Data-driven predictive current modeling method and apparatus for permanent magnet synchronous motor, medium, and device

Abstract

Embodiments of the present application relate to the technical field of converters. Disclosed are a data-driven predictive current modeling method and apparatus for a permanent magnet synchronous motor, a medium, and a device, aiming to solve the problem in the prior art of reduced prediction accuracy due to the fact that a predictive current model is prone to noise interference. In the present application, firstly, a data-driven current prediction model of a motor is obtained by considering the impact of a cross-saturation effect on an electromagnetic field in an actual situation; secondly, a ridge regression-based calculation model of a parameter vector in the model is constructed on the basis of historical data; and then dynamic adjustment of a ridge parameter is established by assessing the severity of noise in the data, so as to realize adaptive penalty of the data noise, a target solution of a penalty term is determined, and the parameter vector is further inferred to obtain a final prediction model, thereby improving the accuracy of current prediction under noise working conditions.

Description

Data-driven predictive current modeling methods, devices, media, and equipment for permanent magnet synchronous motors Technical Field

[0001] This application relates to the field of converter technology, specifically to a data-driven predictive current modeling method, device, medium, and equipment for permanent magnet synchronous motors. Background Technology

[0002] Model predictive control (MPC) has attracted widespread attention in the field of motor drives due to its advantages such as fast dynamic response and high control flexibility, and is a promising alternative to traditional control methods. Traditional MPC predicts system behavior based on motor models, meaning the controller requires precise motor parameters to achieve satisfactory control performance. However, in real-world operating conditions, key motor parameters are highly sensitive, and data-driven technologies are receiving increasing attention for addressing the parameter dependency problem in MPC.

[0003] Least squares has been proven to be a potential alternative to traditional MPC. The basic principle of least squares is to find the optimal estimate of model parameters by minimizing the sum of squares of the errors between predicted and actual values. Its core idea is to match the model's predictions as closely as possible to the observed data. Therefore, this data-driven approach can identify more accurate current prediction models in real time, thereby improving predictive control performance. However, due to potential noise in current sensor measurements, the dependence of least squares on system input and output data presents a challenge. The predicted current model established by least squares under noise interference is easily affected, resulting in poor prediction accuracy and thus impacting control performance. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, medium, and equipment for predicting current modeling of permanent magnet synchronous motors driven by data, aiming to solve the problem that the prediction current model in the prior art is susceptible to noise interference, which leads to a decrease in prediction accuracy.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows:

[0006] In a first aspect, embodiments of this application provide a data-driven predictive current modeling method for permanent magnet synchronous motors, comprising the following steps:

[0007] Obtain a data-driven current prediction model for the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector;

[0008] Based on historical data collected during historical control cycles, a calculation model based on ridge regression parameter vectors is established; the calculation model includes a penalty term.

[0009] The objective solution for the penalty term is determined based on the noise severity of the historical data.

[0010] The target penalty term and target parameter vector are obtained from the target solution to establish a target data-driven current prediction model.

[0011] In one possible implementation of the first aspect, before determining the target solution for the penalty term based on the noise severity of the historical data, the method further includes:

[0012] Construct matrices from the historical data corresponding to the input and output data respectively;

[0013] Based on the matrix, obtain the residual sum of squares and the total sum of squares;

[0014] Multicollinearity of the input data is obtained by using the sum of squared residuals and the sum of squared totals to characterize the noise severity of historical data.

[0015] In one possible implementation of the first aspect, the objective solution of the penalty term is determined based on the noise severity of the historical data, including:

[0016] Based on the positive correlation between multicollinearity and the noise severity of historical data, the relationship between multicollinearity and the penalty term is obtained.

[0017] The relationship between multicollinearity and penalty terms is mapped using a mapping function, and the target solution of the penalty terms is determined.

[0018] In one possible implementation of the first aspect, multicollinearity of the input data is obtained based on the sum of squared residuals and the total sum of squares to characterize the noise severity of the historical data, including:

[0019] The variance inflation factor of the input data is obtained by using the sum of squared residuals and the sum of squared totals to measure the multicollinearity of the input data, thereby characterizing the noise severity of the historical data.

