with the change of current amplitude. If the parameters inside the observer remain constant, it will lead to serious model mismatch and estimation errors. To deal with this problem, existing technologies usually use model reference adaptive systems, extended Kalman filters, or lookup table methods. These conventional technologies are generally based on the "quasi-steady-state linear assumption", that is, the quadrature-axis inductance and stator resistance of the motor are regarded as isolated scalars, and they are considered to be constant or extremely slow-changing physical quantities in one or more control cycles. After these scalars are tracked or identified online by independent algorithm modules, they are fed back to the sliding mode observer for model update.

[0004] When the aforementioned conventional technologies are directly applied to high-speed amorphous alloy motors, although amorphous alloy materials have extremely low iron losses, their unique permeability characteristics make the motor prone to entering a local nonlinear magnetic saturation state under high-speed, high-frequency, and strongly excitation conditions. This causes the stator inductance to no longer be a slowly changing parameter, but rather exhibits extremely high-frequency nonlinear pulsations with the transient current vector. Existing independent parameter adaptive modules are limited by regulator bandwidth or filter lag and cannot track this high-frequency transient nonlinear change in real time. This makes it impossible for the observer to achieve effective compensation when facing cross-coupled disturbances, which in turn leads to a sharp increase in sliding surface chattering and distortion of the back EMF phase resolution.

[0005] Because the nonlinear evolution of parameters of amorphous alloy motors and the analytical process of the observer on the system state are deeply intertwined on the time scale, the traditional serial processing architecture of first identifying scalar parameters separately and then substituting them into the observer is prone to execution phase delay. If multiple strongly coupled high-frequency nonlinear parameters are identified online simultaneously in the traditional scalar equation system, it is easy to cause rank deficiency of the equation or get trapped in local extrema. In the end, it will not only exhaust the real-time computing power of the microprocessor, but also cause parameter identification to diverge. Therefore, this invention designs an adaptive measurement and control method for sliding mode observer parameters of high-speed amorphous alloy motors based on the above-mentioned problems. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an adaptive measurement and control method for sliding mode observer parameters of high-speed amorphous alloy motors. This method solves the problems of model mismatch, easy divergence in online identification of multidimensional strongly coupled parameters, and insufficient computing power for high-frequency real-time updates in sensorless control of high-speed amorphous alloy motors due to high-frequency nonlinearity of materials and strong cross-magnetic saturation.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor, comprising the following steps:

[0008] S1. First, collect the stator voltage signal and stator current signal during the operation of the high-speed amorphous alloy motor, and extract the high-frequency characteristic signal characterizing the transient response of the material from the stator current signal;

[0009] S2. Subsequently, a dynamic parameter tensor that integrates motor resistance parameters and multidimensional inductance parameters is constructed, and the state equation of the amorphous alloy magnetic saturation tensor manifold is established based on the dynamic parameter tensor.

[0010] S3. Based on the state equation of the amorphous alloy magnetic saturation tensor manifold and the current estimated parameter tensor, construct a decoupled sliding mode observer, input the stator voltage signal into the decoupled sliding mode observer for calculation, and obtain the estimated current and estimated back electromotive force.

[0011] S4. Calculate the observation error between the estimated current and the stator current signal, and combine the observation error with the high-frequency characteristic signal to calculate the update direction matrix used to correct the parameters;

[0012] S5. Perform asynchronous decomposition and update operation across time scales on the updated direction matrix to reconstruct a new estimated parameter tensor, and feed the new estimated parameter tensor back to the decoupled sliding mode observer to replace the current estimated parameter tensor, thereby completing the model parameter refresh inside the sliding mode observer.

[0013] S6. Finally, based on the estimated back electromotive force, the estimated rotational speed and estimated electrical angle of the rotor are analyzed. A drive signal is generated according to the estimated rotational speed and the estimated electrical angle, and the high-speed amorphous alloy motor can be controlled to operate in closed loop using the drive signal.

