Current loop adaptive control method and apparatus

Abstract

The application relates to a current loop adaptive control method and device, wherein the method comprises the following steps: collecting sample data and dynamic performance parameters; fitting an approximate relationship function of current loop adjustment parameters corresponding to the dynamic performance parameters based on the sample data; performing offline training on a parameter adjustment network according to the sample data and the approximate relationship function to obtain a target parameter adjustment network; updating a PI regulator according to the target parameter adjustment network to adjust the current loop and output motor control quantity. The application avoids real-time calculation and network weight updating during operation, reduces the calculation resources in the current loop control process of the permanent magnet synchronous motor, realizes the adaptive control performance of the current loop to the speed change, and is beneficial to improving the stability of the current loop control.

Description

Technical Field

[0001] This application relates to the field of permanent magnet motor technology, and in particular to a current loop adaptive control method and device. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) have advantages such as high efficiency and high power factor, and are widely used in various fields. Currently, the control strategy for PMSMs mainly uses dual closed-loop control, which includes a speed loop and a current loop. The current loop, as the inner loop control, directly acts on the controlled object, and its regulation capability greatly affects the motor control performance. In traditional control methods, the current loop often uses a PI controller with fixed parameters. For systems with a small speed range, speed changes have a small impact on the system, and a single adjustment parameter can meet the system's control requirements. However, for systems with a large speed range, as the speed increases, due to factors such as increased coupling between the D and Q axes and increased digital control delay angle, the current loop adjustment parameters originally suitable for low speeds deteriorate in performance at high speeds, and may even lead to instability.

[0003] To address the aforementioned issues, Artificial Neural Networks (ANNs) have been applied to current loop control, leveraging their self-learning and adaptive capabilities to automatically adjust current loop parameters and stabilize current loop performance. However, current applications of ANNs in dual-loop parameter regulation primarily employ online training, meaning that network weights are calculated and updated in real-time during operation. This requires significant computational resources during the control process. Summary of the Invention

[0004] The purpose of this application is to propose a current loop adaptive control method and apparatus to reduce computational resources in the current loop control process of permanent magnet synchronous motor and improve the stability of current loop control.

[0005] To address the aforementioned technical problems, embodiments of this application provide a current loop adaptive control method, comprising:

[0006] Collect sample data and dynamic performance parameters;

[0007] Based on the sample data, an approximate relationship function for the current loop adjustment parameters corresponding to the dynamic performance parameters is fitted.

[0008] The parameter adjustment network is trained offline based on the sample data and the approximate relation function to obtain the target parameter adjustment network.

[0009] The PI controller is updated according to the target parameter adjustment network to adjust the current loop and output motor control quantity.

[0010] To address the aforementioned technical problems, embodiments of this application provide a current loop adaptive control device, comprising:

[0011] The data acquisition unit is used to collect sample data and dynamic performance parameters.

[0012] The parameter fitting unit is used to fit an approximate relationship function between the current loop adjustment parameters and the dynamic performance parameters based on the sample data.

[0013] The model training unit is used to train the parameter adjustment network offline based on the sample data and the approximate relation function to obtain the target parameter adjustment network.

[0014] The current loop control unit is used to update the PI regulator according to the target parameter adjustment network to adjust the current loop and output motor control quantity.

[0015] This invention provides a current loop adaptive control method and apparatus. The method includes: collecting sample data and dynamic performance parameters; fitting an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters based on the sample data; offline training of a parameter adjustment network based on the sample data and the approximate relationship function to obtain a target parameter adjustment network; and updating a PI controller based on the target parameter adjustment network to adjust the current loop and output motor control quantities. This invention avoids real-time calculation and updating of network weights during operation by fitting an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters based on sample data; and offline training of the parameter adjustment network based on the sample data and the approximate relationship function, thus reducing computational resources in the current loop control process of a permanent magnet synchronous motor. Furthermore, by using a target parameter adjustment network for current loop adjustment, the control performance of the current loop adapts to changes in speed, improving the stability of the current loop control. Attached Figure Description

[0016] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an implementation of the current loop adaptive control method provided in this application embodiment;

[0018] Figure 2This is a schematic diagram of the speed adaptive current loop control circuit topology provided in the embodiments of this application;

[0019] Figure 3 This is a schematic diagram of the parameter adjustment network structure provided in the embodiments of this application;

