Motor combination agent model construction method, system, device and storage medium

Abstract

The application discloses a motor combined agent model construction method, system, device and storage medium, and the method comprises the steps of determining a space reduction strategy according to motor initial variables and a motor optimization target based on motor constraint conditions; adjusting the motor initial variables according to the space reduction strategy, and reconstructing motor design variables; and constructing a motor combined agent model based on the motor design variables and multiple meta-agent models. Compared with the prior art, with the increase of variable space dimensions and optimization target diversity, the construction difficulty and complexity of the agent model are inevitably improved, the global prediction accuracy of the motor agent model is reduced, and the motor optimization design efficiency and accuracy are affected. However, the application avoids the optimization design dimension difficulty problem caused by the increase of design variables, can effectively improve the global prediction accuracy of the motor agent model, and further improves the intelligent optimization design efficiency and accuracy of the motor combined agent model.

Description

Technical Field

[0001] This invention relates to the field of motor optimization design technology, and in particular to a method, system, device and storage medium for constructing a motor combination proxy model. Background Technology

[0002] Motor optimization design refers to adjusting relevant design variables to obtain a design scheme with optimal motor performance under given design requirements and constraints. Due to the complexity of motor working principles, flexible topologies, and diverse performance indicators, as well as the constraints and conflicts between different indicators, motor design problems are characterized by non-convexity, nonlinearity, and dimensional complexity, which places higher demands on motor modeling and optimization design techniques.

[0003] Improving the efficiency of motor optimization design in complex parameter dimensions has become a research hotspot in the field of motor design. Currently, the mainstream technical approach in the industry is based on accurate motor performance models, employing appropriate multi-objective intelligent optimization algorithms to quickly find the Pareto front of the optimization objective through the powerful computing capabilities of computers, thereby determining the optimal design scheme for motor performance.

[0004] Existing classical surrogate models each possess unique advantages and disadvantages in terms of application scenarios, sample requirements, construction difficulty, global and local fitting accuracy, and predictive capabilities, making the selection of a suitable model challenging in the multi-objective optimization design of motors. While classical models can effectively solve single-objective optimization problems with low variable space dimensionality, the increasing dimensionality of the design variable space and the diversity of optimization objectives inevitably raises the difficulty and complexity of surrogate model construction, leading to increased uncertainty in model prediction performance and ultimately affecting the efficiency and accuracy of motor optimization design. Therefore, how to construct a motor performance surrogate model with high global prediction accuracy to improve the efficiency and accuracy of motor optimization design has become an urgent problem to be solved.

[0005] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0006] The main objective of this invention is to provide a method, system, device, and storage medium for constructing a motor combination proxy model, aiming to solve the technical problem of how to construct a motor performance proxy model with high global prediction accuracy and improve the efficiency and accuracy of motor optimization design.

[0007] To achieve the above objectives, the present invention provides a method for constructing a motor combination proxy model, the method comprising:

[0008] Determine the initial variables of the motor, the constraints of the motor, and the optimization objective of the motor;

[0009] Based on the motor constraints, a space reduction strategy is determined according to the initial variables of the motor and the motor optimization objective.

[0010] The initial variables of the motor are adjusted according to the space reduction strategy, and the motor design variables are reconstructed.

[0011] A motor combination proxy model is constructed based on the motor design variables and multiple meta-proxy models.

[0012] Optionally, the step of determining the space reduction strategy based on the motor constraints, the initial motor variables, and the motor optimization objective includes:

[0013] Based on the motor constraints, determine the parameter correlation analysis results corresponding to the initial variables of the motor and the optimization objective of the motor;

[0014] The space reduction strategy is determined based on the correlation analysis results of the parameters.

[0015] Optionally, the step of determining the parameter correlation analysis results corresponding to the initial variables of the motor and the optimization objective of the motor based on the motor constraints includes:

[0016] The experimental sample space was determined based on the Latin hypercube method;

[0017] Under the motor constraints, the Pearson correlation coefficient is obtained using the Pearson correlation coefficient calculation formula based on the experimental sample space, the initial variables of the motor, and the motor optimization objective.

[0018] The formula for calculating the Pearson correlation coefficient is as follows:

[0019]

[0020] In the formula, P xy Here, x represents the Pearson correlation coefficient, N is the experimental sample space, and x is the Pearson correlation coefficient. i Let y be the initial variable for the i-th motor. i The optimization objective is for the i-th motor;

[0021] The correlation analysis results of the parameters corresponding to the initial variables of the motor are determined based on the Pearson correlation coefficient.

[0022] Optionally, the step of constructing a motor combination proxy model based on the motor design variables and multiple meta-proxy models includes:

[0023] Determine the meta-model weight factors corresponding to multiple meta-proxy models respectively;

[0024] Based on the motor design variables and the meta-model weight factors, a combined proxy model for the motor is constructed by constructing an expression using a combined proxy model.

