PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems

Abstract

The invention discloses a PMSLM adaptive optimization harmonic current suppression method based on a multi-rotating coordinate system. According to the method, high-precision extraction of target harmonic waves is realized through a multi-rotating reference coordinate system and a complex coefficient filter; an adaptive moment estimation algorithm (AdamW) with weight attenuation is used to quickly solve an optimal harmonic reference current; and harmonic compensation voltage is output by combining the zero static error characteristic of the proportional resonance PR controller, so that the harmonic current is effectively suppressed. The harmonic current of the permanent magnet synchronous linear motor is quickly and effectively suppressed through a harmonic injection method adopting a multi-rotation reference coordinate system, a complex coefficient filter and an AdamW optimization algorithm. The inherent defects of large phase lag and slow dynamic response when a low-pass filter is used in a traditional harmonic current injection method can be effectively overcome by adopting a complex coefficient filter for current extraction; and an AdamW optimization algorithm is adopted to suppress overfitting through a weight attenuation strategy, and an optimal solution is efficiently and stably obtained.

Description

Technical Field

[0001] This invention belongs to the field of permanent magnet synchronous linear motor control technology, specifically relating to a PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems. Background Technology

[0002] Permanent Magnet Synchronous Linear Motor Permanent magnet synchronous linear motors (PMLMs) are widely used in industrial production due to their advantages of zero transmission backlash, fast response speed, and high positioning accuracy. However, during operation, PMSLMs generate a large number of current harmonic components due to factors such as air gap magnetic field distortion, stator winding asymmetry, inverter nonlinearity, and parameter variations. These current harmonic components not only cause current distortion and thrust fluctuations but also affect motor operating efficiency and shorten motor lifespan.

[0003] Harmonic current injection is a commonly used and effective method for harmonic suppression. Its principle involves calculating a harmonic reference current using an optimization algorithm, then controlling the actual harmonic current to follow the reference current using a harmonic current controller. The output harmonic compensation voltage is superimposed on the fundamental voltage to drive the motor, thus suppressing motor harmonics. The main steps of harmonic current injection are harmonic extraction, optimal harmonic reference current calculation, and harmonic current injection. Traditional harmonic current injection methods have several limitations: (1) When extracting harmonics, the method of using a low-pass filter in a single fixed rotating reference coordinate system has the defects of large phase lag and slow dynamic response. It is also difficult to effectively separate harmonic components of different orders, resulting in insufficient extraction accuracy and difficulty in meeting the requirements of high-precision control. (2) When solving for the optimal harmonic current, the traditional gradient descent algorithm has problems such as slow convergence speed and easy to get trapped in local optima, resulting in low efficiency and poor accuracy in solving for the optimal harmonic reference current, which affects the effect of harmonic injection. Summary of the Invention

[0004] The purpose of this invention is to provide a PMSLM adaptive optimization method for harmonic current suppression based on multiple rotating coordinate systems, which solves the problems of low harmonic extraction accuracy and poor efficiency in solving the optimal harmonic reference current in existing harmonic injection techniques.

[0005] The technical solution adopted in this invention is a PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems, which is implemented according to the following steps: Step 1: Establish a permanent magnet synchronous linear motor model and acquire the phase current signal of the motor through a current Hall sensor. After coordinate transformation, the following is obtained d shaft andq shaft current; Step 2: Establish independent rotating coordinate systems for the target harmonic order, and use complex coefficient filters in each rotating coordinate system to output the actual harmonic current; Step 3: Construct an optimal harmonic reference current solution model for harmonic injection into a linear motor, through... AdamW The algorithm is iteratively optimized to obtain the optimal harmonic reference current; Step 4: Use the error between the optimal harmonic reference current and the actual harmonic current extracted by the complex coefficient filter as... PR The controller input generates a harmonic compensation voltage, which is injected into the feedforward channel and works together with the fundamental voltage on the inverter to compensate for the harmonic current.