[0020] Figure 4 This is a flowchart illustrating the implementation of a sub-process in the current loop adaptive control method provided in this application embodiment;

[0021] Figure 5 This is a flowchart illustrating the implementation of a sub-process in the current loop adaptive control method provided in this application embodiment;

[0022] Figure 6 This is a flowchart illustrating the implementation of a sub-process in the current loop adaptive control method provided in this application embodiment;

[0023] Figure 7 This is a flowchart illustrating the implementation of a sub-process in the current loop adaptive control method provided in this application embodiment;

[0024] Figure 8 This is a flowchart illustrating the implementation of a sub-process in the current loop adaptive control method provided in this application embodiment;

[0025] Figure 9 This is a schematic diagram of the current loop adaptive control device provided in the embodiments of this application. Detailed Implementation

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0029] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0030] Please see Figure 1 , Figure 1 This illustrates a specific implementation of the current loop adaptive control method. Figure 2 This is a schematic diagram of the speed adaptive current loop control circuit topology provided in the embodiments of this application; Figure 3 This is a schematic diagram of the parameter adjustment network structure provided in the embodiments of this application.

[0031] It should be noted that if substantially the same result is obtained, the method of this invention is not based on... Figure 1 Limited to the order of the processes shown, this method includes the following steps:

[0032] S1: Collect sample data and dynamic performance parameters.

[0033] This application describes an adaptive current loop control method for permanent magnet synchronous motors. In this embodiment, before offline training of the parameter tuning network, dynamic performance parameters and sample data need to be acquired. When acquiring the dynamic performance parameters, the parameter tuning network needs to be removed from the circuit structure, and a current i is randomly input into the circuit structure. dref and current loop adjustment parameter K dp and K di Dynamic performance parameters are collected, including the maximum peak value M and rise time t during the dynamic process of the current. s When collecting dynamic performance parameters, it is also necessary to collect sample data, which includes the input current i. dref Current loop adjustment parameter K dp and K di and motor speed W e Etc. For example Figure 2 As shown, the parameter adjustment network includes a D-axis parameter adjustment network ANN-d and a Q-axis parameter adjustment network ANN-q. When acquiring dynamic performance parameters, from... Figure 2 Remove the D-axis parameter adjustment network ANN-d and the Q-axis parameter adjustment network ANN-q from the circuit.

[0034] S2: Fit an approximate relationship function between the current loop adjustment parameters and the dynamic performance parameters based on the sample data.

[0035] In this embodiment, the dynamic performance parameters are fitted with respect to the current loop adjustment parameter K based on the collected sample data. dp and K di The approximate relationship function is obtained by fitting the dynamic performance parameters with respect to the current loop adjustment parameter K. dp and Kdi The approximate relationship function can be used to obtain the dynamic performance of the current loop control process, and the dynamic error can be calculated in the subsequent model training process.

[0036] Furthermore, using the least squares method or sampled ridge regression method, an approximate relationship function of the current loop adjustment parameter corresponding to the dynamic performance parameter is fitted based on the sample data;

[0037] The approximate relation function is:

[0038]

[0039] Where M is the maximum peak value, t S For the rise time, f M (·) represents the maximum peak value M and the current loop adjustment parameter K. dp K di relational function; f t (·) represents the rise time t s With current loop adjustment parameter K dp K di Relational function; a n b n and c n f M (·) Weighting coefficients for each power; d n e n and h n f t (·) Weighting coefficients for each power, where m is the highest power of the function, and i dref This is the input current.

[0040] S3: The parameter adjustment network is trained offline based on the sample data and the approximate relation function to obtain the target parameter adjustment network.

[0041] In this embodiment, both the D-axis parameter adjustment network ANN-d and the Q-axis parameter adjustment network ANN-q need to be trained offline simultaneously. The training principles of the two are the same. This embodiment uses the D-axis parameter adjustment network ANN-d as an example for explanation.