[0025] The expression for constructing the combined proxy model is:

[0026]

[0027]

[0028] In the formula, x is the optimization variable. ω represents the predicted response value of the combined surrogate model. i Let i be the weight factor of the i-th meta-model. Let be the predicted response value of the i-th meta-model, and m be the initial number of meta-models used to construct the combined surrogate model.

[0029] Optionally, the step of determining the meta-model weight factors corresponding to the multiple meta-proxy models includes:

[0030] The meta-model weight factors corresponding to multiple meta-surrogate models are determined by minimizing the global prediction root mean square error formula.

[0031] The formula for minimizing the root mean square error of global prediction is:

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038] α<1

[0039] β<0

[0040] In the formula, E e For the motor combination surrogate model, E i Let be the root mean square error of the prediction of the i-th meta-surrogate model. Let y(x) be the average root mean square error of the predicted values ​​from the surrogate model for the motor combination, and N be the updated sample space. k ) represents the true response value of the k-th sample point. Let be the predicted response value of the i-th meta-model at the k-th sample point, and α and β be parameters used to control the importance of the average model and the single surrogate model.

[0041] Optionally, after the step of constructing a motor combination proxy model based on the motor design variables and multiple meta-proxy models, the method further includes:

[0042] Based on leave-one-out cross-validation, the normalized root mean square error corresponding to the motor combination proxy model is determined by a preset error formula.

[0043] The preset error formula is:

[0044]

[0045] In the formula, NRSME is the normalized root mean square error;

[0046] Determine whether the normalized root mean square error is less than a preset threshold;

[0047] When the normalized root mean square error is less than the preset threshold, the motor combination proxy model is used as the target combination proxy model.

[0048] Furthermore, to achieve the above objectives, the present invention also proposes a motor combination proxy model construction system, the motor combination proxy model construction system comprising:

[0049] The determination module is used to determine the initial variables of the motor, the motor constraints, and the motor optimization objectives;

[0050] The determining module is further configured to determine a space reduction strategy based on the motor constraints, the initial variables of the motor, and the motor optimization objective.

[0051] The adjustment module is used to adjust the initial variables of the motor according to the space reduction strategy and reconstruct the motor design variables;

[0052] The module is used to construct a motor combination proxy model based on the motor design variables and multiple meta-proxy models, including a polynomial response surface model, a kriging model, and a radial basis function model.

[0053] Optionally, the motor combination proxy model construction system further includes:

[0054] The detection module is used to determine the normalized root mean square error corresponding to the motor combination proxy model based on the leave-one-out cross-validation method and a preset error formula.

[0055] The preset error formula is:

[0056]

[0057] In the formula, NRSME is the normalized root mean square error;

[0058] The detection module is also used to determine whether the normalized root mean square error is less than a preset threshold.

[0059] The detection module is further configured to use the motor combination proxy model as the target combination proxy model when the normalized root mean square error is less than the preset threshold.

[0060] Furthermore, to achieve the above objectives, the present invention also proposes a motor combination proxy model construction device, the device comprising: a memory, a processor, and a motor combination proxy model construction program stored in the memory and executable on the processor, the motor combination proxy model construction program being configured to implement the steps of the motor combination proxy model construction method described above.

[0061] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing a motor combination proxy model construction program, wherein when the motor combination proxy model construction program is executed by a processor, it implements the steps of the motor combination proxy model construction method described above.

[0062] This invention first determines the initial variables, constraints, and optimization objectives of the motor. Then, based on the constraints, it determines a space reduction strategy according to the initial variables and optimization objectives. Following this, it adjusts the initial variables according to the space reduction strategy, reconstructing the motor design variables. Finally, it constructs a combined motor surrogate model based on the reconstructed motor design variables and multiple meta-surrogate models. Compared to existing technologies where the increasing dimensionality of the variable space and the diversity of optimization objectives inevitably increase the difficulty and complexity of surrogate model construction, reduce the global prediction accuracy of the motor surrogate model, and increase the uncertainty of model prediction performance, thus affecting the efficiency and accuracy of motor optimization design, this invention adjusts the initial variables based on the constraints and the space reduction strategy, and constructs a combined motor surrogate model based on the reduced variables and multiple meta-surrogate models. This effectively improves the global prediction accuracy of the motor surrogate model, avoids the difficulty of optimization design dimensions caused by the increase in design variables, and thus improves the intelligent optimization design efficiency and accuracy of the combined motor surrogate model. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the structure of the motor combination proxy model construction device in the hardware operating environment involved in the embodiments of the present invention;

[0064] Figure 2 This is a flowchart illustrating the first embodiment of the motor combination proxy model construction method of the present invention;

[0065] Figure 3 This is a parameterized model diagram of a permanent magnet motor in the first embodiment of the motor combination proxy model construction method of the present invention;

[0066] Figure 4 This is a correlation analysis diagram of design variables and optimization objectives in the first embodiment of the motor combination proxy model construction method of the present invention;

[0067] Figure 5 This is a comparison chart of the accuracy of the motor combination proxy model and the classical meta-proxy model under different optimization objectives in the first embodiment of the motor combination proxy model construction method of the present invention;

[0068] Figure 6 This is a structural block diagram of the first embodiment of the motor combination proxy model construction system of the present invention.