[0020] In one possible implementation of the first aspect, before obtaining the data-driven current prediction model of the target motor based on the cross-saturation effect, the method further includes:

[0021] Based on the stator voltage and stator current in the current control of the target motor, the input data and output data are obtained respectively.

[0022] Based on the input data, output data, and parameter vector, a data-driven current prediction model for the target motor based on the cross-saturation effect is established.

[0023] In one possible implementation of the first aspect, a calculation model based on ridge regression parameter vectors is established according to historical data collected during historical control cycles, including:

[0024] Based on historical data collected during historical control cycles, historical input data and historical output data are obtained.

[0025] Based on historical input data and historical output data, construct the past subset and the future subset respectively;

[0026] A computational model for parameter vectors based on ridge regression is established based on past and future subsets.

[0027] In one possible implementation of the first aspect, a computational model for the parameter vector based on ridge regression is established according to the past subset and the future subset, including:

[0028] Based on the past and future subsets, the least squares method is applied to estimate the parameter vector, so as to establish a computational model of the parameter vector based on ridge regression.

[0029] Secondly, embodiments of this application provide a data-driven predictive current modeling device for permanent magnet synchronous motors, comprising:

[0030] The acquisition module is used to obtain the data-driven current prediction model of the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector;

[0031] The module is used to build a calculation model based on ridge regression parameter vectors, using historical data collected from historical control cycles; the calculation model includes a penalty term.

[0032] The determination module is used to determine the target solution of the penalty term based on the noise severity of historical data.

[0033] The target module is used to obtain the target penalty term and the target parameter vector based on the target solution, so as to establish a target data-driven current prediction model.

[0034] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when loaded and executed by a processor, implements the permanent magnet synchronous motor data-driven predictive current modeling method as provided in any of the first aspects above.

[0035] Fourthly, embodiments of this application provide an electronic device, including a processor and a memory, wherein,

[0036] Memory is used to store computer programs;

[0037] The processor is used to load and execute computer programs to cause electronic devices to perform a data-driven predictive current modeling method for permanent magnet synchronous motors as provided in any of the first aspects above.

[0038] Compared with the prior art, the beneficial effects of this application are:

[0039] This application proposes a method, apparatus, medium, and device for modeling data-driven predictive current of a permanent magnet synchronous motor. The method includes: obtaining a data-driven current prediction model of the target motor based on cross-saturation effect; wherein the data-driven current prediction model includes a parameter vector; establishing a calculation model of the parameter vector based on ridge regression based on historical data collected from historical control cycles; wherein the calculation model includes a penalty term; determining the target solution of the penalty term based on the noise severity of the historical data; and obtaining the target penalty term and target parameter vector based on the target solution to establish a target data-driven current prediction model. This application first considers the influence of electromagnetic cross-saturation effect under actual conditions to obtain a data-driven current prediction model for the motor. Secondly, it constructs a ridge regression-based calculation model of the parameter vector in the model using historical data. Then, by evaluating the severity of noise in the data, it establishes dynamic adjustment of the ridge parameters to achieve adaptive penalty for data noise, determines the target solution of the penalty term, and further back-calculates the parameter vector to obtain the final prediction model, thereby improving the current prediction accuracy under noisy operating conditions. Attached Figure Description

[0040] Figure 1 is a schematic diagram of least squares overfitting caused by noise;

[0041] Figure 2 is a schematic diagram of the ridge regression fitting;

[0042] Figure 3 is a schematic diagram of the optimal solution space for ridge regression;

[0043] Figure 4 is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of this application;

[0044] Figure 5 is a flowchart illustrating the data-driven predictive current modeling method for permanent magnet synchronous motors provided in an embodiment of this application.

[0045] Figure 6 is a comparative diagram of the waveforms of the predicted current, actual current, and residual current of the predicted current on the dq axis based on the least squares method.

[0046] Figure 7 is a comparative diagram of the dq-axis predicted current, dq-axis actual current, and dq-axis predicted current residual waveforms of the permanent magnet synchronous motor data-driven predicted current modeling method provided in the embodiments of this application.

[0047] Figure 8 is a comparative diagram of the rotational speed, actual dq-axis current, and a-phase current waveforms based on the least squares method.

