Motor active damping noise reduction method, system, device and storage medium

Abstract

The application provides a motor active vibration reduction and noise reduction method, system, device and storage medium, and belongs to the technical field of motor vibration reduction. The method comprises the following steps: acquiring real-time noise of a motor; inputting the real-time noise into a trained flexible Actor-Critic model to obtain target harmonic injection times, a target phase and a target amplitude of a harmonic injection device; and generating a harmonic current for offsetting radial electromagnetic force of the motor based on the target harmonic injection times, the target phase and the target amplitude. The flexible Actor-Critic model outputs different harmonic parameters for different real-time noises of the motor, realizes active updating of the harmonic, and improves the noise reduction efficiency and accuracy of the motor vibration reduction and noise reduction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of motor vibration reduction technology, and in particular to a motor active vibration reduction and noise reduction method, system, device and storage medium. BACKGROUND

[0002] Motors are widely used in military, industrial, agricultural and other fields. During the operation of the motor, a lot of noise is often generated. According to the source of the noise, motor noise can be divided into three categories: aerodynamic noise, mechanical noise and electromagnetic noise. Among them, electromagnetic noise is considered to be the main source of motor noise, and radial electromagnetic waves are the main cause of motor vibration and noise. Electromagnetic noise is related to factors that are difficult to avoid, such as motor magnetic circuit structure, slot pole cooperation, winding distribution, current harmonics, eccentricity and material properties. And as long as the motor is running normally, corresponding electromagnetic vibration and noise will be generated.

[0003] The prior art mainly injects a specific number of harmonic currents to interact with the fundamental magnetic field generated by the permanent magnet to produce radial electromagnetic force of a specific spatial order and frequency, thereby weakening the vibration and noise caused by the electromagnetic force. This method of reducing noise is called harmonic injection method. However, due to the fact that the motor is used in different working conditions in actual use, the noise generated by the motor in different working conditions is different, that is, the harmonic injection parameters need to be adjusted according to the working condition adaptability. However, in the prior art, the adjustment of the harmonic injection parameters is mostly passive adjustment. Specifically, when the noise is detected to be too large, the harmonic injection parameters are adjusted manually to achieve vibration reduction and noise reduction. However, this method has the following technical problems: the adjustment of the harmonic injection parameters is not accurate enough, resulting in poor noise reduction effect.

[0004] Therefore, it is necessary to provide a motor active vibration reduction and noise reduction method, system, device and storage medium to accurately adjust the harmonic injection parameters and ensure the active noise reduction effect of the motor. SUMMARY

[0005] Therefore, it is necessary to provide a motor active vibration reduction and noise reduction method, system, device and storage medium to accurately adjust the harmonic injection parameters and ensure the active noise reduction effect of the motor.

[0006] On the one hand, in order to solve the above technical problems, the present application provides a motor active vibration reduction and noise reduction method, comprising:

[0007] obtaining the real-time noise of the motor;

[0008] inputting the real-time noise into the trained flexible Actor-Critic model to obtain the target harmonic injection frequency, target phase and target amplitude of the harmonic injection device;

[0009] Generate a harmonic current that cancels the radial electromagnetic force of the motor based on the target harmonic injection number, the target phase, and the target amplitude.

[0010] In a possible implementation, the flexible Actor-Critic model includes a soft Q network, a value network, and a policy network; before the real-time noise is input into the trained flexible Actor-Critic model, the method further includes:

[0011] Obtain a training sample; training data in the training sample is five-tuple data, and the five-tuple data includes a state at a first historical moment, an action at the first historical moment, a reward value at the first historical moment, a cumulative reward value at the first historical moment, and a state at a second historical moment; the second historical moment is a next moment of the first historical moment;

[0012] Input the training data into the soft Q network, the value network, and the policy network respectively, and correspondingly obtain a predicted reward value, a predicted cumulative reward value, and a predicted action;

[0013] Determine a reward loss value of the reward value at the first historical moment and the predicted reward value based on a first loss function of the soft Q network, and optimize network parameters of the soft Q network based on the reward loss value to obtain a target soft Q network;

[0014] Determine a predicted reward loss value of the cumulative reward value at the first historical moment and the predicted cumulative reward value based on a second loss function of the value network, and optimize network parameters of the value network based on the predicted reward loss value to obtain a target value network;

[0015] Determine an action loss value based on a third loss function of the policy network, the state at the first historical moment, and the predicted state, and optimize network parameters of the policy network based on the action loss value to obtain a target policy network;

[0016] The target soft Q network, the target value network, and the target policy network constitute a trained flexible Actor-Critic model.

