A method, system and product for low-stagnation data-driven predictive control of a permanent magnet motor

Abstract

The present invention relates to the field of motor drive, and in particular to a method, system and product for low-stagnation data-driven predictive control of a permanent magnet motor, comprising the following steps performed in sequence: S1: acquiring sampled data of a motor stator and constructing a data-driven model; S2: judging whether stagnation occurs in the sampled data; S3: optimizing data components of the sampled data; S4: updating the data-driven model using the sampled data when no stagnation occurs; S5: performing Park transformation on the updated current outputted from step S3 or step S4; S6-S7: updating the data-driven model with the current after Park transformation in step S5; S8: super-localizing the predicted current; S9-S10: designing a control equation, and performing space vector modulation to output an empty control pulse; deeply mining and utilizing the invalid components in the sampled data, reducing the occurrence of stagnation, and improving the quality of current control.

Description

Technical Field

[0001] The present invention relates to the field of motor drive, and in particular to a method, system and product for low-stagnation data-driven predictive control of a permanent magnet motor. Background Art

[0002] In the field of data-driven predictive control, a common problem is the "stagnation" phenomenon, that is, the sampled data remains unchanged for multiple consecutive sampling cycles, resulting in a decrease in the frequency of data updates. In this case, because the gradient value is continuously zero, the predictive control method based on the data gradient cannot obtain effective update data, and thus cannot accurately estimate the model coefficients and update the model in time to fit the real-time state of the motor system. This lack of updates directly leads to a sharp decline in control performance, forming a "stagnation effect." Summary of the invention

[0003] The object of the present invention is to provide a method, system and product for low-stagnation data-driven predictive control of a permanent magnet motor that improves the adaptability of a model and the prediction accuracy.

[0004] In order to achieve the above object, the present invention adopts such technical solution:

[0005] A method for low-stagnation data-driven predictive control of a permanent magnet motor comprises the following steps performed in sequence:

[0006] S1: Acquire sampling data of the motor stator, wherein the sampling data at least includes a current sampling signal in the αβ coordinate system and , and construct a data-driven model as shown in the following formula:

[0007] ;

[0008] Among them, the superscript and is the model order, For the The model coefficients, For the The model coefficients, It represents the d-axis component or q-axis component in the sampling current component coordinate system after Park transformation, and the current gradient is the input signal of the data-driven model, the voltage gradient is the output signal of the data-driven model, Indicates that the current Sampling period, It means the The current gradient of a sampling period, Indicates The voltage gradient of a sampling period;

[0009] S2: Determine whether the data sampling process is stagnant. If so, execute step S3; otherwise, execute step S4;

[0010] S3: Optimizing the data components of the sampled data. The specific optimization process is as follows:

[0011] S3-1: Construct a random gain variable frequency biquad generalized integral structure of the following formula (1) to extract the invalid components of the high frequency components of the sampled data:

[0012] (1);

[0013] in, is the resonant gain, and is the extracted signal in the αβ coordinate system, and is the orthogonal signal in the αβ coordinate system, is the fundamental angular frequency, is the Laplace transform variable;

[0014] Fundamental angular frequency As shown in the following formula (2):

[0015] (2);

[0016] in, is the number of motor pole pairs, For time signal The rotor speed under

[0017] S3-2: When stagnation occurs, combined with the preset random gain , the current signal is injected in the opposite phase to replace the original sampled current to update the data-driven model:

[0018] (3);

[0019] S4: When no stagnation occurs, the current sampling signal is used directly and The data-driven model is updated, that is, the data-driven model update current is:

[0020] (4);

[0021] S5: Update the current output from step S3 or step S4 and Perform Park transformation, and the data in the αβ coordinate system drives the model to update the current and Transformed into the current in the dq coordinate system , Represents the d-axis component or q-axis component in the sampling current component coordinate system after Park transformation;

[0022] S6: The current after Park transformation in step S5 The data driven model is updated, and the specific updating process is as follows:

[0023] S6-1: The coefficients of the data-driven model are estimated by using the recursive least square method of the following formula (5) to keep the data-driven model adaptable to the actual system:

[0024] (5);

[0025] Among them, the matrix and They are all recursive matrices, which are updated cyclically in each sampling period, that is, the matrix result of the kth period is called in the k+1th period. is the estimated error, the superscript represents the matrix transpose, is the identity matrix, is the forgetting factor, vector Includes all model coefficients in formula (5); vector Includes all sampled data in formula (5), and the superscript “^” indicates that the variable is an estimated value;