[0048] Figure 9 is a comparative diagram of the rotational speed, actual dq-axis current, and a-phase current waveforms of the permanent magnet synchronous motor data-driven predictive current modeling method provided in the embodiments of this application.

[0049] Figure 10 is a schematic diagram of the module of the permanent magnet synchronous motor data-driven predictive current modeling device provided in the embodiment of this application;

[0050] The diagram is labeled as follows: 101-Processor, 102-Communication bus, 103-Network interface, 104-User interface, 105-Memory. Detailed Implementation

[0051] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0052] The main solution provided in this application embodiment is: to obtain a data-driven current prediction model of the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector; to establish a calculation model of the parameter vector based on ridge regression according to historical data collected in historical control cycles; wherein, the calculation model includes a penalty term; to determine the target solution of the penalty term according to the noise severity of the historical data; and to obtain the target penalty term and the target parameter vector according to the target solution in order to establish the target data-driven current prediction model.

[0053] Model predictive control (MPC) has attracted widespread attention in the field of motor drives due to its advantages such as fast dynamic response and high control flexibility, and is a promising alternative to traditional control methods (such as field-oriented control and direct torque control). Traditional MPC predicts system behavior based on motor models, meaning the controller requires precise motor parameters to achieve satisfactory control performance. However, in real-world operating conditions, key motor parameters are highly sensitive. For example, the resistance of a permanent magnet synchronous motor (PMSM) may change with temperature, and the inductance value may be affected by magnet saturation and cross-saturation effects. In practical applications, such disturbances make accurate motor modeling and parameter acquisition difficult. Therefore, traditional MPC, which considers fixed motor parameters, suffers from parameter mismatch, leading to performance degradation.

[0054] Against the backdrop of digitalization, networking, and intelligence, data-driven technologies have attracted widespread attention due to their deep integration of next-generation information technology and mathematical models, providing effective solutions for modeling strongly coupled time-varying systems. Among these, least squares techniques are widely used in system identification and intelligent control to predict and explain the relationships between variables. Overall, various solutions have been proposed to improve the robustness of MPC parameters in PMSM drivers. Least squares (LS) has proven to be a potential alternative to traditional MPC. The basic principle of least squares is to find the optimal estimate of model parameters by minimizing the sum of squares of the errors between predicted and actual values. Its core idea is to match the model's predictions as closely as possible to the observed data. Therefore, this data-driven approach can identify more accurate current prediction models in real time, thereby improving predictive control performance.

[0055] Interference immunity is a critical issue for data-driven technologies. The dependence of least squares methods on system input-output data presents a challenge due to the inherent noise in current sensor measurements. Noise sources are diverse, including manufacturing errors, temperature variations, and sensor aging. On the other hand, electromagnetic fields in the motor's operating environment can interfere with signal transmission, resulting in noise in the sampled signal. This interference may originate from the motor's own electromagnetic emissions or from electromagnetic interference caused by other devices. In many industrial applications, motors frequently operate in environments with significant electromagnetic noise, which can easily couple with the sampled signal, leading to persistent measurement errors and affecting model performance. Therefore, to ensure the robustness of the control system, the uncertainties in the observed data in the least squares method must be considered.

[0056] Figure 1 shows a schematic diagram of least squares overfitting caused by noise. In this embodiment, ridge regression technology is applied to the field of permanent magnet synchronous motor drive. Figure 2 shows a schematic diagram of ridge regression fitting, and Figure 3 shows a schematic diagram of the optimal solution space of ridge regression. The severity of noise is evaluated and dynamic ridge coefficient adjustment is applied to achieve adaptive penalty for data noise, thereby improving robustness.

[0057] Therefore, this application provides a method that firstly considers the influence of electromagnetic cross-saturation effect under actual conditions to obtain a data-driven current prediction model for the motor; secondly, constructs a ridge regression-based calculation model of the parameter vector in the model using historical data; then, by evaluating the severity of noise in the data, establishes dynamic adjustment of the ridge parameters to achieve adaptive penalty for data noise, determines the target solution of the penalty term, and further back-infers the parameter vector to obtain the final prediction model, thereby improving the current prediction accuracy under noisy operating conditions.