[0017] In a possible implementation, when a training iteration number of the network parameters of the soft Q network reaches a preset number, or a training error is less than a preset error, or an error change trend is less than a preset change trend, the optimization of the network parameters of the soft Q network is stopped, and the target soft Q network is obtained.

[0018] In a possible implementation, the inputting the real-time noise into the trained flexible Actor-Critic model to obtain the target harmonic injection number, the target phase, and the target amplitude of the harmonic injection device includes:

[0019] determining the real-time state of the motor based on the real-time noise; the real-time state includes a real-time amplitude, a real-time noise frequency, a noise phase constant, and a real-time motor frequency of the motor;

[0020] processing the real-time state based on the target policy network to obtain a real-time target decision action, and controlling the harmonic injection device to execute the target decision action;

[0021] The target decision action includes the target harmonic injection number, the target phase, and the target amplitude.

[0022] In a possible implementation, the optimal policy of the policy network is:

[0023]

[0024]

[0025] wherein, is the optimal policy; is a state , and an action is a reward when is a discount factor of entropy; is an entropy when a state is ; is a policy function of the policy network; is an expectation of the cumulative reward and the maximum entropy; is a maximum operator.

[0026] In a possible implementation, the first loss function is:

[0027]

[0028] The second loss function is:

[0029]

[0030] The third loss function is:

[0031]

[0032] wherein, is the first loss function; is a predicted value of the soft Q network; is a true value of the soft Q network; is the second loss function; is the predicted value of the value network; is the true value of the value network; is the third loss function; is the coefficient; is the output value of the policy network.

[0033] In a possible implementation, the optimization algorithm used during training of the soft Q network, the value network, and the policy network is stochastic gradient descent.

[0034] On the other hand, the present invention also provides a motor active vibration and noise reduction system, comprising:

[0035] A real-time noise acquisition unit, used to acquire the real-time noise of the motor;

[0036] a harmonic injection device parameter determination unit, configured to input the real-time noise into a trained flexible Actor-Critic model to obtain a target harmonic injection order, a target phase, and a target amplitude of the harmonic injection device;

[0037] The vibration and noise reduction unit is configured to generate a harmonic current for counteracting the radial electromagnetic force of the motor based on the target harmonic injection order, the target phase, and the target amplitude.

[0038] On the other hand, the present invention also provides a motor active vibration and noise reduction device, including a memory and a processor, wherein:

[0039] The memory is used to store programs;

[0040] The processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the method for active vibration and noise reduction of a motor described in any one of the possible implementations above.

[0041] On the other hand, the present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the motor active vibration and noise reduction method described in any of the above possible implementation methods.

[0042] The present invention provides the following beneficial effects: The proposed method for active motor vibration and noise reduction uses a harmonic injection device as an intelligent agent. Based on a trained flexible actor-critic model, the parameters of the harmonic injection device (target harmonic injection order, target phase, and target amplitude) under varying real-time noise conditions are determined, achieving active vibration and noise reduction for the motor. Furthermore, the flexible actor-critic model incorporates the entropy of actions into the reward function, enhancing model stability and improving the agent's exploration capabilities. Furthermore, the flexible actor-critic model utilizes an offline policy update, reusing previously collected data to improve efficiency. This ensures the accuracy of the harmonic injection device while increasing injection efficiency, thereby enhancing both the effectiveness and efficiency of vibration and noise reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A schematic flow chart of an embodiment of the method for active vibration and noise reduction of a motor provided by the present invention;

[0045] Figure 2 A schematic diagram of the structure of an embodiment of the flexible Actor-Critic model provided by the present invention;

[0046] Figure 3 A schematic diagram of an embodiment of the process of training a flexible Actor-Critic model provided by the present invention;

[0047] Figure 4 For the present invention Figure 1 A schematic flow chart of an embodiment of obtaining the target harmonic injection number, target phase and target amplitude in step S102;

[0048] Figure 5 A schematic structural diagram of an embodiment of the active vibration and noise reduction system for a motor provided by the present invention;

[0049] Figure 6 This is a schematic structural diagram of an embodiment of the motor active vibration and noise reduction device provided by the present invention. DETAILED DESCRIPTION

[0050] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work are within the scope of protection of the present application.