[0006] The invention is further characterized in that, Step 1 specifically involves: Modeling the permanent magnet synchronous linear motor yields the voltage balance equation, as shown in equation (1): (1); in, , They are respectively d shaft voltage 、q Shaft voltage; , They are respectively d Shaft inductance, q Shaft inductance; It is a permanent magnet flux linkage. This indicates the speed of the mover relative to the stator; Indicates polar moment, This refers to the stator winding resistance. , They are respectively d shaft current 、q shaft current; The actual phase current of the linear motor is collected using a Hall effect current sensor. Three-phase current After coordinate transformation, we obtain d shaft current and q shaft current As shown in equation (2): (2); in, for d shaft and q The included angle of the axis.

[0007] Step 2 specifically involves: Step 2.1: Using the multi-rotating reference coordinate system theory, establish a synchronously rotating coordinate system for the specific harmonics to be suppressed, defining the fundamental angular velocity as... ,in, f If 1 is the fundamental frequency of the linear motor, then the... n The angular velocity of the rotating coordinate system corresponding to the first harmonic is set as follows: ; Step 2.2, establish the complex coefficient filter model, as shown in equation (3); (3) in, s It is a complex variable in the Laplace transform. j It is the imaginary unit. k a The attenuation coefficient; The input to this transfer function contains harmonics of various orders. d, q shaft current ， yes After passing through an angular velocity of After rotating the coordinate system PARK The output obtained by the transformation is the actual harmonic current containing the target order harmonics. That is .

[0008] Step 3 specifically involves: Step 3.1, to achieve harmonic minimization control, establish a system based on the total harmonic distortion of the current (THD). THD The least squares loss function with the objective of minimizing is shown in equation (4): (4) in, for AdamW The parameter vector for algorithm optimization; The calculated number of... n The target reference current for the first harmonic; H This refers to the set of harmonic orders that need to be suppressed. Step 3.2, through AdamW The parameter configuration of the optimization algorithm is obtained when the algorithm converges. This is the optimal parameter vector, i.e., the optimal harmonic reference current.

[0009] (1) Calculate the first t The gradient of the step is shown in equation (5): (5) (2) Update the first-order moment estimate, as shown in equation (6): (6) (3) Update the second-order moment estimate, as shown in equation (7): (7) (4) Correct the bias of the first moment estimate, as shown in equation (8): (8) (5) Correct the bias of the second moment estimate, as shown in equation (9): (9) (6) Update parameters As shown in equation (10): (10) (7) Determine the convergence condition. The iteration terminates when any of the following conditions (11) are met: (11) in, for In the t Parameter values ​​at the next iteration For learning rate, and These are the first-order exponential decay coefficient and the second-order exponential decay coefficient, respectively. , The gradient of the loss function. Here, represents the weight decay coefficient, and N represents the number of iterations for continuous detection. This is the convergence accuracy threshold; Maximum number of iterations yes k Total harmonic distortion of current at any given time.

[0010] Step 4 specifically involves: Step 4.1, Build PR The transfer function of the controller is shown in equation (12): (12) in, The resonant frequency, For cutoff bandwidth, K p This is the proportionality coefficient. K r Resonance coefficient; Step 4.2, take AdamW The error between the optimal harmonic reference current obtained by the optimization algorithm and the actual current extracted by the complex coefficient filter is used as... PR Controller input, PR Controller output harmonic compensation voltage , PRHarmonic compensation voltage output by the controller With the motor feedforward stage Generate after superposition and coordinate transformation , It is applied to the inverter to drive a linear motor, enabling precise and rapid suppression of harmonic currents.

[0011] The beneficial effects of this invention are: to reduce the harmonic current in the permanent magnet synchronous linear motor drive system and thus reduce the impact of the harmonic current on the motor system, this invention designs a PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems. This method employs multiple rotating reference coordinate systems, complex coefficient filters, and... AdamW An optimized harmonic injection method is proposed to achieve rapid and effective suppression of harmonic currents in permanent magnet synchronous linear motors. Specifically, the use of a complex coefficient filter for current extraction effectively overcomes the inherent defects of traditional harmonic current injection methods, such as large phase lag and slow dynamic response when using low-pass filters. AdamW The optimization algorithm addresses the slow convergence speed and susceptibility to local optima when using gradient descent to solve for the optimal harmonic reference current. It also suppresses overfitting through a weight decay strategy, efficiently and stably obtaining the optimal solution. Combined with... PR The controller's zero steady-state error tracking characteristic enables precise and rapid compensation of harmonic currents, thereby suppressing harmonic currents in permanent magnet synchronous linear motors. Attached Figure Description