[0058] Referring to Figure 4, which is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of this application, the electronic device may include: a processor 101, such as a central processing unit (CPU), a communication bus 102, a user interface 104, a network interface 103, and a memory 105. The communication bus 102 is used to realize the connection and communication between these components. The user interface 104 may include a display screen and an input unit such as a keyboard. Optionally, the user interface 104 may also include a standard wired interface and a wireless interface. The network interface 103 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface). The memory 105 may be a storage device independent of the aforementioned processor 101. The memory 105 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as at least one disk storage device. The processor 101 may be a general-purpose processor, including a central processing unit, a network processor, etc., or it may be a digital signal processor, an application-specific integrated circuit, a field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component.

[0059] Those skilled in the art will understand that the structure shown in Figure 4 does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0060] As shown in Figure 4, the memory 105, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a permanent magnet synchronous motor data drive predictive current modeling device.

[0061] In the electronic device shown in Figure 4, the network interface 103 is mainly used for data communication with the network server; the user interface 104 is mainly used for data interaction with the user; the processor 101 and the memory 105 in this application can be set in the electronic device. The electronic device calls the permanent magnet synchronous motor data-driven predictive current modeling device stored in the memory 105 through the processor 101 and executes the permanent magnet synchronous motor data-driven predictive current modeling method provided in the embodiment of this application.

[0062] Referring to Figure 5, based on the hardware device of the foregoing embodiments, this application provides a data-driven predictive current modeling method for permanent magnet synchronous motors, including the following steps:

[0063] S10: Obtain the data-driven current prediction model of the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector.

[0064] In practical implementation, the target motor is a permanent magnet synchronous motor that requires the establishment of a data-driven current prediction model. Cross-saturation refers to the uneven distribution of magnetic flux in the motor's magnetic circuit, causing some magnetic flux lines to reach magnetic saturation simultaneously, while others do not. Data-driven refers to the process of using collected, analyzed, and interpreted data to guide decision-making, optimize processes, and achieve goals.

[0065] In one embodiment, before obtaining a data-driven current prediction model of the target motor based on the cross-saturation effect, the method further includes:

[0066] Based on the stator voltage and stator current in the current control of the target motor, the input data and output data are obtained respectively.

[0067] Based on the input data, output data, and parameter vector, a data-driven current prediction model for the target motor based on the cross-saturation effect is established.

[0068] In practical implementation, in the current control of a permanent magnet synchronous motor, the stator voltage and stator current are the system input and output, respectively. Considering the cross-saturation effect, the state-space representation is as follows:

[0069] y(k+1)=θx(k+1)

[0070]

[0071] Where k is the sampling time, x is the regression matrix, and u = [u d ,u q ] T Let y = [i] be the system input voltage. d i q ] T Let θ be the system output current, θ be the parameter vector, and a, b and e be elements in the parameter vector, corresponding to the coefficients of the current term, voltage term and constant term under the dq axis, respectively.

[0072] S20: Based on historical data collected during historical control cycles, establish a calculation model of parameter vectors based on ridge regression; the calculation model includes a penalty term.

[0073] In practical implementation, parameter vectors are estimated based on historical data collected over multiple control cycles. In the predictive current control environment of the PMSM driver, the applied dq-axis voltage in each control cycle is determined by the controller and thus stored as input data. Conversely, the dq-axis current is sampled at the beginning of each control cycle and stored as output data. That is, historical data is divided into historical input data and historical output data. Based on the historical data collected from previous control cycles, a parameter vector calculation model based on ridge regression is established, including:

[0074] Based on historical data collected during historical control cycles, historical input data and historical output data are obtained.

[0075] Based on historical input data and historical output data, construct the past subset and the future subset respectively;

[0076] A computational model for parameter vectors based on ridge regression is established based on past and future subsets.