[0051] It should be understood that the schematic drawings are not drawn to scale. The flowcharts used in the present application show the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can not be implemented in sequence, and the steps without logical context relationship can be reversed in sequence or implemented simultaneously. In addition, a person skilled in the art can add one or more other operations to the flowcharts or remove one or more operations from the flowcharts under the guidance of the content of the present application. Some block diagrams shown in the drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0052] Reference to "an embodiment" herein means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily independent or alternative embodiments to other embodiments. A person skilled in the art explicitly and implicitly understands that the embodiments described herein can be combined with other embodiments.

[0053] The present application provides a motor active vibration and noise reduction method, system, device and storage medium, which are described below respectively.

[0054] Figure 1 An embodiment flowchart of the motor active vibration and noise reduction method provided by the present application is shown in Figure 1 The motor active vibration and noise reduction method comprises:

[0055] S101, obtaining real-time noise of the motor;

[0056] S102, inputting the real-time noise into a trained Soft Actor-Critic (SAC) model to obtain target harmonic injection frequency, target phase and target amplitude of a harmonic injection device;

[0057] S103, generating a harmonic current for canceling radial electromagnetic force of the motor based on the target harmonic injection frequency, the target phase and the target amplitude.

[0058] Compared to existing technologies, the active motor vibration and noise reduction method provided by the present invention uses a harmonic injection device as an intelligent agent. Based on a trained flexible actor-critic model, the parameters of the harmonic injection device (target harmonic injection order, target phase, and target amplitude) under different real-time noise conditions are determined, thus achieving active vibration and noise reduction for the motor. Furthermore, the flexible actor-critic model incorporates the entropy of the action into the reward function, enhancing model stability and improving the agent's exploration capabilities. Furthermore, the flexible actor-critic model utilizes an offline policy update, reusing previously collected data to improve efficiency. This ensures the accuracy of the harmonic injection device while increasing injection efficiency, thereby enhancing both the effectiveness and efficiency of vibration and noise reduction.

[0059] In some embodiments of the present invention, Figure 2 As shown in Figure 3, the flexible Actor-Critic model includes a soft Q network, a value network, and a policy network.

[0060] It should be noted that the soft Q network and value network are the evaluation networks in the flexible Actor-Critic model. They are "critics" that do not take actions directly, but evaluate the quality of actions. The policy network is an "actor" that is used to determine decision actions based on the input state.

[0061] It should also be noted that: Figure 2 The solid arrows in the middle represent the process of forward parameter transfer, and the dotted arrows represent the process of backpropagation of parameters to update the network iteratively.

[0062] To ensure the validity of the flexible Actor-Critic model, before step S102, the flexible Actor-Critic model needs to be trained to obtain a trained flexible Actor-Critic model. Figure 3 As shown in Figure 2, the process of training the flexible Actor-Critic model is as follows:

[0063] S301. Obtain training samples. The training data in the training samples is a five-tuple data, which includes the state at the first historical moment, the action at the first historical moment, the reward value at the first historical moment, the accumulated reward value at the first historical moment, and the state at the second historical moment; the second historical moment is the moment after the first historical moment.

[0064] S302: Input the training data into the soft Q network, the value network, and the policy network, respectively, to obtain the predicted reward value, the predicted cumulative reward value, and the predicted action;

[0065] S303, determining a reward loss value of the reward value of the first historical moment and the predicted reward value based on a first loss function of the soft Q network, and optimizing the network parameters of the soft Q network based on the reward loss value to obtain a target soft Q network;

[0066] S304, determining a predicted reward loss value of the cumulative reward value of the first historical moment and the predicted cumulative reward value based on a second loss function of the value network, and optimizing the network parameters of the value network based on the predicted reward loss value to obtain a target value network;

[0067] S305, determining an action loss value based on a third loss function of the policy network, the state of the first historical moment and the predicted state, and optimizing the network parameters of the policy network based on the action loss value to obtain a target policy network;

[0068] The target soft Q network, the target value network and the target policy network constitute a trained flexible Actor-Critic model.