[0042] In this embodiment, both the D-axis parameter adjustment network ANN-d and the Q-axis parameter adjustment network ANN-q employ backpropagation (BP) neural networks, and their specific structures are as follows: Figure 3 As shown, the parameter adjustment network consists of three layers: an input layer, a hidden layer, and an output layer. The input layer has four nodes, corresponding to the constant 1, the motor speed W, and so on. eThe network contains the D-axis current error e(t) at time t and the cumulative D-axis current error e(t-1) at time (t-1); the hidden layer has 8 nodes; the output layer has 2 nodes, corresponding to the current loop adjustment parameters Kp and Ki, respectively. In the network, x... i This represents the i-th input quantity of the input layer (i = 1, 2, 3, 4); This represents the k-th output of the output layer (k = 1, 2); w ij This represents the weights between the i-th node in the input layer and the j-th node in the hidden layer (j = 1, 2, ..., 7, 8); w jk This represents the weight between the j-th node in the hidden layer and the k-th node in the output layer.

[0043] Please see Figure 4 , Figure 4 A specific implementation of step S3 is shown below:

[0044] S31: Using the gradient descent method, the sample data is input into the parameter adjustment network in batches.

[0045] In this embodiment, the mini-batch gradient descent method is used during model training. In each iteration, the reference current is randomly input into the current loop control circuit s times, and the weights are updated based on the average of the network weight changes over s iterations. Training is completed after e iterations. The calculation of a single weight change includes forward propagation and backpropagation. The process by which the network calculates the outputs kp and ki based on the input is called forward propagation, while the process by which the network adjusts the weights within the network according to the magnitude of the loss function is called backpropagation.

[0046] S32: Calculate the current loop adjustment parameters for the current iteration based on the input sample data.

[0047] Please see Figure 5 , Figure 5 A specific implementation of step S32 is shown below:

[0048] S321: Adjust the parameters of each node in the hidden layer of the network to multiply the input sample data with the weights between nodes, and accumulate the product results to obtain the accumulated result.

[0049] S322: The accumulated result is processed nonlinearly using the sampled hyperbolic tangent function to obtain the hidden layer output result.

[0050] S323: The accumulated result of the hidden layer output is calculated using a non-negative hyperbolic tangent function to obtain the current loop adjustment parameter for the current iteration.

[0051] This application embodiment describes a forward propagation process, the purpose of which is to calculate the outputs kp and ki based on the input quantities. During the forward propagation process, each node in the hidden layer first accumulates the product of all inputs and weights, and then performs nonlinear processing on the accumulated result using an activation function. Specifically, this application embodiment uses a hyperbolic tangent function as the hidden layer activation function, as follows:

[0052]

[0053] Where x is the input quantity.

[0054] The nonlinear processing process is represented as follows:

[0055]

[0056] in, This indicates that at time t, the j-th node in the hidden layer receives the weighted sum of all nodes in the input layer; x represents the output of the j-th node in the hidden layer at time t. i This represents the i-th input variable of the input layer (i = 1, 2, 3, 4); w ij This represents the weight between the i-th node in the input layer and the j-th node in the hidden layer (j = 1, 2, ..., 7, 8).

[0057] In this embodiment, a non-negative hyperbolic tangent function is used as the output layer excitation function. Specifically, the non-negative hyperbolic tangent function is used to calculate the excitation function of the accumulated output results of the hidden layer, thereby obtaining the current loop adjustment parameters for the current iteration, which are specifically expressed as follows:

[0058]

[0059] Among them, K dp and K di These are the parameters for adjusting the current loop. w represents the first output of the output layer. jk This represents the weight between the j-th node in the hidden layer and the k-th node in the output layer.

[0060] The nonnegative hyperbolic tangent function is specifically expressed as:

[0061]

[0062] S33: Calculate the model loss value based on the current iteration current loop adjustment parameters and the approximate relationship function.

[0063] In this embodiment, the model loss value includes a steady-state error component E1 and a dynamic error component E2. The specific calculation process is as follows:

[0064]

[0065] Where E(t) is the model loss value, f M (k p ,k i Maximum peak value M and current loop adjustment parameter k p k i relational function; f t (k p ,k i ) represents the rise time t s With current loop adjustment parameter k p k i The relationship function is given by m1, m2, and m3, which are the weights of each error component relative to the total error.

[0066] S34: Adjust the weights in the parameter adjustment network according to the loss value to perform offline training on the parameter adjustment network, thereby obtaining the target parameter adjustment network. The target parameter adjustment network includes a D-axis parameter adjustment network and a Q-axis parameter adjustment network.

[0067] Specifically, the purpose of backpropagation is to adjust the weights within the network based on the magnitude of the loss function. Therefore, to ensure the loss function decreases as quickly as possible, the weights are adjusted along the negative gradient of the loss function during training, as follows:

[0068]

[0069] Where γ is the learning rate and α is the momentum factor.