[0069] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0071] Reference Figure 1 , Figure 1 This is a schematic diagram of the device structure for constructing the motor combination proxy model of the hardware operating environment involved in the embodiments of the present invention.

[0072] like Figure 1 As shown, the motor assembly proxy model building device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be high-speed random access memory (RAM) or stable non-volatile memory (NVM), such as a disk storage device. Optionally, the memory 1005 may also be a storage system independent of the aforementioned processor 1001.

[0073] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the device for building the motor combination proxy model, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0074] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a motor combination agent model construction program.

[0075] exist Figure 1 In the motor combination proxy model construction device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the motor combination proxy model construction device of the present invention can be set in the motor combination proxy model construction device, and the motor combination proxy model construction device calls the motor combination proxy model construction program stored in the memory 1005 through the processor 1001 and executes the motor combination proxy model construction method provided in the embodiment of the present invention.

[0076] This invention provides a method for constructing a motor combination proxy model, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the motor combination proxy model construction method of the present invention.

[0077] In this embodiment, the method for constructing the motor combination proxy model includes the following steps:

[0078] Step S10: Determine the initial variables of the motor, the motor constraints, and the motor optimization objective.

[0079] It is easy to understand that the execution subject of this embodiment can be a motor combination proxy model building device with functions such as data processing, network communication and program execution, or other computer devices with similar functions. This embodiment does not limit it.

[0080] It should also be noted that the reference Figure 3 , Figure 3 This is a parameterized model diagram of a permanent magnet motor according to the first embodiment of the motor combination proxy model construction method of the present invention. Figure 3 The design of the motor is based on a spoke-type fractional-slot concentrated winding vernier permanent magnet motor. The initial variables of the motor can be understood as initial design variables, including the stator split ratio k. split stator slot ratio k so stator slot depth ratio k sd stator tooth width ratio k tw Rotor slot ratio k rso Stator tooth tip thickness, stator tooth tip angle ang Permanent magnet width w mag Length h of permanent magnet mag Permanent magnet insertion depth dins Rotor shaping depth h off Stack length l st k Please refer to the table below, which shows the initial design variables and initial adjustment ranges:

[0081] <![CDATA[Stator split ratio k split > 0.2～0.8 <![CDATA[Stator tooth tip angle tip ang > 10°～50° <![CDATA[Stator slot opening ratio k so > 0.2～0.8 <![CDATA[permanent magnet width w mag > 2-5mm <![CDATA[Stator slot depth ratio k sd > 0.2～0.8 <![CDATA[permanent magnet length h mag > 3~8mm <![CDATA[Stator tooth width ratio k tw > 0.2～0.9 <![CDATA[Permanent magnet insertion depth d ins > 0.5~2mm <![CDATA[Rotor slot opening ratio k rso > 0.2～0.8 <![CDATA[Rotor modification depth h off > 0~1mm stator tooth tip thickness 0.5~2mm <![CDATA[Lamination length l stk > 7~10mm

[0082] Motor constraints can be understood as constraints in motor optimization design, and these constraints include: operating voltage U. dc <48V, operating current I r <10A, rated torque T avg >0.8Nm, efficiency η≥88%, power factor PF≥0.9, stator outer diameter R so ≤38mm. Motor optimization objectives include: max{torque density T} den}, min{torque ripple T rip Cogging torque T cog}

[0083] Step S20: Based on the motor constraints, determine the space reduction strategy according to the initial variables of the motor and the motor optimization objective.

[0084] It should be noted that the construction of the motor combination proxy model must be carried out under motor constraints.

[0085] Space reduction strategies include design parameter dimension space reduction strategies, size space reduction strategies, and reducing the dimensions and scale of variable space.

[0086] Furthermore, the processing method for determining the space reduction strategy based on the motor constraints and the initial variables and optimization objectives of the motor is to determine the parameter correlation analysis results corresponding to the initial variables and optimization objectives of the motor based on the motor constraints, and then determine the space reduction strategy based on the parameter correlation analysis results.

[0087] The results of the parameter correlation analysis are the results of the design parameter correlation analysis.

[0088] The processing method for the parameter correlation analysis results corresponding to the initial variables of the motor and the motor optimization objective determined based on motor constraints is as follows: The experimental sample space is determined based on the Latin hypersolution method. Under motor constraints, the Pearson correlation coefficient is obtained using the Pearson correlation coefficient calculation formula based on the experimental sample space, the initial variables of the motor, and the motor optimization objective. The parameter correlation analysis results corresponding to the initial variables of the motor are then determined based on the Pearson correlation coefficient. The formula for calculating the Pearson correlation coefficient is as follows:

[0089]

[0090] In the formula, Pxy Here, x represents the Pearson correlation coefficient, N is the experimental sample space, and x is the Pearson correlation coefficient. i Let y be the initial variable for the i-th motor. i Let i be the optimization objective for the i-th motor.