[0026] It is expressed by the following formula (6):

[0027] (6);

[0028] It is expressed by the following formula (7):

[0029] (7);

[0030] S6-2: Prediction using the time shift method shown in the following formula (8) Predicted current at the moment :

[0031] (8);

[0032] S7: Implemented under finite set, according to the predicted current The main control objective of the control is to design the cost function, which is the vector In The elements are replaced with candidate vectors in turn, and control pulses are directly output;

[0033] Implemented under continuous control set, vector Elements in The Lagrangian extrapolation method is implemented using the following formula (9):

[0034] (9);

[0035] S8: Superlocalize formula (8):

[0036] S9: According to the hyperlocalized formula (8), the intelligent proportional-integral-differential-differential-squared control equation shown in the following formula (10) is designed:

[0037] (10);

[0038] in, is the static error of the current signal updated by the model, that is , , , and They are proportional gain, integral gain, differential gain, and differential square gain respectively. represent The first derivative of represent The second-order derivative of is the current reference value, and They are Input gains and lumped variables of the hyperlocalized model at time instant, is the sampling period;

[0039] S10: Convert the voltage signal output by the control equation of step S9 into a pulse signal output.

[0040] Preferably, in step S8, the following formula is used to super-localize formula (8):

[0041] ;

[0042] ;

[0043] in, and They are Input gains and lumped variables of the hyperlocalized model at time instant, is the sampling period.

[0044] A low-stagnation data-driven predictive control system for a permanent magnet motor comprises a memory and a processor, wherein the memory stores executable commands of the processor, and the processor is configured to implement any one of the methods described above by executing the executable commands.

[0045] A computer-readable storage medium stores a computer program, which implements any of the methods described above when executed.

[0046] A computer product comprises a computer program, wherein when the computer program is executed, the method described in any one of the preceding claims is implemented.

[0047] By adopting the above-mentioned design scheme, the beneficial effects of the present invention are:

[0048] 1. Optimize data utilization and enhance model adaptability: A major highlight of this application is to deeply mine and utilize the invalid components in the sampled data. This innovative strategy first realizes the "purification" of data, which is similar to the data cleaning process, effectively eliminating irrelevant information, so that the data-driven model can more accurately reflect the dynamic characteristics and working status of the actual motor system. This not only significantly improves the adaptability and prediction accuracy of the model, indirectly sets an upper limit for the model error, but also improves the current control quality by reducing the occurrence of stagnation;

[0049] 2. In the process of designing the control equation, the applicant designs an intelligent proportional-integral-differential-differential square control equation. This innovative design cleverly combines the advantages of proportional, integral, differential and other control strategies to achieve dual optimization of dynamic and steady-state performance;

[0050] 3. No dependence on physical models, significantly enhanced system robustness: Another advantage of the method proposed in this application is that it completely gets rid of dependence on physical models and specific parameters. This method directly uses system input and output data to establish a data-driven model and implement a low-stagnation solution. It does not require any prior physical model or parameter information, effectively enhancing the robustness of the system. This data-driven model control strategy enables the system to maintain stable control performance and good dynamic response when facing complex and changeable working conditions and changes in the external environment, providing reliable and flexible control solutions for practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of the control method of the present invention;

[0052] Figure 2 It is a theoretical implementation framework diagram of the present invention;

[0053] Figure 3 It is a schematic diagram of the random gain frequency-variable double second-order generalized integrator structure of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] A low-stagnation data-driven predictive control method for a permanent magnet motor, the method flow is as follows Figure 1 As shown in the figure, the theoretical implementation framework is as follows Figure 2 As shown, the steps include the following steps performed in sequence:

[0056] S1: Acquire sampling data of the motor stator, wherein the sampling data at least includes a current sampling signal in the αβ coordinate system and , and construct a data-driven model as shown in the following formula:

[0057] ;

[0058] Among them, the superscript and is the model order, For the The model coefficients, For the The model coefficients, It represents the d-axis component or q-axis component in the sampling current component coordinate system after Park transformation, and the current gradient is the input signal of the data-driven model, the voltage gradient is the output signal of the data-driven model, Indicates that the current Sampling period, It means the The current gradient of a sampling period, Indicates The voltage gradient of a sampling period;

[0059] S2: Determine whether the sampling data process is stagnant. If so, execute step S3, otherwise execute step S4. In this embodiment, if the sampling data of the current sampling cycle is the same as the sampling data of the previous sampling cycle, it is determined that the sampling data is stagnant.