[0077] In the specific implementation process, the datasets of historical input data and historical output data are respectively denoted as U. ds and Y ds We can construct a subset, defining two past subsets and a future subset for the input and output data:

[0078] U p =[u1u2...u N-1 ]

[0079] U f =[u2u3...u N ]

[0080] Y p =[y1y2...y N-1 ]

[0081] Y f =[y2y3...y N ]

[0082] The subscripts p and f represent the past and future respectively, U p U is a past subset of the historical input data. f Y is a future subset of the historical input data. p Y is a past subset of the historical output data. f These matrices, representing future subsets of historical output data, have N-1 columns. At sampling time k, based on the collected data, a formula for calculating the ridge regression parameter vector θ, i.e., the computational model, can be established:

[0083] θ=Y f H T (HH T +λI)-1

[0084] in:

[0085]

[0086] I is the identity matrix, H T For the transpose of H, if HH T If it is a singular matrix, then adding a λI term can guarantee HH T +λI is full rank, making the matrix invertible while suppressing the effects of noise.

[0087] In one embodiment, a computational model for parameter vectors based on ridge regression is established according to past and future subsets, including:

[0088] Based on the past and future subsets, the least squares method is applied to estimate the parameter vector, so as to establish a computational model of the parameter vector based on ridge regression.

[0089] In practical implementation, the parameter vector is estimated using the constructed past and future subsets and the least squares solution. Generally, the goal of least squares is to achieve global fitting optimization, which can provide accurate unbiased estimates in the absence of noise. However, data in practical applications is not always ideal. In permanent magnet synchronous motor (PMSM) drives, measurement noise is generated due to current induction, and this noise effect is more prominent under some low-speed, low-load conditions. In this case, data noise can cause multicollinearity. Traditional least squares methods can lead to unreliable results due to overfitting. Therefore, the sensitivity of least squares to noisy data will reduce the current prediction accuracy of PMSM drives. Ridge regression is introduced to solve the above-mentioned multicollinearity problem. By adding a regularization term to the loss function, the problem of high correlation between independent variables is solved, making the model's predictive performance more stable.

[0090] In traditional LS regression, the loss function J is defined as:

[0091] J(θ) = ||Y f -θH|| 2

[0092] In ridge regression models, an L2 norm penalty term is added to the loss function to address the multicollinearity problem.

[0093] J(θ) = ||Y f -θH|| 2 +λ||θ 2

[0094] Where λ∈[0,∞] is the ridge coefficient that adjusts the penalty. The main idea of ​​ridge regression is to add an additional penalty reduction term to the loss function of traditional LS, thereby reducing the optimal solution space. The above equation can be further written as:

[0095] J(θ)=(Y f -θH) T (Y f -θH)+λθ T θ

[0096] =Y f T Y f -Y f T θH-H T θ T Y f +H T θ T θH+λθ T θ

[0097] To minimize J(θ), the relation is applied. Once the local minimum point is found, the following relationship can be derived:

[0098] 0-H T Y f -H T Y+2HH T θ+2λθ=0

[0099] The parameter vector under ridge regression can be represented as:

[0100] θ=Y f H T (HH T +λI) -1

[0101] S30: Determine the target solution for the penalty term based on the noise severity of the historical data.

[0102] In the specific implementation process, a prediction model is established to transform the final solution of the model into the final solution of the parameter vector. Then, adjustments are made based on the noise severity to transform the final solution of the model into the final solution of the penalty term. Specifically: based on the noise severity of historical data, the target solution of the penalty term is determined, including:

[0103] Based on the positive correlation between multicollinearity and the noise severity of historical data, the relationship between multicollinearity and the penalty term is obtained.

[0104] The relationship between multicollinearity and penalty terms is mapped using a mapping function, and the target solution of the penalty terms is determined.

[0105] In practice, a ridge coefficient is introduced to compensate for multicollinearity caused by data noise. Since multicollinearity is positively correlated with the severity of data noise, a larger λ should be used to enhance the penalty intensity due to the higher the multicollinearity. Therefore, the penalty intensity λ is related to the multicollinearity severity VIF. iq The relationship between them can be mapped using a sigmoid function. Therefore, the optimal λ, i.e., the objective solution of the penalty term, can be obtained online based on the evaluation results of the data noise, and the relationship is given as:

[0106]

[0107] Where e is the natural constant, and α and β are fixed coefficients used to stretch the sigmoid function horizontally and vertically. The optimal values ​​of α and β can be searched through offline cross-validation of experimental data, thereby skipping unwanted penalty ranges with relatively low multicollinearity severity and limiting the penalty intensity to an appropriate range.