[0070]

[0071] Furthermore, through incremental PID control, it can be expressed as:

[0072]

[0073] Among them, U d (t) represents the control voltage at time t.

[0074] Using formulas (1) to (4) above, we finally obtain the following formula:

[0075]

[0076] Furthermore, following the same steps above, the hidden layer weight adjustment can be derived as follows:

[0077]

[0078] After calculating the weight changes s times, the weighted average is taken to obtain the update amount of each weight in a single iteration:

[0079]

[0080] S4: The PI regulator is updated according to the target parameter adjustment network to adjust the current loop and output motor control quantity.

[0081] In this embodiment, the above steps involve offline training of the parameter tuning network, where the target parameter tuning network, namely the D-axis parameter tuning network and the Q-axis parameter tuning network, has already been trained. In this embodiment, it is the speed adaptive control stage, where the trained D-axis and Q-axis parameter tuning networks are applied to the speed adaptive control stage to perform adaptive control of the current loop.

[0082] Please see Figure 6 , Figure 6 A specific implementation of step S4 is shown below:

[0083] S41: Obtain the reference current and the actual current, and compare the reference current and the actual current to generate a current error.

[0084] Please see Figure 7 , Figure 7 A specific implementation of step S41 is shown below:

[0085] S411: Obtain the sampled three-phase current and perform coordinate transformation on the three-phase current to obtain the D-axis current and the Q-axis current.

[0086] The actual current includes the D-axis current and the Q-axis current, the reference current includes the D-axis reference current and the Q-axis reference current, and the current error includes the D-axis current error and the Q-axis current error.

[0087] S412: Obtain the D-axis reference current and the Q-axis reference current, and use the D-axis reference current and the Q-axis reference current as current loop control quantities.

[0088] S413: Compare the D-axis current with the D-axis reference current to obtain the D-axis current error, and compare the Q-axis current with the Q-axis reference current to obtain the Q-axis current error.

[0089] like Figure 2 As shown, the sampled three-phase current i a i b and i c The D-axis current i is obtained through coordinate transformation. d and Q-axis current i q , respectively with the input D-axis reference current i dref and Q-axis reference current iqref The D-axis current error e is obtained by comparison. d and Q-axis current error e q .

[0090] S42: Obtain the motor speed and input the motor speed and the current error into the target parameter adjustment network for parameter calculation to obtain the PI adjustment parameters.

[0091] Among them, the PI adjustment parameters include the D-axis proportional adjustment parameter K. dp and D-axis integral adjustment parameter K di and Q-axis proportional adjustment parameter K qp Q-axis integral adjustment parameter K qi .

[0092] Specifically, the parameter adjustment networks ANN-d and ANN-q are respectively based on e d e q and rotational speed w e The changes are used to calculate the D-axis PI adjustment parameter (D-axis proportional adjustment parameter K) through forward propagation. dp and D-axis integral adjustment parameter K di ) and Q-axis PI adjustment parameter (Q-axis proportional adjustment parameter K) qp Q-axis integral adjustment parameter K qi ).

[0093] S43: Update the PI regulator using the PI adjustment parameters, and adjust the current loop according to the updated PI regulator to output motor control quantity.

[0094] The updated PI regulator includes a D-axis current regulator and a Q-axis current regulator.

[0095] Please see Figure 8 , Figure 8 A specific implementation of step S43 is shown below:

[0096] S431: Update the PI regulator using the PI adjustment parameters to obtain the D-axis current regulator and the Q-axis current regulator.

[0097] S432: The control voltage is calculated based on the D-axis current error and the Q-axis current error by the D-axis current regulator and the Q-axis current regulator respectively, to obtain the D-axis control voltage and the Q-axis control voltage.

[0098] S433: Convert the D-axis control voltage and the Q-axis control voltage into six-channel switch control quantities, and input the six-channel switch control quantities into the IGBT module to output the motor control quantity.

[0099] The PI controller includes a D-axis current controller PI-d and a Q-axis current controller PI-q.

[0100] In this application example, the D-axis current regulator PI-d and the Q-axis current regulator PI-q are updated through PI adjustment parameters; the D-axis current regulator PI-d and the Q-axis current regulator PI-q respectively receive the input error D-axis current error e. d and Q-axis current error e q The D-axis control voltage Ud and Q-axis control voltage Uq are calculated and then converted into six-channel switch control inputs to the IGBT (Insulated Gate Bipolar Transistor) module to generate motor control signals.