[0060] S3: Optimizing the data components of the sampled data. The specific optimization process is as follows:

[0061] S3-1: The harmonic components generated by harmonic coupling effects and sideband effects are generated by the control strategy. This component in the sampled data is not conducive to establishing a highly adaptive data-driven model to reflect the operating status of the motor system, that is, it is an invalid component, which is usually located in the high-frequency band.

[0062] The random gain frequency-variable double second-order generalized integral (FC-DSOGI) structure of formula (1) is constructed as follows: Figure 3 As shown, it is used to extract the invalid components of the high-frequency components of the sampled data:

[0063] (1);

[0064] in, is the resonant gain, and is the extracted signal in the αβ coordinate system, and is the orthogonal signal in the αβ coordinate system, is the fundamental angular frequency, is the Laplace transform variable;

[0065] Fundamental angular frequency As shown in the following formula (2):

[0066] (2);

[0067] in, is the number of motor pole pairs, For time signal The rotor speed under

[0068] S3-2: When stagnation occurs, combined with the preset random gain In this embodiment, the random gain ; Inject the current signal in the opposite phase to replace the original sampled current to update the data-driven model:

[0069] (3);

[0070] S4: When no stagnation occurs, the current sampling signal is used directly and The data-driven model is updated, that is, the data-driven model update current is:

[0071] (4);

[0072] S5: Update the current output from step S3 or step S4 and Perform Park transformation, and the data in the αβ coordinate system drives the model to update the current and Transformed into the current in the dq coordinate system , Represents the d-axis component or q-axis component in the sampling current component coordinate system after Park transformation;

[0073] S6: The current after Park transformation in step S5 The data driven model is updated, and the specific updating process is as follows:

[0074] S6-1: The coefficients of the data-driven model are estimated using the recursive least squares method (RLS) of the following formula (5) to maintain the adaptation of the data-driven model to the actual system:

[0075] (5);

[0076] Among them, the matrix and They are all recursive matrices, which are updated cyclically in each sampling period, that is, the matrix result of the kth period is called in the k+1th period. is the estimated error, the superscript represents the matrix transpose, is the identity matrix, is the forgetting factor, vector Includes all model coefficients in formula (5); vector Includes all sampled data in formula (5), and the superscript “^” indicates that the variable is an estimated value;

[0077] It is expressed by the following formula (6):

[0078] (6);

[0079] It is expressed by the following formula (7):

[0080] (7);

[0081] S6-2: Prediction using the time shift method shown in the following formula (8) Predicted current at the moment :

[0082] (8);

[0083] S7: Implemented under finite set, according to the predicted current The main control objective of the control is to design the cost function, which is the vector In The elements are replaced with candidate vectors in turn, and control pulses are directly output;

[0084] Implemented under continuous control set, vector Elements in The Lagrangian extrapolation method is implemented using the following formula (9):

[0085] (9);

[0086] S8: Superlocalize formula (8) using the following formula:

[0087] ;

[0088] ;

[0089] in, and They are Input gains and lumped variables of the hyperlocalized model at time instant, is the sampling period.

[0090] S9: According to the hyperlocalized formula (8), design the intelligent proportional-integral-differential-squared (PIDD) as shown in the following formula (10): 2 ) Control equation:

[0091] (10);

[0092] in, is the static error of the current signal updated by the model, that is , , , and They are proportional gain, integral gain, differential gain, and differential square gain respectively. represent The first derivative of represent The second-order derivative of is the current reference value, and They are Input gains and lumped variables of the hyperlocalized model at time instant, is the sampling period;

[0093] S10: Convert the voltage signal output by the control equation of step S9 into a pulse signal output.

[0094] This embodiment also provides a system for implementing the above method.

[0095] A low-stagnation data-driven predictive control system for a permanent magnet motor comprises a memory and a processor, wherein the memory stores executable commands of the processor, and the processor is configured to implement any one of the methods described above by executing the executable commands.

[0096] This embodiment also provides a storage medium storing a computer program capable of implementing the above method.

[0097] A computer-readable storage medium stores a computer program, which implements any of the methods described above when the computer program is executed.

[0098] In this embodiment, a computer product for implementing the above method is also provided.

[0099] A computer product comprises a computer program, and when the computer program is executed, the method described in any one of the above is implemented.