[0069] It should be noted that the execution order of steps S303, S304 and S305 is not limited to sequential execution, but can also be parallel execution or other sequential execution. For example, steps S304, S303 and S305 are executed in sequence, and only the parameters in the soft Q network, the value network and the policy network are trained and optimized.

[0070] In some embodiments of the present application, the training stop condition of the soft Q network includes but is not limited to that the training iteration number reaches a preset number, or the training error is less than a preset error, or the error change trend is less than a preset change trend. That is, when the training iteration number of the network parameters of the soft Q network reaches the preset number, or the training error is less than the preset error, or the error change trend is less than the preset change trend, the optimization of the network parameters of the soft Q network is stopped, and the target soft Q network is obtained.

[0071] Similarly, the training stop conditions of the policy network and the value network are the same as those of the soft Q network, which will not be described here.

[0072] In specific embodiments of the present application, the optimal policy of the policy network is:

[0073]

[0074]

[0075] In the formula, is the optimal policy; is the state , the action is the reward when is the discount factor of entropy; is the state entropy at the time; a policy function of a policy network; expectations of cumulative rewards and maximum entropy; a maximum operator.

[0076] It can be known from the formula of the optimal policy that the embodiment of the present application considers both the cumulative reward and the entropy value of the action in the policy optimization process, and has the dual goals of maximizing the reward and maximizing the entropy (exploration). In other words, the present application introduces an entropy regularization term, so that the policy has greater randomness when making decisions, thereby improving the exploration ability, and further improving the accuracy of the determined harmonic injection parameters and the efficiency of vibration and noise reduction.

[0077] In specific embodiments of the present application, the first loss function is:

[0078]

[0079] The second loss function is:

[0080]

[0081] The third loss function is:

[0082]

[0083] In the formula, is the first loss function; is the predicted value of the soft Q network; is the true value of the soft Q network; is the second loss function; is the predicted value of the value network; is the true value of the value network; is the third loss function; is a coefficient; is the output value of the policy network.

[0084] When training the flexible Actor-Critic model, the loss value is calculated according to the first loss function, the second loss function and the third loss function, and the policy network, the soft Q network and the value network are back propagated to update / optimize the network parameters until convergence, completing the training of the flexible Actor-Critic model.

[0085] In some embodiments of the present application, when the flexible Actor-Critic model is trained, as shown in Figure 4 the real-time noise is input into the trained flexible Actor-Critic model in step S102 to obtain the target harmonic injection frequency, the target phase and the target amplitude of the harmonic injection device, which includes:

[0086] S401, determining a real-time state of the motor based on the real-time noise; the real-time state includes a real-time amplitude, a real-time noise frequency, a noise phase constant and a real-time motor frequency of the motor;

[0087] S402, processing the real-time state based on a target policy network to obtain a real-time target decision action, and controlling the harmonic injection device to execute the target decision action;

[0088] The target decision action includes a target harmonic injection frequency, a target phase and a target amplitude.

[0089] With the change of the real-time noise, the target decision action also changes, realizing dynamic noise reduction of the real-time noise and ensuring the vibration and noise reduction effect.

[0090] It should be noted that: in the training process of the flexible Actor-Critic model, the optimization algorithm of the update / optimization process of the network parameters in the soft Q network, the value network and the policy network is the stochastic gradient descent method.

[0091] The embodiment of the application further improves the training speed of the flexible Actor-Critic model by setting the optimization algorithm as the stochastic gradient descent method, and further improves the vibration and noise reduction efficiency.

[0092] In summary, the motor active vibration and noise reduction method provided by the embodiment of the application takes the harmonic injection device as an intelligent agent, outputs specific harmonic frequency, phase and amplitude according to the noise of the motor to weaken the electromagnetic noise generated by the motor, realizes active update of the injected harmonic, and improves the vibration and noise reduction efficiency.

[0093] In order to better implement the motor active vibration and noise reduction method in the embodiment of the application, on the basis of the motor active vibration and noise reduction method, correspondingly, the embodiment of the application also provides a motor active vibration and noise reduction system, as shown in Figure 5 The motor active vibration and noise reduction system 500 includes:

[0094] A real-time noise acquisition unit 501 is configured to acquire the real-time noise of the motor.