[0101] In this embodiment, sample data and dynamic performance parameters are collected; an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters is fitted based on the sample data; the parameter adjustment network is trained offline according to the sample data and the approximate relationship function to obtain the target parameter adjustment network; the PI controller is updated according to the target parameter adjustment network to adjust the current loop and output the motor control quantity. This embodiment of the invention fits an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters based on sample data; and trains the parameter adjustment network offline according to the sample data and the approximate relationship function, avoiding real-time calculation and updating of network weights during operation, reducing computational resources in the current loop control process of the permanent magnet synchronous motor. Simultaneously, this embodiment of the invention uses neural network calculation to adjust the current loop parameters online, enabling the control performance of the current loop to adapt to changes in speed and ensuring the stability of the current loop control.

[0102] Please refer to Figure 9 As a response to the above Figure 1 The implementation of the method shown in this application provides an embodiment of a current loop adaptive control device, which is similar to... Figure 1 The method embodiments shown correspond to those described.

[0103] like Figure 9 As shown, the current loop adaptive control device 5 in this embodiment includes: a data acquisition unit 51, a parameter fitting unit 52, a model training unit 53, and a current loop control unit 54, wherein:

[0104] Data acquisition unit 51 is used to acquire sample data and dynamic performance parameters;

[0105] The parameter fitting unit 52 is used to fit an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters based on the sample data.

[0106] Model training unit 53 is used to train the parameter adjustment network offline based on the sample data and the approximate relation function to obtain the target parameter adjustment network;

[0107] The current loop control unit 54 is used to update the PI regulator according to the target parameter adjustment network to adjust the current loop and output motor control quantity.

[0108] Furthermore, the current loop control unit 54 includes:

[0109] A current error calculation unit is used to obtain a reference current and a real current, and compare the reference current and the real current to generate a current error;

[0110] The adjustment parameter calculation unit is used to obtain the motor speed and input the motor speed and the current error into the target parameter adjustment network for parameter calculation to obtain the PI adjustment parameters.

[0111] The motor control output unit is used to update the PI regulator through the PI adjustment parameters, and adjust the current loop according to the updated PI regulator to output the motor control quantity.

[0112] Furthermore, the actual current includes the D-axis current and the Q-axis current, the reference current includes the D-axis reference current and the Q-axis reference current, and the current error calculation unit includes:

[0113] The coordinate transformation unit is used to acquire the sampled three-phase current and perform coordinate transformation on the three-phase current to obtain the D-axis current and the Q-axis current.

[0114] A current acquisition unit is used to acquire the D-axis reference current and the Q-axis reference current, and use the D-axis reference current and the Q-axis reference current as current loop control quantities;

[0115] The current error generation unit is used to compare the D-axis current with the D-axis reference current to obtain the D-axis current error, and to compare the Q-axis current with the Q-axis reference current to obtain the Q-axis current error.

[0116] Furthermore, the updated PI regulator includes a D-axis current regulator and a Q-axis current regulator, and the motor control output unit includes:

[0117] The regulator update unit is used to update the PI regulator through the PI regulation parameters to obtain the D-axis current regulator and the Q-axis current regulator;

[0118] The control voltage generation unit is used to calculate the control voltage based on the D-axis current error and the Q-axis current error through the D-axis current regulator and the Q-axis current regulator, respectively, to obtain the D-axis control voltage and the Q-axis control voltage;

[0119] The control quantity generation unit is used to convert the D-axis control voltage and the Q-axis control voltage into six-channel switch control quantities, and input the six-channel switch control quantities into the IGBT module to output the motor control quantity.

[0120] Further, the parameter fitting unit 52 includes:

[0121] An approximate relation function generation unit is used to fit an approximate relation function of the current loop adjustment parameter corresponding to the dynamic performance parameter based on the sample data using the least squares method or the sampled ridge regression method.

[0122] The approximate relation function is:

[0123]

[0124] Where M is the maximum peak value, t S For the rise time, f M (·) represents the maximum peak value M and the current loop adjustment parameter K. dp K di relational function; f t (·) represents the rise time t s With current loop adjustment parameter K dp K di Relational function; a n b n and c n f M (·) Weighting coefficients for each power; d n e n and h n f t (·) Weighting coefficients for each power, where m is the highest power of the function, and i dref This is the input current.