[0100] In summary, a major highlight of this application is the in-depth mining and utilization of invalid components in the sampled data. This innovative strategy first realizes the "purification" of the data, which is similar to the data cleaning process, effectively eliminating irrelevant information, so that the data-driven model can more accurately reflect the dynamic characteristics and working status of the actual motor system. This not only significantly improves the adaptability and prediction accuracy of the model, indirectly sets an upper limit for the model error, but also improves the current control quality by reducing the occurrence of stagnation.

[0101] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A low-stagnation data-driven predictive control method for a permanent magnet motor, characterized in that: The process includes the following steps: S1: Acquire sampling data of the motor stator, wherein the sampling data at least includes a current sampling signal in the αβ coordinate system and , and construct a data-driven model as shown in the following formula: ； Among them, the superscript and is the model order, For the The model coefficients, For the The model coefficients, It represents the d-axis component or q-axis component in the sampling current component coordinate system after Park transformation, and the current gradient is the input signal of the data-driven model, the voltage gradient is the output signal of the data-driven model, Indicates that the current Sampling period, It means the The current gradient of a sampling period, Indicates The voltage gradient of a sampling period; S2: Determine whether the data sampling process is stagnant. If so, execute step S3; otherwise, execute step S4; S3: Optimizing the data components of the sampled data. The specific optimization process is as follows: S3-1: Construct a random gain variable frequency biquad generalized integral structure of the following formula (1) to extract the invalid components of the high frequency components of the sampled data: （1）； in, is the resonant gain, and is the extracted signal in the αβ coordinate system, and is the orthogonal signal in the αβ coordinate system, is the fundamental angular frequency, is the Laplace transform variable; Fundamental angular frequency As shown in the following formula (2): (2)； in, is the number of motor pole pairs, For time signal The rotor speed under S3-2: When stagnation occurs, combined with the preset random gain , the current signal is injected in the opposite phase to replace the original sampled current to update the data-driven model: （3）； S4: When no stagnation occurs, the current sampling signal is used directly and The data-driven model is updated, that is, the data-driven model update current is: （4）； S5: Update the current output from step S3 or step S4 and Perform Park transformation, and the data in the αβ coordinate system drives the model to update the current and Transformed into the current in the dq coordinate system , Represents the d-axis component or q-axis component in the sampling current component coordinate system after Park transformation; S6: The current after Park transformation in step S5 The data driven model is updated, and the specific updating process is as follows: S6-1: The coefficients of the data-driven model are estimated by using the recursive least square method of the following formula (5) to keep the data-driven model adaptable to the actual system: （5）； Among them, the matrix and They are all recursive matrices, which are updated cyclically in each sampling period, that is, the matrix result of the kth period is called in the k+1th period. is the estimated error, the superscript represents the matrix transpose, is the identity matrix, is the forgetting factor, vector Includes all model coefficients in formula (5); vector Includes all sample data in formula (5), and the superscript "^" indicates that the variable is an estimated value; It is expressed by the following formula (6): （6）； It is expressed by the following formula (7): （7）； S6-2: Prediction using the time shift method shown in the following formula (8) Predicted current at the moment : （8）； S7: Implemented under finite set, according to the predicted current The main control objective of the control is to design the cost function, which is the vector In The elements are replaced with candidate vectors in turn, and control pulses are directly output; Implemented under continuous control set, vector Elements in The Lagrangian extrapolation method is implemented using the following formula (9): （9）； S8: Superlocalize formula (8): S9: According to the hyperlocalized formula (8), the intelligent proportional-integral-differential-differential-squared control equation shown in the following formula (10) is designed: （10）； in, is the static error of the current signal updated by the model, that is , , , and They are proportional gain, integral gain, differential gain, and differential square gain respectively. represent The first derivative of represent The second-order derivative of is the current reference value, and They are Input gains and lumped variables of the hyperlocalized model at time instant, is the sampling period; S10: Convert the voltage signal output by the control equation of step S9 into a pulse signal output.

2. The method for low-stagnation data-driven predictive control of a permanent magnet motor according to claim 1, characterized in that: In step S8, the following formula is used to superlocalize formula (8): ； ； in, and They are Input gains and lumped variables of the hyperlocalized model at time instant, is the sampling period.

3. A low-stagnation data-driven predictive control system for a permanent magnet motor, comprising a memory and a processor, characterized in that: The memory stores executable instructions of the processor, and the processor is configured to implement the method according to any one of claims 1 to 2 by executing the executable instructions.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method according to any one of claims 1 to 2 is implemented.

5. A computer product, comprising a computer program, characterized in that: When the computer program is executed, the method according to any one of claims 1 to 2 is implemented.

Images