[0095] A harmonic injection device parameter determination unit 502 is configured to input the real-time noise into the trained flexible Actor-Critic model to obtain a target harmonic injection frequency, a target phase and a target amplitude of the harmonic injection device.

[0096] A vibration and noise reduction unit 503 is configured to generate a harmonic current for offsetting the radial electromagnetic force of the motor based on the target harmonic injection frequency, the target phase and the target amplitude.

[0097] The motor active vibration and noise reduction system 500 provided by the above embodiments can implement the technical solutions described in the motor active vibration and noise reduction method embodiments, and the principles of implementation of the above modules or units can be referred to the corresponding content in the motor active vibration and noise reduction method embodiments, which will not be described here again.

[0098] As shown in Figure 6 The application also correspondingly provides a motor active vibration and noise reduction device 600. The motor active vibration and noise reduction device 600 includes a processor 601, a memory 602, and a display 603. Figure 6 Only part of the components of the motor active vibration and noise reduction device 600 are shown, but it should be understood that all the shown components are not required to be implemented, and more or less components can be alternatively implemented.

[0099] The processor 601 can be a central processing unit (CPU), a microprocessor, or other data processing chip in some embodiments, used to run the program code or process data stored in the memory 602, such as the motor active vibration and noise reduction method in the application.

[0100] In some embodiments of the application, the processor 601 can be a single server or a server group. The server group can be centralized or distributed. In some embodiments, the processor 601 can be local or remote. In some embodiments, the processor 601 can be implemented in a cloud platform. In an embodiment, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multiple cloud, etc., or any combination thereof.

[0101] The memory 602 can be an internal storage unit of the motor active vibration and noise reduction device 600 in some embodiments, such as a hard disk or memory of the motor active vibration and noise reduction device 600. The memory 602 can also be an external storage device of the motor active vibration and noise reduction device 600 in other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the motor active vibration and noise reduction device 600.

[0102] Further, the memory 602 can include both the internal storage unit and the external storage device of the motor active vibration and noise reduction device 600. The memory 602 is used to store application software and various data installed on the motor active vibration and noise reduction device 600.

[0103] The display 603 may, in some embodiments, be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) toucher, and the like. The display 603 is used to display information of the motor active damping and noise reduction device 600 and to display a visualized user interface. The components 601-603 of the motor active damping and noise reduction device 600 communicate with each other through a system bus.

[0104] In some embodiments of the present application, when the processor 601 executes the motor active damping and noise reduction program in the memory 602, the following steps can be implemented:

[0105] Obtaining real-time noise of the motor;

[0106] Inputting the real-time noise into the trained flexible Actor-Critic model to obtain a target harmonic injection frequency, a target phase, and a target amplitude of the harmonic injection device;

[0107] Generating a harmonic current for canceling radial electromagnetic force of the motor based on the target harmonic injection frequency, the target phase, and the target amplitude.

[0108] It should be understood that, in addition to the above functions, the processor 601 can also implement other functions when executing the motor active damping and noise reduction program in the memory 602. For details, refer to the description of the corresponding method embodiments.

[0109] Further, the type of the motor active damping and noise reduction device 600 referred to in the embodiments of the present application is not specifically limited, and the motor active damping and noise reduction device 600 can be a portable motor active damping and noise reduction device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop, and the like. Exemplary embodiments of the portable motor active damping and noise reduction device include, but are not limited to, a portable motor active damping and noise reduction device running an IOS, an Android, a Microsoft, or another operating system. The portable motor active damping and noise reduction device described above can also be another portable motor active damping and noise reduction device. It should also be understood that, in some other embodiments of the present application, the motor active damping and noise reduction device 600 can also not be a portable motor active damping and noise reduction device, but a desktop computer having a touch-sensitive surface (such as a touch panel).

[0110] Correspondingly, the embodiments of the present application also provide a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps or functions in the motor active damping and noise reduction method provided by the above method embodiments.

[0111] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiment methods can be instructed by a computer program to relevant hardware (such as a processor, a controller, etc.) to be completed, and the computer program can be stored in a computer readable storage medium. Wherein, the computer readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory, etc.

[0112] The motor active vibration and noise reduction method, system, device and storage medium provided by the application are described in detail above, and the principles and implementation manners of the application are described by applying specific examples. The above embodiment description is only used to help understand the method of the application and its core idea; meanwhile, for those skilled in the art, according to the idea of the application, the specific implementation manner and application range will be changed, and the above description should not be understood as a limitation on the application.

