Rotating speed control method and system under load interference of magnetic suspension turbo molecular pump

Abstract

The invention discloses a rotating speed control method and system for a magnetic suspension turbo molecular pump under load disturbance. The method comprises the following steps: constructing a mathematical model based on a rotor dynamic behavior of a driving motor of the magnetic suspension turbo molecular pump under load disturbance; the driving motor is a permanent magnet synchronous motor; constructing a bandwidth-variable self-adaptive linear expansion state observer based on the mathematical model, wherein the bandwidth-variable self-adaptive linear expansion state observer is used for observing the angular velocity and lumped disturbance of a motor rotor; the observer is improved by increasing a rotating speed observation error compensation item; the bandwidth adaptive rate and the coefficient of the observation error compensation item are optimized based on a reinforcement learning algorithm; and when load disturbance is generated, observing and obtaining a lumped disturbance estimated value based on the optimized observer, calculating a current feed-forward compensation amount, and carrying out feed-forward compensation on the disturbance. According to the invention, through the improved adaptive extended state observer, accurate estimation and feed-forward compensation are carried out on disturbance, and the control performance of the permanent magnet synchronous motor is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of permanent magnet synchronous motor control, and more particularly to a speed control method and system for a magnetic suspension turbo-molecular pump under load disturbance. BACKGROUND

[0002] A magnetic suspension turbo-molecular pump is an important device for providing a high vacuum environment and is widely used in scientific exploration, space technology, integrated circuit manufacturing, and other fields. A magnetic suspension turbo-molecular pump obtains a high vacuum environment by transferring the momentum of a high-speed rotating blade to thin gas molecules and making them move in a directional manner. Therefore, in order to obtain better vacuum performance, the turbo rotor needs to have a relatively high speed.

[0003] In the semiconductor manufacturing process, the pressure in the cavity is not stable, but changes with the production process, which changes the gas medium in the cavity, thereby causing the cavity pressure to change. Therefore, the load of the impeller and the motor often changes with the process. When the load disturbance occurs, suppressing the disturbance and reducing the speed fluctuation not only can maintain a constant pumping speed, but also can reduce the disturbance to the magnetic suspension bearing, reduce the risk of rotor instability, and ensure the reliable and stable operation of the magnetic suspension turbo-molecular pump.

[0004] Therefore, it is an urgent problem for those skilled in the art to provide a method for suppressing disturbance and reducing speed fluctuation when load disturbance occurs. SUMMARY

[0005] Therefore, the present application provides a speed control method and system for a magnetic suspension turbo-molecular pump under load disturbance, which observes the load disturbance of a permanent magnet synchronous motor and reduces the adverse effects of the load disturbance through feedforward compensation, thereby improving the observation accuracy of the disturbance and the control performance of the permanent magnet synchronous motor.

[0006] To achieve the above object, the present application adopts the following technical solutions: In a first aspect, the present application provides a speed control method for a magnetic suspension turbo-molecular pump under load disturbance, comprising the following steps: S1, constructing a mathematical model based on the rotor dynamic behavior of a drive motor of the magnetic suspension turbo-molecular pump under load disturbance; the drive motor is a permanent magnet synchronous motor; S2, constructing a bandwidth-variable adaptive linear extended state observer based on the mathematical model, for observing the angular velocity of the motor rotor and the lumped disturbance; S3, improving the observer by adding a speed observation error compensation term; and optimizing the bandwidth adaptive rate of the observer and the coefficient of the observation error compensation term based on a reinforcement learning algorithm; S4, when the load disturbance, for the improved and optimized observer observation magnetic suspension turbine molecular pump and get the lumped disturbance estimate, disturbance feedforward compensation, improve the permanent magnet synchronous motor load disturbance under the angular velocity.

[0007] Further, in step S1, the mathematical model is:

[0008] wherein, is the angular acceleration; J is the moment of inertia of the turbine rotor; is the angular velocity of the rotor; is the load torque of the rotor; is the cogging torque; is the electromagnetic torque output by the motor; is q the stator current of the shaft, is the permanent magnet flux linkage, p is the number of pole pairs of the motor; is the lumped disturbance of the permanent magnet synchronous motor speed loop.