[0014] Preferably, the process of extracting high-frequency feature signals in step S1 specifically includes:

[0015] The stator voltage signal and the stator current signal are converted to a two-phase stationary coordinate system to obtain the stator voltage vector and the stator current vector;

[0016] By utilizing the inherent pulse width modulation natural switching harmonics of the inverter, a high-frequency current ripple vector is extracted from the stator current vector through a bandpass filter;

[0017] The phase offset angle of the high-frequency current ripple vector relative to the fundamental wave is extracted by a phase-locked loop, and the phase offset angle is used as the high-frequency characteristic signal.

[0018] Preferably, the process of establishing the state equation in step S2 specifically includes:

[0019] The stator resistance matrix of the high-speed amorphous alloy motor is integrated with the inductance matrix containing self-inductance and cross-inductance to construct a second-order dynamic parameter vector matrix as the dynamic parameter tensor.

[0020] The generalized state transition matrix and input gain matrix, which include cross-magnetic saturation characteristics, are defined using the dynamic parameter tensor.

[0021] By combining the generalized state transition matrix, the input gain matrix, the stator current vector, the stator voltage vector, and the extended back electromotive force vector, a state equation for the amorphous alloy magnetic saturation tensor manifold with the derivative of the stator current vector with respect to time as the dependent variable is established.

[0022] Preferably, the process of constructing the decoupled sliding mode observer in step S3 specifically includes:

[0023] Substitute the current estimated parameter tensor into the generalized state transition matrix and the input gain matrix to construct the state equation of the decoupled sliding mode observer;

[0024] A sliding mode control law consisting of a smoothing function and a sliding mode gain diagonal matrix is ​​introduced into the decoupled sliding mode observer to suppress observation chattering and output the estimated current;

[0025] Based on the equivalent control principle, the sliding mode control law is processed by low-pass filtering to extract the estimated back electromotive force containing rotor position information.

[0026] Preferably, the process of calculating the observation error in step S4 specifically includes:

[0027] The difference between the estimated current and the stator current signal is used to obtain the current observation error vector as the observation error.

[0028] The difference between the current estimated parameter tensor and the actual dynamic parameter tensor is used to obtain the parameter tensor estimation error matrix.

[0029] Using a positive definite weight matrix, a Lyapunov composite energy function containing the current observation error vector and the trace of the parameter tensor estimation error matrix is ​​constructed.

[0030] Preferably, the process of calculating the update direction matrix used to correct the parameters further includes:

[0031] Let the time derivative of the Lyapunov composite energy function be less than or equal to zero;

[0032] The high-frequency characteristic signal is mapped and converted into an amorphous iron loss weighted constraint matrix;

[0033] Based on the current observation error vector, the transpose matrix of the estimated current, the positive definite weight matrix, and the amorphous iron loss weighting constraint matrix, the update direction matrix that satisfies the system stability condition is obtained.

[0034] Preferably, the process of asynchronously decomposing the update direction matrix across time scales in step S5 specifically includes:

[0035] Perform singular value decomposition on the updated direction matrix to obtain an orthogonal matrix and a singular value diagonal matrix;

[0036] The singular value diagonal matrix is ​​split into a primary singular value matrix representing high-frequency dynamics and a secondary singular value matrix representing slow-varying dynamics.

[0037] By combining the orthogonal matrix, the fast-variable derivative term is reconstructed using the principal singular value matrix, and the slow-variable derivative term is reconstructed using the secondary singular value matrix.

[0038] Preferably, the process of updating the direction matrix in step S5 specifically includes:

[0039] In the high-frequency master control layer, the fast variable derivative term is discretized and integrated within the microsecond-level master interrupt cycle to obtain the fast parameter subtensor.

[0040] In the low-frequency background layer, the slow-variable derivative term is discretized and integrated during the background cycle when the processor is idle, resulting in the slow parameter subtensor.

[0041] Preferably, the process of reconstructing the new estimated parameter tensor and feeding it back in step S5 specifically includes:

[0042] Before the decoupled sliding mode observer is called to perform calculations in the main control cycle, the fast parameter subtensor and the slow parameter subtensor calculated at the moment are added together to form the new estimated parameter tensor.