[0091] In this embodiment, if the traditional univariate scanning method is used, the computation time is enormous, and because the influence of interactions between variables is ignored, a locally optimal rather than globally optimal design may be obtained. The method proposed in this invention first uses the optimal Latin hypercube method to design the experimental sample space, and then uses the Pearson correlation coefficient to perform correlation analysis of the design variables, referencing... Figure 4 , Figure 4 This is a correlation analysis diagram of design variables and optimization objectives in the first embodiment of the motor combination proxy model construction method of the present invention. Figure 4 The results of the correlation analysis of design parameters can be obtained.

[0092] In the specific implementation, based on the correlation analysis results of design parameters, a design parameter dimensional space reduction strategy and a size space reduction strategy are executed to reduce the dimension and scale of the variable space. The design parameter dimensional space reduction strategy divides the design parameter space into two design spaces: one with high correlation and the other with low correlation, based on the correlation analysis results. For variables in the low correlation space, based on the correlation analysis results, multiple variables with strong interactions are reorganized into new design variables, or discarded directly. According to mathematical statistics, when the sample size is greater than 30, if the correlation coefficient between two variables is greater than 0.8, the two variables are considered to have a strong correlation. It has been confirmed that in this embodiment, the variables in the low correlation space have no strong correlation, therefore, there is no need to reorganize them into new variables.

[0093] The size and space reduction strategy combines motor design experience with a database of design parameters for similar motors, and comprehensively evaluates conditions such as electromagnetic load, heat dissipation and cooling, topology, and manufacturing process to reasonably reduce the adjustment range of design variables.

[0094] Step S30: Adjust the initial variables of the motor according to the space reduction strategy and reconstruct the motor design variables.

[0095] It should be understood that the motor design variables are the motor design variables after the parameter space has been reduced (i.e., optimization variables).

[0096] Based on the aforementioned space reduction strategy, this embodiment updates the permanent magnet motor optimization design variable space. This space only contains highly correlated variables and does not include new variables reorganized based on low-correlation variables. Furthermore, the adjustment ranges of variables within the new space are effectively reduced. Refer to the table below, which shows the design variables and adjustment ranges after motor reconstruction:

[0097] <![CDATA[Stator split ratio k split > 0.5～0.8 <![CDATA[permanent magnet width w mag > 3-4mm <![CDATA[Stator slot opening ratio k so > 0.3～0.5 <![CDATA[permanent magnet length h mag > 6~8mm <![CDATA[Stator slot depth ratio k sd > 0.6～0.8 <![CDATA[Rotor modification depth h off > 0~1mm <![CDATA[Stator tooth width ratio k tw > 0.6～0.9 <![CDATA[Lamination length l stk > 8~9mm

[0098] Step S40: Construct a motor combination proxy model based on the motor design variables and multiple meta-proxy models.

[0099] Furthermore, the process of constructing a combined motor proxy model based on motor design variables and multiple meta-proxy models involves determining the meta-model weight factors corresponding to the multiple meta-proxy models, and constructing the combined motor proxy model by constructing expressions based on the motor design variables and meta-model weight factors through the combined proxy model.

[0100] In practical implementation, multiple meta-surrogate models include three classic meta-surrogate models: the polynomial response surface model (PRS), the Kriging-KRG model, and the radial basis function model (RBF).

[0101] This embodiment is based on three classic meta-surrogate models: the polynomial response surface model, the Kriging model, and the radial basis function model. The polynomial response surface model is characterized by its simple construction, low computational cost, and high linear fitting accuracy; therefore, a second-order polynomial response surface model is chosen here. The Kriging model is characterized by its high local fitting accuracy; therefore, a combination of a constant term and a Gaussian kernel correlation function is used here. The radial basis function model is characterized by its high nonlinear fitting accuracy; therefore, a multivariate quadratic basis function with the best predictive performance is chosen here. By assigning different weights to the multiple meta-models, a motor combination surrogate model based on weight allocation is constructed.

[0102] The expression for constructing the composite proxy model is:

[0103]

[0104]

[0105] In the formula, x is the optimization variable. ω represents the predicted response value of the combined surrogate model. i Let i be the weight factor of the i-th meta-model. Let be the predicted response value of the i-th meta-model.

[0106] Furthermore, the method for determining the meta-model weight factors corresponding to multiple meta-surrogate models is to determine the meta-model weight factors corresponding to multiple meta-surrogate models by minimizing the global prediction root mean square error formula, wherein the global prediction root mean square error formula is:

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113] α<1

[0114] β<0

[0115] In the formula, E e For the motor combination surrogate model, E i Let be the root mean square error of the prediction of the i-th meta-surrogate model. Let y(x) be the average root mean square error of the predicted values ​​from the surrogate model for the motor combination, and N be the updated sample space. k ) represents the true response value of the k-th sample point. Let be the predicted response value of the i-th meta-model at the k-th sample point, and α and β be parameters used to control the importance of the average model and the single surrogate model.