[0009] Further, in step S2, the bandwidth-variable adaptive linear extended state observer is expressed by the formula:

[0010] gain and satisfy the condition:

[0011]

[0012] wherein, is the estimated value of the angular velocity, is the estimate of the lumped disturbance; and are the derivatives of and respectively; and are the speed observer and lumped disturbance observer gains respectively; is the estimation error; is the controller output signal; is the bandwidth of the observer.

[0013] Further, in step S3, the observer is improved by adding a speed observation error compensation term; the improved observer is expressed by the formula:

[0014] gain 、 and satisfy the condition:

[0015]

[0016]

[0017]

[0018] Substituting the formula, the transfer function expressions of and are:

[0019] wherein, is the adjustment gain and is the proportional coefficient, and s is a complex frequency domain variable in the Laplace transform of the transfer function.

[0020] Further, the bandwidth of the observer is adaptively adjusted based on the estimated value of the angular velocity and the bandwidth adaptive parameter; and the expression is:

[0021]

[0022] wherein, is the bandwidth adaptive parameter, and are the upper and lower limit values of the observer bandwidth.

[0023] Further, in step S3, the bandwidth adaptive rate and the coefficient of the observation error compensation term are optimized based on the reinforcement learning algorithm; specifically including: The deep deterministic policy gradient algorithm is used to optimize the parameters and . The deep deterministic policy gradient algorithm includes a policy network and a value network; the policy network gives the suggested value of the parameters and at the current time according to the current operating state of the permanent magnet synchronous motor; the value network calculates the action value according to the current operating state and the parameter suggestion value, and evaluates the output parameter suggestion value according to the reward function; the current operating state includes the reference value and the feedback value of the current and the angular velocity.

[0024] Further, the expression of the reward function is:

[0025]

[0026] wherein R is a reward function value; , and are reward, slight penalty and major penalty values, respectively; is a reference value of angular velocity; is a feedback value of angular velocity.

[0027] Further, the step S4 specifically comprises: when the load is disturbed, observing the magnetic suspension turbo molecular pump based on the improved and optimized observer and obtaining a lumped disturbance estimation value ;a current feedforward compensation amount is calculated ; the current feedforward compensation amount is superimposed on the current reference value of the input permanent magnet synchronous motor, so as to realize early cancellation of the disturbance of the permanent magnet synchronous motor and improve the rotor angular velocity.

[0028] In a second aspect, the present application provides a speed control system under load disturbance of a magnetic suspension turbo molecular pump, which adopts the speed control method under load disturbance of the magnetic suspension turbo molecular pump according to any one of the first aspect, and comprises the following modules: a model construction module: used for constructing a mathematical model based on the rotor dynamic behavior of a driving motor of the magnetic suspension turbo molecular pump under load disturbance; the driving motor is a permanent magnet synchronous motor; an observer construction module: used for constructing a bandwidth-variable adaptive linear extended state observer based on the mathematical model, for observing the angular velocity of the motor rotor and the lumped disturbance; an observer improvement module: used for improving the observer by adding a rotor speed observation error compensation term; and optimizing the bandwidth self-adaptive rate of the observer and the coefficient of the observation error compensation term based on a reinforcement learning algorithm; a feedforward compensation module: when the load is disturbed, used for observing the magnetic suspension turbo molecular pump based on the improved and optimized observer and obtaining a lumped disturbance estimation value, calculating a current feedforward compensation amount, and performing feedforward compensation on the disturbance.

[0029] According to the above technical solution, compared with the prior art, the present application provides a speed control method under load disturbance of a magnetic suspension turbo molecular pump, which has the following beneficial effects: The present application utilizes an adaptive linear extended state observer to observe the load disturbance of a permanent magnet synchronous motor, and reduces the adverse effects caused by the load disturbance through feedforward compensation.