[0043] The model parameters inside the decoupled sliding mode observer are refreshed by replacing the current estimated parameter tensor with the new estimated parameter tensor.

[0044] Preferably, the process of generating the drive signal and controlling the closed-loop operation of the high-speed amorphous alloy motor in step S6 specifically includes:

[0045] The estimated back electromotive force is input into the phase-locked loop to calculate the estimated rotational speed and the estimated electrical angle.

[0046] The error between the estimated rotational speed and the target given rotational speed is input into the speed loop regulator to generate a current command;

[0047] The estimated electrical angle is used to perform a coordinate synchronous rotation transformation, and a reference voltage command is generated by the current loop regulator based on the current command.

[0048] The reference voltage command is converted into a three-phase duty cycle signal via space vector pulse width modulation and output to the inverter to generate the drive signal to control the closed-loop operation of the high-speed amorphous alloy motor.

[0049] This invention provides an adaptive measurement and control method for sliding mode observer parameters of high-speed amorphous alloy motors. It offers the following advantages:

[0050] 1. This invention collects the stator current signal during motor operation and extracts high-frequency characteristic signals that characterize the transient response of amorphous alloys from it using the inherent pulse width modulation natural switching harmonics of the inverter. This effectively avoids the problems of additional torque pulsation, electromagnetic noise, and core loss caused by high-frequency injection, and accurately obtains the high-frequency magnetization state of the material while ensuring the smooth operation of the motor.

[0051] 2. This invention constructs a dynamic parameter tensor that integrates motor resistance parameters and multidimensional inductance parameters, and establishes the state equation of the amorphous alloy magnetic saturation tensor manifold based on it. It directly and explicitly characterizes the deep cross-magnetic saturation and dynamic coupling effect of the amorphous alloy motor under high-speed and strong magnetic conditions in the underlying mathematical model, thereby eliminating the risk of sliding mode observer mismatch caused by nonlinear parameter mutations in traditional models from the root.

[0052] 3. This invention calculates the observation error between the estimated current and the actual stator current signal, and combines the weighted constraint matrix of the amorphous iron loss generated by the extracted high-frequency feature signal to calculate the update direction matrix. It directly introduces the real physical high-frequency boundary conditions of the amorphous alloy material into the Lyapunov stability energy function, realizing the deep integration of physical properties and mathematical algorithms, and effectively preventing the divergence and instability of the algorithm's adaptive iteration process.

[0053] 4. This invention performs singular value decomposition on the calculated parameter update direction matrix, precisely splitting it into a primary singular value matrix representing high-frequency dynamics and a secondary singular value matrix representing slow-varying dynamics. Mathematically, this invention successfully decouples the rapid inductance transients caused by high-speed cross magnetic saturation from the slow resistance drift caused by temperature accumulation, reducing parameter cross interference in online identification of nonlinear strongly coupled systems.

[0054] 5. This invention solves the severe microprocessor computing bottleneck problem caused by real-time adaptive calculation of high-dimensional tensor matrices by performing asynchronous update operations across time scales on the split fast-variable derivative terms and slow-variable derivative terms, that is, processing high-frequency terms within the microsecond-level main control cycle and processing low-frequency terms within the millisecond-level background cycle and finally reconstructing and refreshing. This enables the high-order complex algorithm to be successfully applied to conventional industrial-grade controller hardware, breaking through the engineering limitation that traditional parameter identification and state observation must be synchronously closed-loop within the same high-frequency interrupt. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the method architecture of the present invention;

[0056] Figure 2 This is one of the schematic diagrams of the method flow of the present invention;

[0057] Figure 3 This is a second schematic diagram of the method flow of the present invention;

[0058] Figure 4 This is the third schematic diagram of the method flow of the present invention;

[0059] Figure 5 This is the fourth schematic diagram of the method flow of the present invention;

[0060] Figure 6 This is the fifth schematic diagram of the method flow of the present invention;

[0061] Figure 7 This is the sixth schematic diagram of the method flow of the present invention. Detailed Implementation