[0012] Figure 1 This is a control block diagram of the PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems of the present invention; Figure 2 This is a control block diagram of the combination of the multi-rotation reference coordinate system and the complex coefficient filter of the present invention; Figure 3 This is the invention AdamW Optimization algorithm flowchart. Detailed Implementation

[0013] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings.

[0014] Example 1 This invention is based on multiple rotating coordinate systems. PMSLM The adaptive optimization method for harmonic current suppression is as follows: Step 1: Establish a permanent magnet synchronous linear motor model and acquire the phase current signal of the motor through a current Hall sensor. After coordinate transformation, the following is obtained d shaft current i d and q shaft current iq .

[0015] Step 2: In the digital signal processing unit, based on the theory of multiple rotating reference coordinate systems, independent rotating coordinate systems are established for the target harmonic orders (such as the 5th, 7th, 11th, and 13th). Complex coefficient filters are used in each rotating coordinate system to replace traditional low-pass filters, filtering out non-target frequency components and achieving high-precision separation of the fundamental wave and each harmonic order.

[0016] Step 3: Construct the optimal harmonic reference current solution model for harmonic injection of the linear motor, and use the adaptive moment estimation algorithm with weighted attenuation ( Adaptive Moment Estimation with weight decay, AdamW Iterative optimization is performed, and overfitting is suppressed through a weight decay strategy, which can quickly, stably and accurately obtain the optimal solution.

[0017] Step 4: Use a proportional resonant controller, according to AdamW The error between the optimal harmonic reference current obtained by the optimization algorithm and the actual current value extracted by the complex coefficient filter is used as... PR Controller input, settings PR The controller's resonant frequency and proportional coefficient generate a harmonic compensation voltage, which is injected into the feedforward channel. This voltage, together with the fundamental voltage, acts on the inverter to drive the linear motor, thereby compensating for harmonic currents and suppressing them.

[0018] Example 2 This invention is based on a PMSLM adaptive optimization method for harmonic current suppression using multiple rotating coordinate systems, and is implemented according to the following steps: Step 1: Establish a permanent magnet synchronous linear motor model and acquire the phase current signal of the motor through a current Hall sensor. After coordinate transformation, the following is obtained d shaft current i d and q shaft current i q Specifically: Modeling the permanent magnet synchronous linear motor yields the voltage balance equation, as shown in equation (1): (1); in, , They are respectively d shaft voltage 、q Shaft voltage; , They are respectively d Shaft inductance, q Shaft inductance; It is a permanent magnet flux linkage. This indicates the speed of the mover relative to the stator; Indicates polar moment, This refers to the stator winding resistance. , They are respectively d shaft current 、q shaft current; The actual phase current of the linear motor is collected using a Hall effect current sensor. Three-phase current After coordinate transformation, we obtain d shaft current and q shaft current As shown in equation (2): (2); in, for d shaft and q The included angle of the axis; Step 2: In the digital signal processing unit, based on the theory of multiple rotating reference coordinate systems, independent rotating coordinate systems are established for the target harmonic orders (e.g., 5th, 7th, 11th, and 13th). Complex coefficient filters are used in each rotating coordinate system to replace traditional low-pass filters, filtering out non-target frequency components and achieving high-precision separation of the fundamental frequency from each harmonic order. Specifically: Step 2.1: Using the multi-rotating reference coordinate system theory, establish a synchronously rotating coordinate system for the specific harmonics to be suppressed (such as the 5th, 7th, 11th, 13th, etc.), and define the fundamental angular velocity as... (in, f (where 1 is the fundamental frequency of the linear motor), then the... n First harmonic ( n The angular velocity of the rotating coordinate system corresponding to the target harmonic order is set to... Since the rotation speed of the coordinate system is synchronized with the electric angular velocity of the harmonic components, the current is a DC current.