[0030] In order to improve the performance of the adaptive linear extended state observer, the speed observation error compensation term is added in the disturbance observation to improve the observer, so that the observation effect of the observer on the complex disturbance is ensured. The adaptive rate and the coefficient of the observation error compensation term of the improved adaptive linear extended state observer are optimized and trained by using the reinforcement learning algorithm, so that the disturbance observation performance is better and the adverse effects of noise on the system are reduced. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0032] Figure 1 A flow chart of a speed control method of a magnetic suspension turbo molecular pump under load disturbance is provided for the embodiments of the present application.

[0033] Figure 2 A structure diagram of a bandwidth-variable adaptive linear extended state observer is provided for the embodiments of the present application.

[0034] Figure 3 A principle block diagram of a deep deterministic policy gradient algorithm is provided for the embodiments of the present application.

[0035] Figure 4 A structure diagram of a reinforcement learning algorithm is provided for the embodiments of the present application.

[0036] Figure 5 A block diagram of a permanent magnet synchronous motor control system of a magnetic suspension turbo molecular pump is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0038] Embodiment 1 The embodiments of the present application disclose a speed control method of a magnetic suspension turbo molecular pump under load disturbance, as shown in Figure 1 , including the following steps: S1, obtaining the rotor dynamic behavior of the driving motor of the magnetic suspension turbo molecular pump under load disturbance and constructing a mathematical model; the driving motor is a permanent magnet synchronous motor; S2, constructing a bandwidth-variable adaptive linear extended state observer based on the mathematical model, for observing the angular velocity of the motor rotor and the lumped disturbance; S3, improving the observer by adding a rotational speed observation error compensation term; and optimizing the bandwidth self-adaptation rate of the observer and the coefficient of the observation error compensation term based on a reinforcement learning algorithm; S4, when the load is disturbed, observing the magnetic suspension turbo molecular pump based on the improved and optimized observer and obtaining the lumped disturbance estimation value, to feed forward compensate the disturbance and improve the angular velocity of the permanent magnet synchronous motor under load disturbance.

[0039] The embodiment is applied to semiconductor manufacturing and used for maintaining a high-vacuum environment in a semiconductor etching device. In a plasma etching process, a magnetic suspension turbo molecular pump is used to achieve a high-vacuum environment, and the rotor thereof needs to rotate at a super-high speed of 60,000 r / min to continuously pump away gas molecules in the cavity. However, during the etching process, process gases such as CF4 and O2 are periodically injected and discharged, causing the pressure in the cavity to fluctuate dramatically and further causing the aerodynamic load on the turbine blade to frequently change. This load disturbance can cause the rotor speed to fluctuate, affect the pumping stability, and even cause the magnetic suspension bearing control to be unstable. The control method of the “adaptive extended state observer + feed forward compensation based on reinforcement learning optimization” proposed in the embodiment is integrated into the drive control system of the magnetic suspension turbo molecular pump, and the specific implementation process is as follows: Step 1, constructing a mathematical model of a permanent magnet synchronous motor.

[0040] In the embodiment, the rotor dynamic behavior of the drive motor of the magnetic suspension turbo molecular pump under load disturbance is the basis for constructing the mathematical model of the permanent magnet synchronous motor.

[0041] During the operation of the permanent magnet synchronous motor, two forces are usually involved. First, the disturbance torques such as load torque, friction torque, and cogging torque hinder the smooth rotation of the motor rotor. In addition, the electromagnetic torque generated after the current is passed through the three-phase winding provides energy for the motor to overcome the disturbance torque and drive the motor to rotate smoothly.