[0062] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Please see the appendix Figure 1 -Appendix Figure 7 This invention provides an adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor, comprising the following steps:

[0064] S1. First, acquire the stator voltage signal and stator current signal during the operation of the high-speed amorphous alloy motor, and extract the high-frequency characteristic signal characterizing the transient response of the material from the stator current signal. The process of extracting the high-frequency characteristic signal specifically includes: converting the stator voltage signal and stator current signal to a two-phase stationary coordinate system to obtain the stator voltage vector and stator current vector; using the inherent pulse width modulation natural switching harmonics of the inverter, extracting the high-frequency current ripple vector from the stator current vector through a bandpass filter; extracting the phase offset angle of the high-frequency current ripple vector relative to the fundamental wave through a phase-locked loop, and using the phase offset angle as the high-frequency characteristic signal.

[0065] Specifically, this step aims to obtain the physical state of the motor operation and extract constraint signals deeply bound to the physical properties of the amorphous alloy material. First, during the main cycle of the control system, the DC bus voltage on the inverter side and the three-phase current of the motor stator are collected in real time through voltage and current sensors. Then, using Clark coordinate transformation, the signals in the three-phase stationary coordinate system are converted to the two-phase stationary coordinate system. (Coordinate system) to obtain the stator voltage vector With stator current vector To characterize the eddy current and transient magnetization response of amorphous alloys under extremely low iron loss and high entropy conditions, this method does not employ additional external guided injection. Instead, it directly utilizes the inverter's inherent pulse width modulation (PWM) natural switching waveform as the excitation source. The acquired stator current waveform is input into a digital phase-locked loop (PLL) for phase analysis, extracting the phase shift angle of the ripple relative to the fundamental current. The phase offset angle As a high-frequency characteristic signal, it reflects the high-frequency leakage flux and local saturation state inside the amorphous stator core.

[0066] S2. Subsequently, a dynamic parameter tensor that integrates the motor resistance parameters and multidimensional inductance parameters is constructed, and the state equation of the amorphous alloy magnetic saturation tensor manifold is established based on the dynamic parameter tensor. The process of establishing the state equation specifically includes: integrating the stator resistance matrix of the high-speed amorphous alloy motor with the inductance matrix containing self-inductance and cross-inductance, and constructing a second-order dynamic parameter vector matrix as the dynamic parameter tensor.

[0067] The generalized state transition matrix and input gain matrix containing cross-magnetic saturation characteristics are defined using dynamic parameter tensors. By combining the generalized state transition matrix, input gain matrix, stator current vector, stator voltage vector and extended back electromotive force vector, the state equation of the amorphous alloy magnetic saturation tensor manifold with the derivative of the stator current vector with respect to time as the dependent variable is established.

[0068] Specifically, this step breaks away from the conventional setting of treating motor resistance and quadrature-axis inductance as independent scalars, and constructs a high-dimensional mathematical model to cover cross-coupling effects;

[0069] First, define a second-order dynamic parameter vector matrix as the dynamic parameter tensor. This tensor represents the stator resistance matrix of a high-speed amorphous alloy motor. With a multidimensional inductance matrix including self-inductance and cross-inductance To achieve comprehensive integration, a generalized state transition matrix is ​​defined based on the dynamic parameter tensor. and the input gain matrix Based on this, combined with the extended back electromotive force vector Establish a stator current vector The equation of state for the magnetic saturation tensor manifold of an amorphous alloy, with its derivative with respect to time as the dependent variable, is expressed as follows:

[0070] ;

[0071] In this equation, the matrix and The off-diagonal elements explicitly characterize , The axis is affected by cross-magnetic saturation and coupling caused by high-speed strong magnetic fields.