[0125] Furthermore, the model training unit 53 includes:

[0126] The data input unit is used to input the sample data into the parameter adjustment network in batches using the gradient descent method;

[0127] The current loop adjustment parameter calculation unit is used to calculate the current loop adjustment parameters based on the input sample data.

[0128] The loss calculation unit is used to calculate the model loss value based on the current iteration current loop adjustment parameters and the approximate relationship function;

[0129] A target parameter adjustment network generation unit is used to adjust the weights in the parameter adjustment network according to the loss value, so as to perform offline training on the parameter adjustment network to obtain the target parameter adjustment network, wherein the target parameter adjustment network includes a D-axis parameter adjustment network and a Q-axis parameter adjustment network.

[0130] Furthermore, the target parameter adjustment network includes:

[0131] The weighted product unit is used to adjust the weights between the input sample data and the weights between the nodes in the hidden layer of the network by adjusting the parameters, and to accumulate the product results to obtain the accumulated result.

[0132] The nonlinear processing unit is used to sample the hyperbolic tangent function and perform nonlinear processing on the accumulated result to obtain the hidden layer output result.

[0133] The excitation function calculation unit is used to calculate the excitation function of the accumulated result of the hidden layer output using a non-negative hyperbolic tangent function, so as to obtain the current loop adjustment parameter of the current iteration.

[0134] In this embodiment, sample data and dynamic performance parameters are collected; an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters is fitted based on the sample data; the parameter adjustment network is trained offline according to the sample data and the approximate relationship function to obtain the target parameter adjustment network; the PI controller is updated according to the target parameter adjustment network to adjust the current loop and output the motor control quantity. This embodiment of the invention fits an approximate relationship function of the current loop adjustment parameters corresponding to the dynamic performance parameters based on sample data; and trains the parameter adjustment network offline according to the sample data and the approximate relationship function, avoiding real-time calculation and updating of network weights during operation, thus reducing computational resources in the current loop control process of the permanent magnet synchronous motor. Simultaneously, this embodiment of the invention uses neural network calculation to adjust the current loop parameters online, enabling the control performance of the current loop to adapt to changes in speed and ensuring the stability of the current loop control.

[0135] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A current loop adaptive control method, characterized in that, include: Collect sample data and dynamic performance parameters, wherein the sample data includes input current, current loop adjustment parameters and motor speed, and the dynamic performance parameters include the maximum peak value and rise time of the current during the dynamic process; Based on the sample data, an approximate relationship function for the current loop adjustment parameters corresponding to the dynamic performance parameters is fitted. The parameter adjustment network is trained offline based on the sample data and the approximate relation function to obtain the target parameter adjustment network. The PI controller is updated according to the target parameter adjustment network to adjust the current loop and output motor control quantity.

2. The current loop adaptive control method according to claim 1, characterized in that, The step of updating the PI controller according to the target parameter adjustment network to adjust the current loop and output motor control quantity includes: Obtain a reference current and a real current, and compare the reference current and the real current to generate a current error; The motor speed is obtained, and the motor speed and the current error are input into the target parameter adjustment network for parameter calculation to obtain the PI adjustment parameters; The PI regulator is updated by the PI adjustment parameters, and the current loop is adjusted according to the updated PI regulator to output motor control quantity.

3. The current loop adaptive control method according to claim 2, characterized in that, The actual current includes D-axis current and Q-axis current, the reference current includes D-axis reference current and Q-axis reference current, and the current error includes D-axis current error and Q-axis current error. The process of acquiring the reference current and the actual current, comparing the reference current and the actual current, and generating a current error includes: The sampled three-phase currents are acquired, and the three-phase currents are transformed into coordinates to obtain the D-axis current and the Q-axis current. Obtain the D-axis reference current and the Q-axis reference current, and use the D-axis reference current and the Q-axis reference current as current loop control quantities; The D-axis current is compared with the D-axis reference current to obtain the D-axis current error, and the Q-axis current is compared with the Q-axis reference current to obtain the Q-axis current error.