[0042] The magnetic suspension turbo molecular pump of the embodiment has the characteristic of no friction, and the mathematical model of the permanent magnet synchronous motor thereof is represented as:

[0043] wherein, is the angular acceleration; is the rotational inertia of the turbo rotor; is the angular velocity of the rotor; is the load torque of the rotor; is the cogging torque; The electromagnetic torque output by the motor; since the driving motor of the magnetic suspension turbo molecular pump is a surface-mounted permanent magnet synchronous motor, a field-oriented control (FOC) is adopted, the core idea of which is to convert the three-phase alternating current of the motor into a direct current through coordinate transformation, so that the control mode of the motor is similar to that of a direct current motor, realizing independent regulation of the flux linkage and torque. At this time, the driving control strategy of excitation current can be expressed as:

[0044] wherein, represents the stator current of the shaft, is the permanent magnet flux linkage, is the pole pair number of the motor.

[0045] From the above two formulas, we can get:

[0046] The model of the surface-mounted permanent magnet synchronous motor in the rotor field-oriented coordinate system is obtained.

[0047] Step 2, design a bandwidth-variable adaptive linear extended state observer to observe the speed and disturbance.

[0048] The motor model can be further arranged in this embodiment as follows:

[0049] wherein, is the lumped disturbance of the speed loop of the permanent magnet synchronous motor.

[0050] Before constructing the linear extended state observer for estimating the speed and lumped disturbance of the permanent magnet synchronous motor under load disturbance, the dynamic behavior of the system must be abstracted into a state space model.

[0051] In this embodiment, the state variable is selected, and wherein, x 1 = ω represents the angular speed of the motor rotor, which is the main state that needs to be accurately observed in this embodiment. x 2 =d w represents the lumped disturbance (i.e. the equivalent value of the load torque and the cogging torque), which is an unknown but external disturbance affecting the dynamic behavior of the system. f(t) represents the rate of change of the disturbance, which is assumed to be an unknown but bounded time function.

[0052] ​The embodiment will make the disturbance that is originally difficult to measure directly As an "extended state" of the system, so that it can be estimated by the observer.

[0053] Available:

[0054] Where, The derivative of the state vector x, i.e. the rate of change of state; y represents the output of the system; The output of the controller; The state transition matrix describes the coupling relationship between the internal states of the system; The input matrix represents the influence of the controller output u on the state; The output matrix determines which states can be measured; The disturbance matrix of the system shows how the external disturbance f(t) affects the system.

[0055] Based on the above state space model, a linear extended state observer is constructed, and its expression is:

[0056] Where, The estimate of The estimate of the lumped disturbance And And And The estimated error, And And The observer gain can be expressed as:

[0057] Where, The bandwidth of the linear extended state observer.

[0058] In the actual control process, in order to obtain better observer performance and improve the response speed of the observer, the bandwidth of the linear extended state observer needs to be increased. However, with the increase of the bandwidth, the observer will introduce noise into the control system. Therefore, when constructing the observer, a trade-off needs to be made between the observation performance and the measurement noise.

[0059] The embodiment constructs an adaptive linear extended state observer, which adaptively adjusts the bandwidth of the observer, and its adaptive rule can be expressed as:

[0060] And satisfy:

[0061] in, To set parameters for bandwidth adaptation, by adjusting... The adaptive adjustment enables adaptive adjustment of the observer bandwidth. and These are the upper and lower limits of the bandwidth for the adaptive linearly expanding state observer.

[0062] Step 3: First, the adaptive linear extended state observer is improved by adding a rotational speed observation error compensation term to the disturbance observation, thereby improving the disturbance observation performance.

[0063] The improved observer is expressed by the formula:

[0064] Among them, gain Conditions met: ,Right now:

[0065] in, Used to adjust gain and The proportion, and satisfying the condition: .

[0066] at this time, and The transfer function is expressed as:

[0067] From the above formula, we can see that The introduction of this does not affect the performance of rotational speed observation. However, when observing disturbances, The introduction of this increases the molecular order, thereby improving the observation performance of perturbations.

[0068] The structure of the adaptive linear extended state observer in this embodiment is referenced. Figure 2 As shown, the gain satisfies the following condition: , ,at the same time, The adaptive rate changes in real time as the rotational speed changes.