[0116] It should also be noted that the weight factor ω of the combined proxy model i The weighting factor is obtained by minimizing the root mean square error of global prediction. Parameters α and β control the importance of the average model and the single surrogate model, respectively, with α < 1 and β < 0. The smaller both α and β are, the larger the weighting factor is assigned to the optimal surrogate model; the larger both α and β are, the higher the average importance of the surrogate model. The free choice of α and β allows for high flexibility in the weighting factor calculation method, and reasonable parameter values ​​will improve the prediction accuracy of the combined surrogate model. In this embodiment, parameters α and β are set to 0.05 and -1, respectively.

[0117] After constructing a motor combination surrogate model based on motor design variables and multiple meta-surrogate models, the normalized root mean square error corresponding to the motor combination surrogate model is determined using a preset error formula based on leave-one-out cross-validation. The preset error formula is:

[0118]

[0119] In the formula, NRSME represents the normalized root mean square error. It is then determined whether the normalized root mean square error is less than a preset threshold. If the normalized root mean square error is less than the preset threshold, the motor combination proxy model is used as the target combination proxy model. The preset threshold can be user-defined, and this embodiment does not impose any restrictions.

[0120] In the specific implementation, leave-one-out cross-validation is used to evaluate the combined surrogate model. The evaluation index for model accuracy is the normalized root mean square error (NRSME), and a smaller NRSME value indicates higher global prediction accuracy. In this embodiment, the NRSME threshold for the model is set to 0.1, and iterations are performed until the model accuracy reaches the convergence requirement. Figure 5 , Figure 5 This is a comparison chart of the accuracy of the motor combination proxy model and the classical meta-proxy model under different optimization objectives in the first embodiment of the motor combination proxy model construction method of the present invention. After adopting the motor combination proxy model construction method proposed in this invention, within a limited sample space, for torque density T... den Torque fluctuation T rip and cogging torque T cog For optimization objectives, the proposed motor combination surrogate model has high global prediction accuracy.

[0121] In this embodiment, the initial variables, constraints, and optimization objectives of the motor are first determined. Then, based on the constraints, a space reduction strategy is determined according to the initial variables and optimization objectives. Following this, the initial variables are adjusted according to the space reduction strategy to reconstruct the motor design variables. Finally, a combined motor proxy model is constructed based on the motor design variables and multiple meta-proxy models. Compared to existing technologies where the increasing dimensionality of the variable space and the diversity of optimization objectives inevitably increase the difficulty and complexity of proxy model construction, leading to increased uncertainty in model prediction performance and affecting the efficiency and accuracy of motor optimization design, this embodiment reconstructs the dimensions and scale of the new design variable space through a space reduction strategy. For different motor performance optimization objectives, a combined motor proxy model with high global prediction accuracy is constructed. Based on the combined motor proxy model, the problem of balancing simulation speed and accuracy in complex dimensional spaces can be effectively solved, significantly improving the efficiency and accuracy of intelligent optimization design based on motor performance proxy models and avoiding the dimensionality disaster of optimization design caused by an increase in design variables.

[0122] Reference Figure 6 , Figure 6 This is a structural block diagram of the first embodiment of the motor combination proxy model construction system of the present invention.

[0123] like Figure 6 As shown, the motor combination proxy model construction system proposed in this embodiment of the invention includes:

[0124] The determination module 6001 is used to determine the initial variables of the motor, the motor constraints, and the motor optimization objectives.

[0125] It should also be noted that the reference Figure 3 , Figure 3This is a parameterized model diagram of a permanent magnet motor according to the first embodiment of the motor combination proxy model construction method of the present invention. Figure 3 The design of the motor is based on a spoke-type fractional-slot concentrated winding vernier permanent magnet motor. The initial variables of the motor can be understood as initial design variables, including the stator split ratio k. split stator slot ratio k so stator slot depth ratio k sd stator tooth width ratio k tw Rotor slot ratio k rso Stator tooth tip thickness, stator tooth tip angle ang Permanent magnet width w mag Length h of permanent magnet mag Permanent magnet insertion depth d ins Rotor shaping depth h off Stack length l st k Please refer to the table below, which shows the initial design variables and initial adjustment ranges:

[0126] <![CDATA[Stator split ratio k split > 0.2～0.8 <![CDATA[Stator tooth tip angle tip ang > 10°～50° <![CDATA[Stator slot opening ratio k so > 0.2～0.8 <![CDATA[permanent magnet width w mag > 2-5mm <![CDATA[Stator slot depth ratio k sd > 0.2～0.8 <![CDATA[Permanent magnet length h mag > 3~8mm <![CDATA[Stator tooth width ratio k tw > 0.2～0.9 <![CDATA[Magnet insertion depth d ins > 0.5~2mm <![CDATA[Rotor slot opening ratio k rso > 0.2～0.8 <![CDATA[Rotor modification depth h off > 0~1mm stator tooth tip thickness 0.5~2mm <![CDATA[Lamination length l stk > 7~10mm

[0127] Motor constraints can be understood as constraints in motor optimization design, and these constraints include: operating voltage U. dc <48V, operating current I r <10A, rated torque T avg >0.8Nm, efficiency η≥88%, power factor PF≥0.9, stator outer diameter R so ≤38mm. Motor optimization objectives include: max{torque density T} den}, min{torque ripple T rip Cogging torque T cog}

[0128] The determining module 6001 is further configured to determine a space reduction strategy based on the motor constraints, the initial variables of the motor, and the motor optimization objective.