[0108] In one embodiment, before determining the target solution for the penalty term based on the noise severity of historical data, the method further includes:

[0109] Construct matrices from the historical data corresponding to the input and output data respectively;

[0110] Based on the matrix, obtain the residual sum of squares and the total sum of squares;

[0111] Multicollinearity of the input data is obtained by using the sum of squared residuals and the sum of squared totals to characterize the noise severity of historical data.

[0112] In practical implementation, the data-driven current prediction model involves elements i in both the input voltage and the output current. d i q u d and u q Collect the data corresponding to each variable into a matrix, which can be represented as X, and then calculate i. q The regression coefficient b is:

[0113]

[0114] Among them, X iq For X to contain i q The row of data, X no_iq For those not containing i q The remaining rows of data, and X iq and X no_iq The matrix transpose. Then we obtain i. qThe correlation coefficient with other variables. The sum of squared residuals (SSR) can be expressed as:

[0115]

[0116] in, for The average of the data. The total sum of squares (SST) can be expressed as:

[0117]

[0118] Therefore, i q The variance inflation factor (VIF) can be expressed as:

[0119]

[0120] Among them, R 2 =SSR / SST.

[0121] The VIF value indicates the severity of multicollinearity associated with data noise. Specifically, it uses the sum of squared residuals and the total sum of squares to determine the multicollinearity of the input data, thus characterizing the noise severity of historical data. This includes:

[0122] The variance inflation factor of the input data is obtained by using the sum of squared residuals and the sum of squared totals to measure the multicollinearity of the input data, thereby characterizing the noise severity of the historical data.

[0123] When VIF exceeds a threshold, traditional LS regression becomes unreliable. Ridge regression is introduced to improve prediction accuracy. In this embodiment, the average VIF is used to assess the severity of data noise.

[0124] S40: Obtain the target penalty term and target parameter vector based on the target solution to establish a target data-driven current prediction model.

[0125] In the specific implementation process, according to the aforementioned steps, by determining the target solution of the penalty term, the penalty term can be determined and then the target solution of the parameter vector can be derived, that is, the target parameter vector. Then, the parameter vector in the initially established model is updated to obtain the final target data-driven current model.

[0126] In this embodiment, the electromagnetic field is first affected by the cross saturation effect under actual conditions, and a data-driven current prediction model for the motor is obtained. Then, a ridge regression-based calculation model of the parameter vector in the model is constructed using historical data. Then, by evaluating the severity of noise in the data, dynamic adjustment of the ridge parameter is established to achieve adaptive penalty for data noise. The target solution of the penalty term is determined and the parameter vector is further back-calculated to obtain the final prediction model, thereby improving the current prediction accuracy under noisy conditions.

[0127] To verify the effectiveness of the method provided in this application, ridge regression and least squares method were compared under different noise levels. The comparison of the dq-axis predicted current residuals is shown in Table 1:

[0128] Table 1 Comparison of the variance of predicted residuals at 300 rpm

[0129]

[0130] Specifically, under different noise levels, the prediction residuals of the proposed method are significantly smaller than those of traditional methods (MPC and LS) in both rd(A) and rq(A), demonstrating higher prediction accuracy and robustness. Furthermore, both the LS method and the proposed method, as data-driven approaches, can accurately identify the current prediction model online. They show significant advantages over the model-based MPC method under both no-noise (var=0) and low-noise (var=0.02) conditions. However, as the noise level increases (var=0.05), the negative impact of noise on the prediction model leads to a larger prediction residual for the LS method compared to the MPC method, while the proposed method maintains a significant advantage under the same conditions.