[0069] The amplitude and phase characteristics of the observer and the gain adjustment coefficient Simulations show that as... The increase in [value] improves the observer's ability to detect disturbances and also reduces phase delay. However, excessive [value] This will generate high-frequency fluctuations, causing the system to overshoot. The settings require a compromise. To ensure good speed tracking performance under different operating conditions, it is necessary to... and The values of the two parameters change according to the change of working conditions.

[0070] Secondly, a reinforcement learning algorithm is designed to optimize and train the coefficients of the bandwidth adaptive rate and the observation error compensation term.

[0071] In the embodiment, and The values of the two parameters are not given exactly, and need to be further determined. Using the traditional trial and error method to obtain the values of the two parameters needs a large number of tedious experimental processes, and it is difficult to obtain the best parameters under different working conditions.

[0072] In the embodiment, the reinforcement learning method is used to optimize and train the parameters and , so as to obtain better adaptive linear extended state observer performance.

[0073] In order to further implement the above technical solutions, the present application uses a deep deterministic policy gradient (DDPG) algorithm for reinforcement learning training.

[0074] As shown in Figure 3 , the DDPG contains a policy network Actor and a value network Critic. Among them, the Actor network determines the output action according to the state. The Critic network evaluates the action according to the state and the action output by the Actor. By comparing the evaluation result given by the Critic network and the reward value obtained by the reward function, the parameters of the Critic network and the Actor network can be optimized and adjusted. In the process of continuous interaction between the agent and the environment, the learning parameters of the two networks will be updated until the policy converges to a relatively ideal situation.

[0075] Compared with the traditional method, the reinforcement learning parameter optimization is not to find the optimal parameter under a certain condition, but to learn the parameter setting strategy through the interaction between itself and the model. Therefore, when the optimization target or condition changes, the optimal parameter can still be quickly calculated.

[0076] As shown in Figure 4 , in the embodiment, the adaptive linear extended state observer based on reinforcement learning takes the reference value and the feedback value of the current and the speed as the observation value of the environment, and takes the bandwidth setting parameter and the coefficient of the observer error compensation term in the adaptive rule as the action value for reinforcement training.

[0077] In the embodiment, the reinforcement learning action value The conditions need to be met:

[0078] In order to further implement the above technical solutions, the reward function of the reinforcement learning algorithm of the application is designed as:

[0079]

[0080] Wherein, , and are reward, slight punishment and major punishment values respectively. is the reference value of angular velocity; is the feedback value of angular velocity. When the speed error , the agent approaches the target value and performs well, so a high reward of 10 is given. When the speed error , the agent deviates from the target value and performs poorly, so a slight punishment of-1 is given. When the speed error , the system speed control has a serious deviation, and the agent should be strongly constrained to trigger such a state, so a serious punishment of-100 is given.

[0081] Step 4, according to the improved and optimized observer observation of the estimated value of the lumped disturbance, the current feedforward compensation amount is calculated, the disturbance is feedforward compensated, and the angular velocity control precision under the load disturbance of the permanent magnet synchronous motor is improved.

[0082] In this embodiment, the lumped disturbance estimation value output by the improved and optimized extended state observer is used to calculate the corresponding current feedforward compensation amount, which is injected into the motor control system to offset the disturbance to the speed in advance, thereby significantly improving the angular velocity control precision.

[0083] Specifically includes: Extract disturbance information: get disturbance acceleration from the lumped disturbance estimation value output by the observer.

[0084] Torque mapping: multiply the moment of inertia J to get the equivalent disturbance torque.

[0085] Current conversion: according to the formula , the required compensation current is back calculated.

[0086] Feedforward injection: superimpose the compensation current on the q-axis current reference value to realize zero-delay disturbance cancellation.

[0087] In the specific implementation process of this embodiment, as shown in Figure 5 , the feedback current signal i of the permanent magnet synchronous motor is obtained through the current detector and the angular velocity detector.qfdb and the angular velocity of the magnetic suspension turbo molecular pump; is an angular velocity value observed by an adaptive linear extended state observer, the adaptive linear extended state observer performs state expansion according to a feedback current signal and an angular velocity to obtain a lumped disturbance of the system; and the current loop controller calculates a current feedforward compensation amount according to the disturbance observed by the adaptive linear extended state observer and forms a driving current acting on the permanent magnet synchronous motor.