[0072] S3. Based on the state equation of the amorphous alloy magnetic saturation tensor manifold and the current estimated parameter tensor, a decoupled sliding mode observer is constructed. The stator voltage signal is input into the decoupled sliding mode observer for calculation to obtain the estimated current and estimated back electromotive force. The specific process of constructing the decoupled sliding mode observer includes: substituting the current estimated parameter tensor into the generalized state transition matrix and the input gain matrix to construct the state equation of the decoupled sliding mode observer; introducing a sliding mode control law composed of a smoothing function and a sliding mode gain diagonal matrix into the decoupled sliding mode observer to suppress observation chattering and output the estimated current; according to the equivalent control principle, the sliding mode control law is processed by low-pass filtering to extract the estimated back electromotive force containing rotor position information.

[0073] Specifically, this step is used to construct the core module for state observation, enabling online estimation of unknown current and back electromotive force. A current estimation parameter tensor is set in the control program. Substituting these into the generalized state transition matrix and the input gain matrix, we construct the state equation for the decoupled observer:

[0074] ;

[0075] In the formula, To estimate the current, To estimate the back electromotive force, To overcome the severe chattering of conventional sign functions under high-speed computation, a smoothing function (such as the sigmoid function) is designed as the observer gain matrix. ) and sliding mode gain diagonal matrix Sliding mode control law: When the system state reaches the sliding surface, according to the equivalent control principle, the sliding control law is... After low-pass filtering, high-frequency switching noise can be removed, and the estimated back electromotive force containing precise rotor position information can be extracted. .

[0076] S4. Calculate the observation error between the estimated current and the stator current signal. Combining the observation error with the high-frequency characteristic signal, calculate the update direction matrix used to correct the parameters. The process of calculating the observation error specifically includes: subtracting the estimated current from the stator current signal to obtain the current observation error vector as the observation error; subtracting the current estimated parameter tensor from the actual dynamic parameter tensor to obtain the parameter tensor estimation error matrix; using the positive definite weight matrix, constructing a Lyapunov composite energy function containing the current observation error vector and the trace of the parameter tensor estimation error matrix. The process of calculating the update direction matrix used to correct the parameters also includes: setting the derivative of the Lyapunov composite energy function with respect to time to be less than or equal to zero; mapping the high-frequency characteristic signal into an amorphous iron loss weighted constraint matrix; and solving for the update direction matrix that satisfies the system stability condition based on the current observation error vector, the transpose matrix of the estimated current, the positive definite weight matrix, and the amorphous iron loss weighted constraint matrix.

[0077] Specifically, this step aims to solve for the parameter tensor adjustment direction that converges the system error, based on the rigorous Lyapunov stability theorem. First, the estimated current... With stator current signal Difference, define the current observation error vector ; transfer the current estimated parameter tensor Tensor of the actual dynamic parameters of the motor Difference, define the parameter tensor estimation error matrix Using two positive definite weight matrices and The Lyapunov composite energy function is constructed as follows:

[0078] ;

[0079] in The expression represents finding the trace of a matrix. To ensure the global asymptotic stability of the observer error system, let the time derivative of the composite energy function be... Simultaneously, the phase shift angle (high-frequency characteristic signal) obtained in the above steps The mapping is transformed into an amorphous iron loss weighted constraint matrix. As the physical boundary condition for tensor adaptation, the update direction matrix that satisfies stability can be obtained through comprehensive derivation. :

[0080] .

[0081] S5. Perform asynchronous decomposition and update operations across time scales on the update direction matrix to reconstruct a new estimated parameter tensor. Feed this new estimated parameter tensor back to the decoupled sliding mode observer to replace the current estimated parameter tensor, thus refreshing the model parameters within the sliding mode observer. The process of asynchronous decomposition across time scales on the update direction matrix specifically includes: performing singular value decomposition on the update direction matrix to obtain an orthogonal matrix and a singular value diagonal matrix; splitting the singular value diagonal matrix into a principal singular value matrix representing high-frequency dynamics and a secondary singular value matrix representing slow-varying dynamics; combining the orthogonal matrix, reconstructing the fast-varying derivative term using the principal singular value matrix and the slow-varying derivative term using the secondary singular value matrix, and then updating the update direction... The process of updating the matrix specifically includes: in the high-frequency main control layer, discretization integration is performed on the fast-variable derivative term within the microsecond-level main interrupt cycle to obtain the fast parameter sub-tensor; in the low-frequency background layer, discretization integration is performed on the slow-variable derivative term within the processor idle background cycle to obtain the slow parameter sub-tensor. The process of reconstructing a new estimated parameter tensor and feeding it back specifically includes: before calling the decoupled sliding mode observer to perform the operation in the main control cycle, the currently calculated fast parameter sub-tensor and the slow parameter sub-tensor are added together to form a new estimated parameter tensor; the new estimated parameter tensor replaces the current estimated parameter tensor, completing the model parameter refresh inside the decoupled sliding mode observer;