Claims

1. A method for active vibration and noise reduction of an electric machine, characterized in that The method comprises: acquiring real-time noise of the motor; inputting the real-time noise into a trained flexible Actor-Critic model to obtain target harmonic injection frequency, target phase and target amplitude of a harmonic injection device; generating a harmonic current for offsetting radial electromagnetic force of the motor based on the target harmonic injection frequency, the target phase and the target amplitude; the flexible Actor-Critic model comprises a soft Q network, a value network and a policy network; before the inputting, the method further comprises: acquiring training samples; training data in the training samples are five-tuple data, which comprise state at a first historical time, action at the first historical time, reward value at the first historical time, cumulative reward value at the first historical time and state at a second historical time; the second historical time is a next time of the first historical time; inputting the training data into the soft Q network, the value network and the policy network respectively to correspondingly obtain predicted reward value, predicted cumulative reward value and predicted action; determining reward loss value of the reward value at the first historical time and the predicted reward value based on a first loss function of the soft Q network, and optimizing network parameters of the soft Q network based on the reward loss value to obtain a target soft Q network; determining predicted reward loss value of the cumulative reward value at the first historical time and the predicted cumulative reward value based on a second loss function of the value network, and optimizing network parameters of the value network based on the predicted reward loss value to obtain a target value network; determining action loss value based on a third loss function of the policy network, the state at the first historical time and the predicted state, and optimizing network parameters of the policy network based on the action loss value to obtain a target policy network; the target soft Q network, the target value network and the target policy network constitute the trained flexible Actor-Critic model; the first loss function is: the second loss function is: the third loss function is: wherein is a first loss function; is a predicted value of the soft Q network; is a true value of the soft Q network; is a second loss function; is a predicted value of the value network; is a true value of the value network; is a third loss function; is a coefficient; is an output value of the policy network.

2. The active vibration and noise reduction method of an electric machine according to claim 1, characterized by, when the number of training iterations of the network parameters of the soft Q network reaches a preset number, or the training error is less than a preset error, or the error change trend is less than a preset change trend, the optimization of the network parameters of the soft Q network is stopped to obtain the target soft Q network.

3. The motor active damping and noise reduction method of claim 1, wherein, The inputting the real-time noise into the trained flexible Actor-Critic model to obtain the target harmonic injection frequency, the target phase and the target amplitude of the harmonic injection device comprises: determining real-time state of the motor based on the real-time noise; the real-time state comprises real-time amplitude, real-time noise frequency, noise phase constant and real-time motor frequency of the motor; processing the real-time state based on the target policy network to obtain a real-time target decision action, and controlling the harmonic injection device to execute the target decision action; wherein the target decision action comprises the target harmonic injection frequency, the target phase and the target amplitude.

4. The motor active damping and noise reduction method of claim 1, wherein, An optimal policy of the policy network is: where is the optimal policy; is the state is the action is the reward when is the discount factor for entropy; is the entropy when the state is the entropy; is the policy function of the policy network; is the expectation over the cumulative reward and the maximum entropy; is the maximum operator.

5. The active vibration and noise reduction method of an electric machine according to claim 1, characterized by, An optimization algorithm for training the soft Q network, the value network and the policy network is stochastic gradient descent.

6. An active vibration and noise reduction system for an electric machine, characterized in that The system is suitable for the motor active vibration and noise reduction method in any one of claims 1-5, and the system comprises: A real-time noise acquisition unit is configured to acquire real-time noise of the motor. A harmonic injection device parameter determination unit is configured to input the real-time noise into the trained flexible Actor-Critic model to obtain a target harmonic injection frequency, a target phase and a target amplitude of the harmonic injection device. A vibration and noise reduction unit is configured to generate a harmonic current for offsetting the radial electromagnetic force of the motor based on the target harmonic injection frequency, the target phase and the target amplitude.

7. A motor active vibration and noise reducing device, characterized by, The memory is configured to store a program. The processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the motor active vibration and noise reduction method in any one of claims 1-5. The computer readable program or instructions are stored in the memory and are executed by the processor to implement the steps of the motor active vibration and noise reduction method in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, ​

Images