[0129] It should be noted that the construction of the motor combination proxy model must be carried out under motor constraints.

[0130] Space reduction strategies include design parameter dimension space reduction strategies, size space reduction strategies, and reducing the dimensions and scale of variable space.

[0131] Furthermore, the processing method for determining the space reduction strategy based on the motor constraints and the initial variables and optimization objectives of the motor is to determine the parameter correlation analysis results corresponding to the initial variables and optimization objectives of the motor based on the motor constraints, and then determine the space reduction strategy based on the parameter correlation analysis results.

[0132] The results of the parameter correlation analysis are the results of the design parameter correlation analysis.

[0133] The processing method for the parameter correlation analysis results corresponding to the initial variables of the motor and the motor optimization objective determined based on motor constraints is as follows: The experimental sample space is determined based on the Latin hypersolution method. Under motor constraints, the Pearson correlation coefficient is obtained using the Pearson correlation coefficient calculation formula based on the experimental sample space, the initial variables of the motor, and the motor optimization objective. The parameter correlation analysis results corresponding to the initial variables of the motor are then determined based on the Pearson correlation coefficient. The formula for calculating the Pearson correlation coefficient is as follows:

[0134]

[0135] In the formula, P xy Here, x represents the Pearson correlation coefficient, N is the experimental sample space, and x is the Pearson correlation coefficient. i Let y be the initial variable for the i-th motor. i Let i be the optimization objective for the i-th motor.

[0136] In this embodiment, if the traditional univariate scanning method is used, the computation time is enormous, and because the influence of interactions between variables is ignored, a locally optimal rather than globally optimal design may be obtained. The method proposed in this invention first uses the optimal Latin hypercube method to design the experimental sample space, and then uses the Pearson correlation coefficient to perform correlation analysis of the design variables, referencing... Figure 4 , Figure 4 This is a correlation analysis diagram of design variables and optimization objectives in the first embodiment of the motor combination proxy model construction method of the present invention. Figure 4 The results of the correlation analysis of design parameters can be obtained.

[0137] In the specific implementation, based on the correlation analysis results of design parameters, a design parameter dimensional space reduction strategy and a size space reduction strategy are executed to reduce the dimension and scale of the variable space. The design parameter dimensional space reduction strategy divides the design parameter space into two design spaces: one with high correlation and the other with low correlation, based on the correlation analysis results. For variables in the low correlation space, based on the correlation analysis results, multiple variables with strong interactions are reorganized into new design variables, or discarded directly. According to mathematical statistics, when the sample size is greater than 30, if the correlation coefficient between two variables is greater than 0.8, the two variables are considered to have a strong correlation. It has been confirmed that in this embodiment, the variables in the low correlation space have no strong correlation, therefore, there is no need to reorganize them into new variables.

[0138] The size and space reduction strategy combines motor design experience with a database of design parameters for similar motors, and comprehensively evaluates conditions such as electromagnetic load, heat dissipation and cooling, topology, and manufacturing process to reasonably reduce the adjustment range of design variables.

[0139] The adjustment module 6002 is used to adjust the initial variables of the motor according to the space reduction strategy and reconstruct the motor design variables.

[0140] It should be understood that the motor design variables are the target design variables after the parameter space has been reduced.

[0141] Based on the aforementioned space reduction strategy, this embodiment updates the permanent magnet motor optimization design variable space. This space only contains highly correlated variables and does not include new variables reorganized based on low-correlation variables. Furthermore, the adjustment ranges of variables within the new space are effectively reduced. Refer to the table below, which shows the target design variables and initial adjustment ranges:

[0142] <![CDATA[Stator split ratio k split > 0.5～0.8 <![CDATA[permanent magnet width w mag > 3-4mm <![CDATA[Stator slot opening ratio k so > 0.3～0.5 <![CDATA[permanent magnet length h mag > 6~8mm <![CDATA[Stator slot depth ratio k sd > 0.6～0.8 <![CDATA[Rotor modification depth h off > 0~1mm <![CDATA[Stator tooth width ratio k tw > 0.6～0.9 <![CDATA[Lamination length l stk > 8~9mm

[0143] Module 6003 is used to construct a motor combination proxy model based on the motor design variables and multiple meta-proxy models.

[0144] Furthermore, the process of constructing a combined motor proxy model based on motor design variables and multiple meta-proxy models involves determining the meta-model weight factors corresponding to the multiple meta-proxy models, and constructing the combined motor proxy model by constructing expressions based on the motor design variables and meta-model weight factors through the combined proxy model.