4. The current loop adaptive control method according to claim 3, characterized in that, The updated PI controller includes a D-axis current controller and a Q-axis current controller. The process of updating the PI controller using the PI adjustment parameters and adjusting the current loop according to the updated PI controller to output motor control quantities includes: The PI regulator is updated by the PI adjustment parameters to obtain the D-axis current regulator and the Q-axis current regulator; The control voltage is calculated by the D-axis current regulator and the Q-axis current regulator based on the D-axis current error and the Q-axis current error, respectively, to obtain the D-axis control voltage and the Q-axis control voltage. The D-axis control voltage and the Q-axis control voltage are converted into six-channel switch control quantities, and the six-channel switch control quantities are input into the IGBT module to output the motor control quantity.

5. The current loop adaptive control method according to claim 1, characterized in that, The approximate relationship function for fitting the current loop adjustment parameters corresponding to the dynamic performance parameters based on the sample data includes: Using the least squares method or sampled ridge regression method, an approximate relationship function of the current loop adjustment parameter corresponding to the dynamic performance parameter is fitted based on the sample data; The approximate relation function is: ； in, The maximum peak value, For the rising time, The maximum peak value M and the current loop adjustment parameters , Relational functions; Ascending time With current loop adjustment parameters , Relational functions; and for Weighting coefficients for each power; , and for The weighting coefficients for each power, where m is the highest power of the function. For input current, For power index.

6. The current loop adaptive control method according to any one of claims 1 to 5, characterized in that, The step of offline training of the parameter adjustment network based on the sample data and the approximate relation function to obtain the target parameter adjustment network includes: The sample data is input into the parameter adjustment network in batches using the gradient descent method. The current loop adjustment parameters for the current iteration are calculated based on the input sample data. The model loss value is calculated based on the current iterative current loop adjustment parameters and the approximate relationship function. The weights in the parameter adjustment network are adjusted according to the loss value to train the parameter adjustment network offline, thereby obtaining the target parameter adjustment network, wherein the target parameter adjustment network includes a D-axis parameter adjustment network and a Q-axis parameter adjustment network.

7. The current loop adaptive control method according to claim 6, characterized in that, The step of calculating the current loop adjustment parameters based on the input sample data includes: The parameters are adjusted so that each node in the hidden layer of the network multiplies the input sample data with the weights between nodes, and the product results are accumulated to obtain the accumulated result. The accumulated result is nonlinearly processed using the sampled hyperbolic tangent function to obtain the hidden layer output result; The accumulated results of the hidden layer output are calculated using a non-negative hyperbolic tangent function to obtain the current loop adjustment parameters for the current iteration.

8. A current loop adaptive control device, characterized in that, include: The data acquisition unit is used to acquire sample data and dynamic performance parameters. The sample data includes input current, current loop adjustment parameters, and motor speed. The dynamic performance parameters include the maximum peak value and rise time of the current during the dynamic process. The parameter fitting unit is used to fit an approximate relationship function between the current loop adjustment parameters and the dynamic performance parameters based on the sample data. The model training unit is used to train the parameter adjustment network offline based on the sample data and the approximate relation function to obtain the target parameter adjustment network. The current loop control unit is used to update the PI regulator according to the target parameter adjustment network to adjust the current loop and output motor control quantity.

9. The current loop adaptive control device according to claim 8, characterized in that, The current loop control unit includes: A current error calculation unit is used to obtain a reference current and a real current, and compare the reference current and the real current to generate a current error; The adjustment parameter calculation unit is used to obtain the motor speed and input the motor speed and the current error into the target parameter adjustment network for parameter calculation to obtain the PI adjustment parameters. The motor control output is used to update the PI regulator through the PI adjustment parameters, and to adjust the current loop according to the updated PI regulator, thereby outputting the motor control quantity.

10. The current loop adaptive control device according to claim 9, characterized in that, The actual current includes D-axis current and Q-axis current, the reference current includes D-axis reference current and Q-axis reference current, and the current error calculation unit includes: The coordinate transformation unit is used to acquire the sampled three-phase current and perform coordinate transformation on the three-phase current to obtain the D-axis current and the Q-axis current. A current acquisition unit is used to acquire the D-axis reference current and the Q-axis reference current, and use the D-axis reference current and the Q-axis reference current as current loop control quantities; The current error generation unit is used to compare the D-axis current with the D-axis reference current to obtain the D-axis current error, and to compare the Q-axis current with the Q-axis reference current to obtain the Q-axis current error.

Images