[0019] Step 2.2, establish the complex coefficient filter model, as shown in equation (3); (3) in, s It is a complex variable in the Laplace transform. j It is the imaginary unit. The angular frequency of the target signal that the filter needs to extract; k a The attenuation coefficient is... k a The value needs to be chosen in a trade-off between filtering accuracy and dynamic response: a smaller value... k a For example, a value of (0.01-0.05) can achieve better high-frequency suppression and steady-state accuracy, but the dynamic response is slow; a larger value...k a (e.g., 0.05-2) can speed up the response, but weakens the anti-interference capability. Therefore... k a The value is adjusted to be around 0.05.

[0020] The input to this transfer function contains harmonics of various orders. d, q shaft current ， yes After passing through an angular velocity of After rotating the coordinate system PARK The output obtained by the transformation is a pure harmonic current containing only the target order harmonics. That is .

[0021] The poles of the transfer function are the roots of the denominator: The poles have complex coordinates: the real part is The imaginary part is When the angular frequency of the input signal is close to At this time, the filter's response to the signal is strongest. (Real part) The damping characteristics of the poles are determined when When the value is greater than 0, the real part is negative, and the transient response will decay. k a The larger the value, the larger the absolute value of the real part, the stronger the damping, and the better the response to frequency deviation. The signal decays faster, and the transient response decays even faster.

[0022] The harmonic current signals of each order are input into the corresponding complex coefficient filters. The complex coefficient filters are used to replace the traditional low-pass filters to extract the harmonic current components in each harmonic rotating coordinate system.

[0023] Complex coefficient filters are bandpass filters designed in the complex domain. Their transfer function exhibits asymmetry in the complex coordinate system, thus possessing the inherent ability to selectively extract positive-order or negative-order rotation vectors. Compared to traditional... LPF (The phase lag is usually >30°), while the phase lag of the complex coefficient filter can be controlled within 5°, and the dynamic response speed is improved by 2-3 times.

[0024] Step 3: Construct the optimal harmonic reference current solution model for harmonic injection of the linear motor, and use the adaptive moment estimation algorithm with weighted attenuation ( Adaptive Moment Estimation with weight decay, AdamW Iterative optimization is performed, and overfitting is suppressed through a weight decay strategy, which can quickly, stably and accurately obtain the optimal solution.

[0025] First, by constructing a system with the total harmonic distortion of the current (THD) as the threshold value, the system can achieve the desired effect. THDMinimizing the loss function, which reflects the magnitude of harmonic content in the motor output current, transforms the harmonic current suppression problem into an optimization problem. THD The smaller the value, the closer the current waveform is to an ideal sine wave. AdamW In each iteration, the algorithm evaluates the impact of the current control parameters on harmonic suppression by calculating the gradient of the loss function relative to each harmonic injection parameter (amplitude and phase). Then, it dynamically adjusts the learning rate using the first moment (weighted average of gradients) and the second moment (weighted average of squared gradients), ensuring that parameter updates have both global convergence and avoid oscillations. Simultaneously, a weight decay term suppresses overfitting, preventing the parameters from growing indefinitely. Through this process, AdamW The algorithm continuously updates the harmonic injection parameters, gradually reducing the loss function value. THD When the change approaches zero, the algorithm considers it to have converged to the optimal solution, thus obtaining the optimal harmonic reference current.