[0145] In practical implementation, multiple meta-surrogate models include three classic meta-surrogate models: the polynomial response surface model (PRS), the Kriging-KRG model, and the radial basis function model (RBF).

[0146] This embodiment is based on three classic meta-surrogate models: the polynomial response surface model, the Kriging model, and the radial basis function model. The polynomial response surface model is characterized by its simple construction, low computational cost, and high linear fitting accuracy; therefore, a second-order polynomial response surface model is chosen here. The Kriging model is characterized by its high local fitting accuracy; therefore, a combination of a constant term and a Gaussian kernel correlation function is used here. The radial basis function model is characterized by its high nonlinear fitting accuracy; therefore, a multivariate quadratic basis function with the best predictive performance is chosen here. By assigning different weights to the multiple meta-models, a motor combination surrogate model based on weight allocation is constructed.

[0147] The expression for constructing the composite proxy model is:

[0148]

[0149]

[0150] In the formula, x is the optimization variable. ω represents the predicted response value of the combined surrogate model. iLet i be the weight factor of the i-th meta-model. Let be the predicted response value of the i-th meta-model.

[0151] Furthermore, the method for determining the meta-model weight factors corresponding to multiple meta-surrogate models is to determine the meta-model weight factors corresponding to multiple meta-surrogate models by minimizing the global prediction root mean square error formula, wherein the global prediction root mean square error formula is:

[0152]

[0153]

[0154]

[0155]

[0156]

[0157]

[0158] α<1

[0159] β<0

[0160] In the formula, E e For the motor combination surrogate model, E i Let be the root mean square error of the prediction of the i-th meta-surrogate model, E be the average root mean square error of the predictions of the motor combination surrogate model, N be the updated sample space, and y(x) be the root mean square error of the predictions. k ) represents the true response value of the k-th sample point. Let be the predicted response value of the i-th meta-model at the k-th sample point, and α and β be parameters used to control the importance of the average model and the single surrogate model.

[0161] It should also be noted that the weight factor ω of the combined proxy model i The weighting factor is obtained by minimizing the root mean square error of global prediction. Parameters α and β control the importance of the average model and the single surrogate model, respectively, with α < 1 and β < 0. The smaller both α and β are, the larger the weighting factor is assigned to the optimal surrogate model; the larger both α and β are, the higher the average importance of the surrogate model. The free choice of α and β allows for high flexibility in the weighting factor calculation method, and reasonable parameter values ​​will improve the prediction accuracy of the combined surrogate model. In this embodiment, parameters α and β are set to 0.05 and -1, respectively.

[0162] After constructing a motor combination surrogate model based on motor design variables and multiple meta-surrogate models, the normalized root mean square error corresponding to the motor combination surrogate model is determined using a preset error formula based on leave-one-out cross-validation. The preset error formula is:

[0163]

[0164] In the formula, NRSME represents the normalized root mean square error. It is then determined whether the normalized root mean square error is less than a preset threshold. If the normalized root mean square error is less than the preset threshold, the motor combination proxy model is used as the target combination proxy model. The preset threshold can be user-defined, and this embodiment does not impose any restrictions.

[0165] In the specific implementation, leave-one-out cross-validation is used to evaluate the combined surrogate model. The evaluation index for model accuracy is the normalized root mean square error (NRSME), and a smaller NRSME value indicates higher global prediction accuracy. In this embodiment, the NRSME threshold for the model is set to 0.1, and iterations are performed until the model accuracy reaches the convergence requirement. Figure 5 , Figure 5 This is a comparison chart of the accuracy of the motor combination proxy model and the classical meta-proxy model under different optimization objectives in the first embodiment of the motor combination proxy model construction method of the present invention. After adopting the motor combination proxy model construction method proposed in this invention, within a limited sample space, for torque density T... den Torque fluctuation T rip and cogging torque T cog For optimization objectives, the proposed motor combination surrogate model has high global prediction accuracy.

[0166] In this embodiment, the initial variables, constraints, and optimization objectives of the motor are first determined. Then, based on the constraints, a space reduction strategy is determined according to the initial variables and optimization objectives. Following this, the initial variables are adjusted according to the space reduction strategy to reconstruct the motor design variables. Finally, a combined motor proxy model is constructed based on the motor design variables and multiple meta-proxy models. Compared to existing technologies where the increasing dimensionality of the variable space and the diversity of optimization objectives inevitably increase the difficulty and complexity of proxy model construction, leading to increased uncertainty in model prediction performance and affecting the efficiency and accuracy of motor optimization design, this embodiment reconstructs the dimensions and scale of the new design variable space through a space reduction strategy. For different motor performance optimization objectives, a combined motor proxy model with high global prediction accuracy is constructed. Based on the combined motor proxy model, the problem of balancing simulation speed and accuracy in complex dimensional spaces can be effectively solved, significantly improving the efficiency and accuracy of intelligent optimization design based on motor performance proxy models and avoiding the dimensionality disaster of optimization design caused by an increase in design variables.

[0167] Other embodiments or specific implementations of the motor combination proxy model construction system of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0168] 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.