[0082] Specifically, regarding the aforementioned update direction matrix To address the issues of high computational dimensionality and excessive computational demands on conventional MCUs / DSPs, this step utilizes Singular Value Decomposition (SVD) to decouple and physically divide complex matrix integration operations along the time axis.

[0083] First, the calculated update direction matrix is... Perform singular value decomposition: Where U and V are orthogonal matrices, Given a singular value diagonal matrix, based on the magnitude of the values, Decomposed into principal singular value matrices representing the fast transients of the cross-saturated inductance. And the subsingular value matrix characterizing the slow effect of temperature on stator resistance. Based on this, the fast variable derivative term is reconstructed. With slow-varying derivative terms Then, perform cross-timescale update operations:

[0084] High-frequency main control layer: high-speed main interrupt cycle in the microsecond range (e.g., tens of microseconds). Within this framework, only the fast variable derivative term is discretized and integrated to obtain the fast parametric subtensor. ;

[0085] Low-frequency background layer: Processor idle task cycles in milliseconds (e.g., several milliseconds). Within this process, the slow-varying derivative term is discretized and integrated asynchronously to obtain the slow-parameter subtensor. ;

[0086] Before each call to the decoupled sliding mode observer for a new round of state prediction, the system adds the latest fast parameter subtensor to the slow parameter subtensor. These parameters are combined to form a new estimated parameter tensor, which is then fed back to step S3 to achieve zero-delay refresh of the model parameters inside the observer.

[0087] S6. Finally, based on the estimated back EMF, the estimated rotor speed and estimated electrical angle are analyzed. A drive signal is generated based on the estimated speed and estimated electrical angle, and this drive signal is used to control the closed-loop operation of the high-speed amorphous alloy motor. The specific process of generating the drive signal and controlling the closed-loop operation of the high-speed amorphous alloy motor includes: inputting the estimated back EMF into the phase-locked loop to calculate the estimated speed and estimated electrical angle; inputting the error between the estimated speed and the target given speed into the speed loop regulator to generate a current command; performing coordinate synchronous rotation transformation using the estimated electrical angle, and generating a reference voltage command based on the current command through the current loop regulator; converting the reference voltage command into a three-phase duty cycle signal via space vector pulse width modulation and outputting it to the inverter to generate the drive signal to control the closed-loop operation of the high-speed amorphous alloy motor.

[0088] Specifically, this step is responsible for converting the calibrated and extracted observation signal into actual physical drive commands for the hardware inverter, and converting the estimated back electromotive force calculated in step S3 into... The input is fed into the phase-locked loop architecture of the control system, where the estimated rotor speed is calculated through the internal PI controller and integrator. With estimation of electrical angle The extracted estimated speed is compared with the target speed given by the system, and the difference is input to the speed loop PI regulator to generate a corresponding quadrature-axis current command. Then, using the calculated estimated electrical angle, a coordinate synchronous rotation transformation (Park transformation) is performed on the feedback stator current vector to obtain the current in the rotating coordinate system. The current loop PI regulator generates a reference voltage command based on the current command. Finally, the reference voltage command is converted into a three-phase duty cycle signal (i.e., drive signal) by the space vector pulse width modulation (SVPWM) module, which is applied to the power switches of the inverter to drive the high-speed amorphous alloy motor to operate stably, realizing sensorless closed-loop control of the entire system.