[0131] Referring to Figures 6-9, a comparison is made between the least squares method and the method provided in the embodiments of this application. Figure 6 is a comparative schematic diagram of the dq-axis predicted current, actual dq-axis current, and dq-axis predicted current residual waveforms based on the least squares method. Figure 7 is a comparative schematic diagram of the dq-axis predicted current, actual dq-axis current, and dq-axis predicted current residual waveforms of the data-driven predicted current modeling method for permanent magnet synchronous motors provided in the embodiments of this application. Figure 8 is a comparative schematic diagram of the speed, actual dq-axis current, and a-phase current waveforms based on the least squares method. Figure 9 is a comparative schematic diagram of the speed, actual dq-axis current, and a-phase current waveforms of the data-driven predicted current modeling method for permanent magnet synchronous motors provided in the embodiments of this application. Here, Measured represents the actual measured data, and Predicted represents the predicted data. As can be seen from the comparison of the figures, under noise interference conditions, the method of this application has better predicted current performance than traditional least squares modeling, reflected in reduced current ripple and improved robustness under predicted current closed-loop control.

[0132] Referring to Figure 10, based on the same inventive concept as in the foregoing embodiments, this application also provides a data-driven predictive current modeling device for permanent magnet synchronous motors, comprising:

[0133] The acquisition module is used to obtain the data-driven current prediction model of the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector;

[0134] The module is used to build a calculation model based on ridge regression parameter vectors, using historical data collected from historical control cycles; the calculation model includes a penalty term.

[0135] The determination module is used to determine the target solution of the penalty term based on the noise severity of historical data.

[0136] The target module is used to obtain the target penalty term and the target parameter vector based on the target solution, so as to establish a target data-driven current prediction model.

[0137] Those skilled in the art should understand that the division of the various modules in the embodiments is merely a logical functional division. In actual applications, they can be fully or partially integrated onto one or more actual carriers. These modules can be implemented entirely in software through processing unit calls, entirely in hardware, or a combination of software and hardware. It should be noted that each module in the permanent magnet synchronous motor data-driven predictive current modeling device in this embodiment corresponds one-to-one with each step in the permanent magnet synchronous motor data-driven predictive current modeling method in the aforementioned embodiments. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned permanent magnet synchronous motor data-driven predictive current modeling method, which will not be repeated here.

[0138] Based on the same inventive concept as in the foregoing embodiments, embodiments of this application also provide a computer-readable storage medium storing a computer program. When the computer program is loaded and executed by a processor, it implements the permanent magnet synchronous motor data-driven predictive current modeling method provided in the embodiments of this application.

[0139] Based on the same inventive concept as in the foregoing embodiments, embodiments of this application also provide an electronic device, including a processor and a memory, wherein,

[0140] Memory is used to store computer programs;

[0141] The processor is used to load and execute computer programs to enable electronic devices to perform the permanent magnet synchronous motor data-driven predictive current modeling method provided in the embodiments of this application.

[0142] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a device including one or any combination of the above-mentioned memories. The computer may be a variety of computing devices, including smart terminals and servers.

[0143] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0144] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0145] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0146] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0147] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0148] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a multimedia terminal device (which may be a mobile phone, computer, television receiver, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0149] In summary, this application provides a method, apparatus, medium, and device for modeling data-driven predictive current of a permanent magnet synchronous motor. The method includes: obtaining a data-driven current prediction model of the target motor based on cross-saturation effect; wherein the data-driven current prediction model includes a parameter vector; establishing a calculation model of the parameter vector based on ridge regression based on historical data collected from historical control cycles; wherein the calculation model includes a penalty term; determining the target solution of the penalty term based on the noise severity of the historical data; and obtaining the target penalty term and target parameter vector based on the target solution to establish the target data-driven current prediction model. This application first considers the influence of electromagnetic cross-saturation effect under actual conditions to obtain a data-driven current prediction model for the motor. Secondly, it constructs a ridge regression-based calculation model of the parameter vector in the model using historical data. Then, by evaluating the severity of noise in the data, it establishes dynamic adjustment of the ridge parameters to achieve adaptive penalty for data noise, determines the target solution of the penalty term, and further back-calculates the parameter vector to obtain the final prediction model, thereby improving the current prediction accuracy under noisy operating conditions.