[0026] Specifically: Step 3.1, to achieve harmonic minimization control, establish a system based on the total harmonic distortion of the current (THD). THD The least squares loss function with the objective of minimizing is shown in equation (4): (4) in, for AdamW The parameter vector for algorithm optimization; The calculated number of... n The target reference current for the first harmonic; For the first n The actual harmonic current obtained from first harmonic measurement; H This is the set of harmonic orders that need to be suppressed, for example, {5,7,11,13}; Step 3.2, through AdamW Optimize the parameter configuration of the algorithm; (1) Calculate the first t The gradient of the step (the partial derivative of the loss function with respect to the parameters) is shown in equation (5): (5) (2) Update the first-order moment estimate, as shown in equation (6): (6) (3) Update the second-order moment estimate, as shown in equation (7): (7) (4) Correct the bias of the first moment estimate, as shown in equation (8): (8) (5) Correct the bias of the second moment estimate, as shown in equation (9): (9) (6) Update parameters As shown in equation (10): (10) (7) Determine the convergence condition. The iteration terminates when any of the following conditions (11) are met: (11) in, for In the t Parameter values ​​at the next iteration The learning rate (usually set to 0.001-0.01, controlling the step size of parameter updates; too large a value can cause oscillations, while too small a value can lead to slow convergence). and These are the first-order exponential decay coefficient and the second-order exponential decay coefficient (usually...). Take 0.9, Take 0.99), It is a very small value to prevent the denominator from being 0 when updating parameters (usually taken as...). ), The gradient of the loss function. This is the weight decay coefficient (usually set to 0.01-0.1 to prevent overfitting). N is the number of iterations for continuous detection (usually set to 10). The convergence accuracy threshold is set to 0.1% in this invention. Maximum number of iterations (200 in this invention). yes k Total harmonic distortion of current at any given time (tc) THD ).

[0027] When the algorithm converges, we get This is the optimal parameter vector, i.e., the optimal harmonic reference current.

[0028] Step 4: Use proportional resonance ( Proportional Resonant, PR ) controller, according to AdamW The error between the optimal harmonic reference current obtained by the optimization algorithm and the actual current value extracted by the complex coefficient filter is used as... PR Controller input, settings PR The controller's resonant frequency and proportional coefficient generate a harmonic compensation voltage, which is injected into the feedforward channel. This voltage, together with the fundamental voltage, acts on the inverter to drive the linear motor, thereby compensating for harmonic currents and suppressing them.

[0029] Specifically: Step 4.1, Build PRThe transfer function of the controller is shown in equation (12): (12) in, The resonant frequency, For cutoff bandwidth, K p This is the proportionality coefficient. K r Resonance coefficient, s It is a complex variable in the Laplace transform. j It is the imaginary unit.

[0030] Step 4.2, take AdamW The error between the optimal harmonic reference current obtained by the optimization algorithm and the actual current extracted by the complex coefficient filter is used as... PR Controller input, PR Controller output harmonic compensation voltage , PR Harmonic compensation voltage output by the controller With the motor feedforward stage Generate after superposition and coordinate transformation , It is applied to the inverter to drive a linear motor, enabling precise and rapid suppression of harmonic currents.

[0031] The coordinate transformation angle can be obtained by calculating the mover displacement. The mover velocity reference value is also available. With the measured value of the moving part velocity The difference signal is obtained after comparison, and then used as the input of the velocity loop proportional-integral controller to obtain the thrust current reference value. . and respectively with and The current deviation is obtained by comparison, and the current deviation is then processed by the proportional-integral regulator of the current loop. and .

[0032] Example 3 Figure 1 This is a control block diagram of the PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems of the present invention, which includes an inner current loop, an outer velocity loop, and... SVPWM Modules; This invention employs In the vector control method, during system operation, the current is first sampled to obtain... Then, the thrust current feedback value is obtained through coordinate transformation. Excitation current feedback value The angle of coordinate transformation. The reference value of the mover velocity can be obtained through mover displacement calculation. With the measured value of the moving part velocity The difference signal is obtained after comparison, and then used as the input of the velocity loop proportional-integral controller to obtain the thrust current reference value. , respectively with The current deviation is obtained by comparison, and the current deviation is then processed by the proportional-integral regulator of the current loop. Then through Park Inverse transformation yields Meanwhile, the harmonic detection module extracts the harmonic current components of each order through a multi-rotating reference coordinate system and a complex coefficient filter, using... AdamW The optimization algorithm obtains the optimal harmonic reference current, and then... PR The controller outputs a harmonic compensation voltage, which, when superimposed on the fundamental voltage, acts together with the fundamental voltage. SVPWM The modulation module will generate PWM The signal is input to the inverter, which drives the permanent magnet synchronous linear motor to achieve rapid suppression of harmonic current and high-precision operation control of the system.