[0089] In summary, this invention provides an adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor. By acquiring the stator current signal during motor operation and utilizing the inherent pulse width modulation natural switching harmonics of the inverter, high-frequency characteristic signals characterizing the transient response of the amorphous alloy are extracted. This eliminates the need to inject additional high-frequency excitation voltage or current into the control system, effectively avoiding the additional torque ripple, electromagnetic noise, and core loss problems caused by high-frequency injection. It accurately obtains the high-frequency magnetization state of the material while ensuring stable motor operation. Furthermore, by constructing a dynamic parameter tensor that integrates motor resistance parameters and multi-dimensional inductance parameters, and establishing the state equation of the amorphous alloy magnetic saturation tensor manifold based on this tensor, this invention breaks away from the existing technology that treats stator resistance and quadrature-axis inductance as independent parameters. The traditional assumptions about quantity are directly and explicitly characterized in the underlying mathematical model the deep cross-magnetic saturation and dynamic coupling effect of amorphous alloy motors under high-speed and strong magnetic conditions. This eliminates the risk of sliding mode observer mismatch caused by abrupt changes in nonlinear parameters in traditional models. Furthermore, by calculating the observation error between the estimated current and the actual stator current signal, and combining the weighted constraint matrix of amorphous iron loss generated by the extracted high-frequency feature signal mapping, the update direction matrix is ​​calculated. The real physical high-frequency boundary conditions of amorphous alloy materials are directly introduced into the Lyapunov stability energy function, realizing the deep integration of physical properties and mathematical algorithms. This ensures the global asymptotic stability of multidimensional tensor parameters in online analysis under complex high-frequency conditions, effectively preventing divergence and instability in the adaptive iteration process of the algorithm.

[0090] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor, characterized in that, Includes the following steps: S1. First, collect the stator voltage signal and stator current signal during the operation of the high-speed amorphous alloy motor, and extract the high-frequency characteristic signal characterizing the transient response of the material from the stator current signal; S2. Subsequently, a dynamic parameter tensor that integrates motor resistance parameters and multidimensional inductance parameters is constructed, and the state equation of the amorphous alloy magnetic saturation tensor manifold is established based on the dynamic parameter tensor. S3. Based on the state equation of the amorphous alloy magnetic saturation tensor manifold and the current estimated parameter tensor, construct a decoupled sliding mode observer, input the stator voltage signal into the decoupled sliding mode observer for calculation, and obtain the estimated current and estimated back electromotive force. S4. Calculate the observation error between the estimated current and the stator current signal, and combine the observation error with the high-frequency characteristic signal to calculate the update direction matrix used to correct the parameters; S5. Perform asynchronous decomposition and update operation across time scales on the updated direction matrix to reconstruct a new estimated parameter tensor, and feed the new estimated parameter tensor back to the decoupled sliding mode observer to replace the current estimated parameter tensor, thereby completing the model parameter refresh inside the sliding mode observer. S6. Finally, based on the estimated back electromotive force, the estimated rotational speed and estimated electrical angle of the rotor are analyzed. A drive signal is generated according to the estimated rotational speed and the estimated electrical angle, and the high-speed amorphous alloy motor can be controlled to operate in closed loop using the drive signal.

2. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 1, characterized in that, The process of extracting high-frequency feature signals in step S1 specifically includes: The stator voltage signal and the stator current signal are converted to a two-phase stationary coordinate system to obtain the stator voltage vector and the stator current vector; By utilizing the inherent pulse width modulation natural switching harmonics of the inverter, a high-frequency current ripple vector is extracted from the stator current vector through a bandpass filter; The phase offset angle of the high-frequency current ripple vector relative to the fundamental wave is extracted by a phase-locked loop, and the phase offset angle is used as the high-frequency characteristic signal.

3. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 2, characterized in that, The process of establishing the state equations in step S2 specifically includes: The stator resistance matrix of the high-speed amorphous alloy motor is integrated with the inductance matrix containing self-inductance and cross-inductance to construct a second-order dynamic parameter vector matrix as the dynamic parameter tensor. The generalized state transition matrix and input gain matrix, which include cross-magnetic saturation characteristics, are defined using the dynamic parameter tensor. By combining the generalized state transition matrix, the input gain matrix, the stator current vector, the stator voltage vector, and the extended back electromotive force vector, a state equation for the amorphous alloy magnetic saturation tensor manifold with the derivative of the stator current vector with respect to time as the dependent variable is established.

4. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 3, characterized in that, The process of constructing the decoupled sliding mode observer in step S3 specifically includes: Substitute the current estimated parameter tensor into the generalized state transition matrix and the input gain matrix to construct the state equation of the decoupled sliding mode observer; A sliding mode control law consisting of a smoothing function and a sliding mode gain diagonal matrix is ​​introduced into the decoupled sliding mode observer to suppress observation chattering and output the estimated current; Based on the equivalent control principle, the sliding mode control law is processed by low-pass filtering to extract the estimated back electromotive force containing rotor position information.

5. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 1, characterized in that, The process of calculating the observation error in step S4 specifically includes: The difference between the estimated current and the stator current signal is used to obtain the current observation error vector as the observation error. The difference between the current estimated parameter tensor and the actual dynamic parameter tensor is used to obtain the parameter tensor estimation error matrix. Using a positive definite weight matrix, a Lyapunov composite energy function containing the current observation error vector and the trace of the parameter tensor estimation error matrix is ​​constructed.

6. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 5, characterized in that, The process of calculating the update direction matrix used to correct the parameters also includes: Let the time derivative of the Lyapunov composite energy function be less than or equal to zero; The high-frequency characteristic signal is mapped and converted into an amorphous iron loss weighted constraint matrix; Based on the current observation error vector, the transpose matrix of the estimated current, the positive definite weight matrix, and the amorphous iron loss weighting constraint matrix, the update direction matrix that satisfies the system stability condition is obtained.

7. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 1, characterized in that, Step S5, which involves asynchronous decomposition of the update direction matrix across time scales, specifically includes: Perform singular value decomposition on the updated direction matrix to obtain an orthogonal matrix and a singular value diagonal matrix; The singular value diagonal matrix is ​​split into a primary singular value matrix representing high-frequency dynamics and a secondary singular value matrix representing slow-varying dynamics. By combining the orthogonal matrix, the fast-variable derivative term is reconstructed using the principal singular value matrix, and the slow-variable derivative term is reconstructed using the secondary singular value matrix.

8. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 1, characterized in that, The process of updating the direction matrix in step S5 specifically includes: In the high-frequency master control layer, the fast variable derivative term is discretized and integrated within the microsecond-level master interrupt cycle to obtain the fast parameter subtensor. In the low-frequency background layer, the slow-variable derivative term is discretized and integrated during the background cycle when the processor is idle, resulting in the slow parameter subtensor.

9. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 1, characterized in that, The process of reconstructing the new estimated parameter tensor and feeding it back in step S5 specifically includes: Before the decoupled sliding mode observer is called to perform calculations in the main control cycle, the fast parameter subtensor and the slow parameter subtensor calculated at the moment are added together to form the new estimated parameter tensor. The model parameters inside the decoupled sliding mode observer are refreshed by replacing the current estimated parameter tensor with the new estimated parameter tensor.

10. The adaptive measurement and control method for sliding mode observer parameters of a high-speed amorphous alloy motor according to claim 1, characterized in that, The process of generating drive signals and controlling the closed-loop operation of the high-speed amorphous alloy motor in step S6 specifically includes: The estimated back electromotive force is input into the phase-locked loop to calculate the estimated rotational speed and the estimated electrical angle. The error between the estimated rotational speed and the target given rotational speed is input into the speed loop regulator to generate a current command; The estimated electrical angle is used to perform a coordinate synchronous rotation transformation, and a reference voltage command is generated by the current loop regulator based on the current command. The reference voltage command is converted into a three-phase duty cycle signal via space vector pulse width modulation and output to the inverter to generate the drive signal to control the closed-loop operation of the high-speed amorphous alloy motor.

Images