[0150] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A data-driven predictive current modeling method for permanent magnet synchronous motors, characterized in that, Includes the following steps: Obtain a data-driven current prediction model for the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector; Based on historical data collected during historical control cycles, a calculation model for the parameter vector based on ridge regression is established; wherein, the calculation model includes a penalty term; The method further includes determining the target solution of the penalty term based on the noise severity of the historical data; prior to determining the target solution of the penalty term based on the noise severity of the historical data, the method also includes: Construct matrices from the historical data corresponding to the input and output data respectively; Based on the matrix, obtain the residual sum of squares and the total sum of squares; Based on the sum of squared residuals and the sum of squared totals, the multicollinearity of the input data is obtained to characterize the noise severity of the historical data; Determining the target solution of the penalty term based on the noise severity of the historical data includes: Based on the positive correlation between the multicollinearity and the noise severity of the historical data, the relationship between the multicollinearity and the penalty term is obtained; The relationship between the multicollinearity and the penalty term is mapped based on the mapping function, and the target solution of the penalty term is determined; Based on the target solution, the target penalty term and the target parameter vector are obtained to establish a target data-driven current prediction model.

2. The data-driven predictive current modeling method for permanent magnet synchronous motors according to claim 1, characterized in that, The step of obtaining multicollinearity of the input data based on the residual sum of squares and the total sum of squares to characterize the noise severity of the historical data includes: The variance inflation factor of the input data is obtained based on the residual sum of squares and the total sum of squares to measure the multicollinearity of the input data, thereby characterizing the noise severity of the historical data.

3. The data-driven predictive current modeling method for permanent magnet synchronous motors according to claim 1, characterized in that, Before obtaining the data-driven current prediction model of the target motor based on the cross-saturation effect, the method further includes: Based on the stator voltage and stator current in the current control of the target motor, input data and output data are obtained respectively. Based on the input data, the output data, and the parameter vector, a data-driven current prediction model for the target motor based on the cross-saturation effect is established.

4. The data-driven predictive current modeling method for permanent magnet synchronous motors according to claim 1, characterized in that, The calculation model for the parameter vector based on ridge regression, established using historical data collected from historical control cycles, includes: Based on historical data collected during historical control cycles, historical input data and historical output data are obtained. Based on the historical input data and the historical output data, construct the past subset and the future subset respectively; Based on the past subset and the future subset, a computational model for the parameter vector based on ridge regression is established.

5. The data-driven predictive current modeling method for permanent magnet synchronous motors according to claim 4, characterized in that, The step of establishing a computational model for the parameter vector based on ridge regression, according to the past subset and the future subset, includes: Based on the past subset and the future subset, the parameter vector is estimated by applying the solution of the least squares method to establish a computational model of the parameter vector based on ridge regression.

6. A data-driven predictive current modeling device for permanent magnet synchronous motors, characterized in that, include: An acquisition module is used to acquire a data-driven current prediction model of the target motor based on the cross-saturation effect; wherein, the data-driven current prediction model includes a parameter vector; A modeling module is provided, which is used to establish a calculation model of the parameter vector based on ridge regression according to historical data collected in historical control cycles; wherein, the calculation model includes a penalty term; The determining module is configured to determine the target solution of the penalty term based on the noise severity of the historical data; prior to determining the target solution of the penalty term based on the noise severity of the historical data, the module further includes: Construct matrices from the historical data corresponding to the input and output data respectively; Based on the matrix, obtain the residual sum of squares and the total sum of squares; Based on the sum of squared residuals and the sum of squared totals, the multicollinearity of the input data is obtained to characterize the noise severity of the historical data; Determining the target solution of the penalty term based on the noise severity of the historical data includes: Based on the positive correlation between the multicollinearity and the noise severity of the historical data, the relationship between the multicollinearity and the penalty term is obtained; The relationship between the multicollinearity and the penalty term is mapped based on the mapping function, and the target solution of the penalty term is determined; The target module is used to obtain the target penalty term and the target parameter vector based on the target solution, so as to establish a target data-driven current prediction model.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is loaded and executed by the processor, it implements the data-driven predictive current modeling method for permanent magnet synchronous motors as described in any one of claims 1-5.

8. An electronic device, characterized in that, Including processor and memory, among which, The memory is used to store computer programs; The processor is used to load and execute the computer program to cause the electronic device to perform the permanent magnet synchronous motor data-driven predictive current modeling method as described in any one of claims 1-5.

Images