[0088] The embodiment also includes a speed loop controller, which calculates a reference current loop signal according to an angular velocity and a reference angular velocity i qref and inputs the reference current loop signal to the current loop controller. The current loop controller calculates a driving current in combination with the feedforward compensation amount, the current loop reference signal and the feedback current signal.

[0089] The embodiment accurately estimates the disturbance by using the improved adaptive extended state observer; and the improved adaptive extended state observer parameters are optimized and trained by using the reinforcement learning algorithm, so that the control performance of the permanent magnet synchronous motor is improved by accurate estimation of the disturbance and feedforward compensation.

[0090] Embodiment 2 The embodiment of the application discloses a rotational speed control system of a magnetic suspension turbo molecular pump under load disturbance, adopts a rotational speed control method of a magnetic suspension turbo molecular pump under load disturbance as described in Embodiment 1, and includes the following modules. A model construction module is used to construct a mathematical model based on the rotor dynamic behavior of the driving motor of the magnetic suspension turbo molecular pump under load disturbance; the driving motor is a permanent magnet synchronous motor. An observer construction module is used to construct a bandwidth-variable adaptive linear extended state observer based on the mathematical model, for observing the angular velocity of the motor rotor and the lumped disturbance. An observer improvement module is used to improve the observer by adding a rotational speed observation error compensation term; and the bandwidth adaptive rate of the observer and the coefficient of the observation error compensation term are optimized based on a reinforcement learning algorithm. A feedforward compensation module is used to obtain a lumped disturbance estimation value of the magnetic suspension turbo molecular pump based on the improved and optimized observer when the load disturbance occurs, calculate a current feedforward compensation amount, and perform feedforward compensation on the disturbance.

[0091] The rotational speed control system of the embodiment is composed of four core function modules, forming a closed loop control chain of "modeling -> observation -> optimization -> compensation"; among them, the model construction module is responsible for constructing the mathematical model of the rotor dynamic behavior of the permanent magnet synchronous motor under load disturbance. The model needs to consider factors such as load disturbance, electromagnetic torque, cogging torque and moment of inertia. The observer construction module constructs a bandwidth-variable adaptive linear extended state observer based on the mathematical model; the observer is used to observe the angular velocity of the motor rotor and the lumped disturbance. The observer improvement module improves the observer by adding a rotational speed observation error compensation term, and optimizes the bandwidth adaptive rate of the observer and the coefficient of the observation error compensation term using reinforcement learning algorithm. When the load disturbance occurs, the feedforward compensation module uses the improved and optimized observer to observe the magnetic suspension turbo molecular pump to obtain the lumped disturbance estimate; and according to the observed disturbance estimate, the current feedforward compensation amount is calculated to feedforward compensate the disturbance.

[0092] The embodiment uses a current control signal superimposed with a feedforward compensation amount to control the operation of the permanent magnet synchronous motor. In actual operation, the performance of the control system is monitored and adjusted as needed.

[0093] Each of the embodiments in the specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0094] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for controlling the rotational speed of a magnetically levitated turbomolecular pump under load disturbance, characterized in that, Includes the following steps: S1. A mathematical model is constructed for the rotor dynamic behavior of the drive motor based on the magnetic levitation turbomolecular pump under load disturbance; the drive motor is a permanent magnet synchronous motor. S2. Based on the mathematical model, construct an adaptive linear extended state observer with variable bandwidth to observe the angular velocity and lumped disturbance of the motor rotor; S3. Improve the observer by adding a rotational speed observation error compensation term; and optimize the bandwidth adaptive rate and the coefficient of the observation error compensation term of the observer based on a reinforcement learning algorithm; S4. When the load is disturbed, the magnetic levitation turbomolecular pump is observed based on the improved and optimized observer and the lumped disturbance estimate is obtained. The current feedforward compensation is calculated and the disturbance is fedforward compensated.

2. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 1, characterized in that, In step S1, the mathematical model is: in, Angular acceleration; J The moment of inertia of the turbine rotor; ω is the angular velocity of the rotor; This represents the load torque of the rotor; This refers to the cogging torque; This refers to the electromagnetic torque output by the motor. for q Stator current of the shaft, It is a permanent magnet flux linkage. p This represents the number of pole pairs of the motor. This refers to the lumped disturbance in the speed loop of the permanent magnet synchronous motor.

3. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 1, characterized in that, In step S2, the bandwidth-variable adaptive linear extended state observer is expressed by the following formula: Gain and Conditions met: in, This is an estimate of the angular velocity. For the estimation of lumped disturbances; and They are respectively and The derivative; and These are the gains of the speed observer and the lumped disturbance observer, respectively. To estimate the deviation; Output signals to the controller; This represents the bandwidth of the observer.

4. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 3, characterized in that, In step S3, the observer is improved by adding a rotational speed observation error compensation term; the improved observer is expressed by the formula: Gain , and Conditions met: Substituting into the formula, we get and The transfer function is expressed as: in, To adjust the gain and The scaling factor is s, where s is the complex frequency domain variable in the Laplace transform of the transfer function.

5. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 4, characterized in that, The bandwidth of the observer is adaptively adjusted based on the estimated angular velocity and a bandwidth adaptive parameter; expressed by the formula: in, For bandwidth adaptive parameters, and These are the upper and lower limits of the observer bandwidth.

6. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 5, characterized in that, In step S3, the coefficients of the bandwidth adaptation rate and the observation error compensation term are optimized based on the reinforcement learning algorithm; specifically, this includes: A deep deterministic gradient algorithm is used to process the parameters. and Optimize; The deep deterministic policy gradient algorithm comprises a policy network and a value network; the policy network provides parameters for the current moment based on the current operating state of the permanent magnet synchronous motor. and The suggested values; the value network calculates the action value based on the current operating state and the suggested parameter values, and evaluates the suggested output parameter values ​​according to the reward function; the current operating state includes reference values ​​and feedback values ​​of current and angular velocity.

7. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 6, characterized in that, The expression for the reward function is: Where R is the reward function value; , and These are respectively reward, minor penalty, and major penalty values; This is a reference value for angular velocity; This is the feedback value of the angular velocity.

8. The speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in claim 1, characterized in that, Step S4 specifically includes: When the load is disturbed, the magnetic levitation turbomolecular pump is observed using the improved and optimized observer to obtain the lumped disturbance estimate. ; The current feedforward compensation amount is calculated. The current feedforward compensation is superimposed on the current reference value of the input permanent magnet synchronous motor to achieve early cancellation of disturbances to the permanent magnet synchronous motor and improve the rotor angular velocity.

9. A speed control system for a magnetically levitated turbomolecular pump under load disturbance, comprising the speed control method for a magnetically levitated turbomolecular pump under load disturbance as described in any one of claims 1-8, characterized in that, Includes the following modules: Model building module: used to construct a mathematical model of the rotor dynamic behavior of the drive motor of a magnetically levitated turbomolecular pump under load disturbance; the drive motor is a permanent magnet synchronous motor; Observer construction module: used to construct a bandwidth-variable adaptive linear extended state observer based on the mathematical model, for observing the angular velocity and lumped disturbance of the motor rotor; Observer improvement module: used to improve the observer by adding a rotational speed observation error compensation term; and to optimize the bandwidth adaptive rate and the coefficient of the observation error compensation term of the observer based on a reinforcement learning algorithm; Feedforward compensation module: When the load is disturbed, the magnetic levitation turbomolecular pump is observed based on the improved and optimized observer and the lumped disturbance estimate is obtained. The current feedforward compensation amount is calculated and the disturbance is fedforward compensated.

Images