[0033] Example 4 Figure 2 This invention presents a control block diagram combining multiple rotating reference coordinate systems and complex coefficient filters. For specific harmonics (e.g., 5th, 7th, 11th, and 13th orders) in the current of a permanent magnet synchronous motor, multiple reference coordinate systems rotating at corresponding angular velocities are established. This allows each order of harmonic to be DC-DC-ized within its respective rotating coordinate system, achieving spatial separation of harmonics. Subsequently, a complex coefficient filter is introduced in each rotating coordinate system to replace the traditional low-pass filter. Utilizing the selective bandpass characteristic of complex poles on positive-sequence or negative-sequence rotating vectors, high-precision extraction of the target harmonic components is achieved. This combined method overcomes the shortcomings of traditional low-pass filters, such as large phase lag and poor accuracy, enabling harmonic current detection to have faster dynamic response and higher accuracy, providing accurate harmonic information for subsequent optimal harmonic injection and compensation control.

[0034] Example 5 Figure 3 In this invention AdamW The algorithm's flowchart. First, a mechanism is constructed to minimize the total harmonic distortion (THD) of the current. THD The loss function is defined as the objective function (based on the sum of squared deviations between the harmonic reference current and the actual measured current for each harmonic); subsequently, parameters such as the harmonic reference current are initialized, and the learning rate is set. First-order moment attenuation coefficient Second-order moment attenuation coefficient Weight decay coefficient The hyperparameters are initialized and the iteration count is set to zero. Then, the gradient of the loss function with respect to the current harmonic reference current is iteratively calculated. The first moment (accumulating gradient momentum information to accelerate the update of the stable direction) and the second moment (accumulating the magnitude information of the squared gradient to assign adaptive step sizes to different harmonic reference currents) are updated using an exponentially weighted moving average. The first and second moments are then corrected for bias (eliminating early estimation bias caused by initial values ​​of zero). The harmonic reference current is then updated based on the corrected moment estimates. Finally, convergence is checked, such as whether the loss change and parameter update magnitude meet the threshold. If convergence is not achieved, the iteration count is incremented and the loop continues. If convergence is achieved, the loop terminates and the optimal harmonic reference current is output. This reference current serves as... PR The controller input generates a compensation voltage, which, when superimposed on the fundamental voltage, drives the motor to achieve precise harmonic compensation.

[0035] Example 6 The method of this invention achieves efficient suppression of current harmonics in permanent magnet synchronous linear motors, solves the defects of slow convergence and easy getting trapped in local optima in traditional gradient descent algorithms, and suppresses overfitting through weight decay strategy, so as to obtain the optimal solution efficiently and stably.

Claims

1. A PMSLM adaptive optimization method for harmonic current suppression based on multiple rotating coordinate systems, characterized in that, The specific steps are as follows: Step 1: Establish a permanent magnet synchronous linear motor model. Obtain the phase current signal of the motor through a current Hall sensor, and obtain it through coordinate transformation. d shaft and q shaft current; Step 2: Establish independent rotating coordinate systems for the target harmonic order, and use complex coefficient filters in each rotating coordinate system to output the actual harmonic current; Step 3: Construct an optimal harmonic reference current solution model for harmonic injection into a linear motor, through... AdamW The algorithm is iteratively optimized to obtain the optimal harmonic reference current; Step 4: Use the error between the optimal harmonic reference current and the actual harmonic current extracted by the complex coefficient filter as... PR The controller input generates a harmonic compensation voltage, which is injected into the feedforward channel and works together with the fundamental voltage on the inverter to compensate for the harmonic current.

2. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 1, characterized in that, In step 1, specifically: Modeling the permanent magnet synchronous linear motor yields the voltage balance equation, as shown in equation (1): （1）； in, , They are respectively d shaft voltage 、q Shaft voltage; , They are respectively d Shaft inductance, q Shaft inductance; It is a permanent magnet flux linkage. This indicates the speed of the mover relative to the stator; Indicates polar moment, This refers to the stator winding resistance. , They are respectively d shaft current 、q shaft current; The actual phase current of the linear motor is collected using a Hall effect current sensor. Three-phase current After coordinate transformation, we obtain d shaft current and q shaft current .

3. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 2, characterized in that, The d shaft current and q shaft current The calculation formula is shown in equation (2): （2）； in, for d shaft and q The included angle of the axis.

4. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 3, characterized in that, Step 2 specifically involves: Step 2.1: Using the multi-rotating reference coordinate system theory, establish a synchronously rotating coordinate system for the specific harmonics to be suppressed, defining the fundamental angular velocity as... ,in, f If 1 is the fundamental frequency of the linear motor, then the... n The angular velocity of the rotating coordinate system corresponding to the first harmonic is set as follows: ; Step 2.2, establish the complex coefficient filter model, as shown in equation (3); （3） in, s It is a complex variable in the Laplace transform. j It is the imaginary unit. k a The attenuation coefficient; The input to this transfer function contains harmonics of various orders. d,q shaft current ， yes After passing through an angular velocity of After rotating the coordinate system PARK The output obtained by the transformation is the actual harmonic current containing the target order harmonics. That is .

5. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 4, characterized in that, Step 3 specifically involves: Step 3.1, in order to achieve harmonic minimization control, a least-squares loss function is established with the objective of minimizing the total harmonic distortion rate of the current, as shown in equation (4): （4） in, for AdamW The parameter vector for algorithm optimization; The calculated number of... n The target reference current for the first harmonic; H This refers to the set of harmonic orders that need to be suppressed. Step 3.2, through AdamW The parameter configuration of the optimization algorithm is obtained when the algorithm converges. This is the optimal parameter vector, i.e., the optimal harmonic reference current.

6. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 5, characterized in that, In step 3.2, specifically, it involves: through AdamW Optimize the parameter configuration of the algorithm; (1) Calculate the first t The gradient of the step is shown in equation (5): （5） (2) Update the first-order moment estimate, as shown in equation (6): （6） (3) Update the second-order moment estimate, as shown in equation (7): （7） (4) Correct the bias of the first moment estimate, as shown in equation (8): （8） (5) Correct the bias of the second moment estimate, as shown in equation (9): （9） (6) Update parameters As shown in equation (10): （10） (7) Determine the convergence condition. The iteration terminates when any of the following conditions (11) are met: （11） in, for In the t Parameter values ​​at the next iteration For learning rate, and These are the first-order exponential decay coefficient and the second-order exponential decay coefficient, respectively. , The gradient of the loss function. Here, represents the weight decay coefficient, and N represents the number of iterations for continuous detection. This is the convergence accuracy threshold; Maximum number of iterations yes k Total harmonic distortion of current at any given time.

7. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 6, characterized in that, Step 4 specifically involves: Step 4.1, Build PR The transfer function of the controller is shown in equation (12): （12） in, The resonant frequency, For cutoff bandwidth, K p This is the proportionality coefficient. K r Resonance coefficient; Step 4.2, take AdamW The error between the optimal harmonic reference current obtained by the optimization algorithm and the actual current extracted by the complex coefficient filter is used as... PR Controller input, PR Controller output harmonic compensation voltage , PR Harmonic compensation voltage output by the controller With the motor feedforward stage Generate after superposition and coordinate transformation , It is applied to the inverter to drive a linear motor, enabling precise and rapid suppression of harmonic currents.

8. The PMSLM adaptive optimization harmonic current suppression method based on multiple rotating coordinate systems as described in claim 7, characterized in that, In step 4.2, the motor feedforward circuit... The calculation process is as follows: reference value of moving part velocity With the measured value of the moving part velocity The difference signal is obtained after comparison, and then used as the input of the velocity loop proportional-integral controller to obtain the thrust current reference value. ; and respectively with and The current deviation is obtained by comparison, and the current deviation is then processed by the proportional-integral regulator of the current loop. and .

Images