[0169] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0170] 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 the present invention, 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 terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0171] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for constructing a motor combination proxy model, characterized in that, The method for constructing the motor combination proxy model includes the following steps: Determine the initial variables of the motor, the constraints of the motor, and the optimization objective of the motor; Based on the motor constraints, a space reduction strategy is determined according to the initial variables of the motor and the motor optimization objective. The initial variables of the motor are adjusted according to the space reduction strategy, and the motor design variables are reconstructed. A motor combination proxy model is constructed based on the motor design variables and multiple meta-proxy models; The step of constructing a motor combination proxy model based on the motor design variables and multiple meta-proxy models includes: The meta-model weight factors corresponding to multiple meta-surrogate models are determined respectively. The multiple meta-surrogate models include polynomial response surface model, Kriging model and radial basis function model. Based on the motor design variables and the meta-model weight factors, a combined proxy model for the motor is constructed by constructing an expression using a combined proxy model. The expression for constructing the combined proxy model is: In the formula, x To optimize variables, The predicted response value of the combined proxy model, ω i For the first i Weighting factors of individual models For the first i The predicted response value of the individual model. m The initial number of meta-models for constructing the combined proxy model; The step of determining the meta-model weight factors corresponding to multiple meta-proxy models includes: The meta-model weight factors corresponding to multiple meta-surrogate models are determined by minimizing the global prediction root mean square error formula. The formula for minimizing the root mean square error of global prediction is: In the formula, E e For motor combination proxy model, E i For the first i The root mean square error of prediction in the individual surrogate model This represents the average root mean square error of the predicted values ​​from the motor combination surrogate model. N For the updated sample space, For the first k The true response value of each sample point. For the first i The meta-model in the first k The predicted response value for each sample point. α and β These are parameters used to control the importance of the average model and the single surrogate model.

2. The method as described in claim 1, characterized in that, The step of determining the space reduction strategy based on the motor constraints, the initial variables of the motor, and the motor optimization objective includes: Based on the motor constraints, determine the parameter correlation analysis results corresponding to the initial variables of the motor and the optimization objective of the motor; The space reduction strategy is determined based on the correlation analysis results of the parameters.

3. The method as described in claim 2, characterized in that, The step of determining the parameter correlation analysis results corresponding to the initial variables of the motor and the optimization objective of the motor based on the motor constraints includes: The experimental sample space was determined based on the Latin hypercube method; Under the motor constraints, the Pearson correlation coefficient is obtained using the Pearson correlation coefficient calculation formula based on the experimental sample space, the initial variables of the motor, and the motor optimization objective. The formula for calculating the Pearson correlation coefficient is as follows: In the formula, P xy The Pearson correlation coefficient is used. N For the experimental sample space, x i For the first i Initial variables for each motor, y i For the first i Individual motor optimization objectives; The correlation analysis results of the parameters corresponding to the initial variables of the motor are determined based on the Pearson correlation coefficient.

4. The method as described in claim 1, characterized in that, Following the step of constructing a motor combination proxy model based on the motor design variables and multiple meta-proxy models, the method further includes: Based on leave-one-out cross-validation, the normalized root mean square error corresponding to the motor combination proxy model is determined by a preset error formula. The preset error formula is: In the formula, NRSME This is the normalized root mean square error; Determine whether the normalized root mean square error is less than a preset threshold; When the normalized root mean square error is less than the preset threshold, the motor combination proxy model is used as the target combination proxy model.

5. A motor combination proxy model construction system, applied to the motor combination proxy model construction method of claim 1, characterized in that, The motor combination proxy model construction system includes: The determination module is used to determine the initial variables of the motor, the motor constraints, and the motor optimization objectives; The determining module is further configured to determine a space reduction strategy based on the motor constraints, the initial variables of the motor, and the motor optimization objective. The adjustment module is used to adjust the initial variables of the motor according to the space reduction strategy and reconstruct the motor design variables; A construction module is used to construct a motor combination proxy model based on the motor design variables and multiple meta-proxy models.

6. The system as described in claim 5, characterized in that, The motor combination proxy model construction system also includes: The detection module is used to determine the normalized root mean square error corresponding to the motor combination proxy model based on the leave-one-out cross-validation method and a preset error formula. The preset error formula is: In the formula, NRSME This is the normalized root mean square error; The detection module is also used to determine whether the normalized root mean square error is less than a preset threshold. The detection module is further configured to use the motor combination proxy model as the target combination proxy model when the normalized root mean square error is less than the preset threshold.

7. A device for constructing a proxy model for a motor combination, characterized in that, The device includes: a memory, a processor, and a motor combination proxy model building program stored in the memory and executable on the processor, the motor combination proxy model building program being configured to implement the steps of the motor combination proxy model building method as described in any one of claims 1 to 4.

8. A storage medium, characterized in that, The storage medium stores a motor combination proxy model construction program, which, when executed by a processor, implements the steps of the motor combination proxy model construction method as described in any one of claims 1 to 4.

Images