Vehicle motor electromagnetic performance optimization method and system, electronic equipment and storage medium

By using machine learning and the Grey Wolf optimization algorithm to establish a surrogate model in the drive motor of new energy vehicles, the motor design parameters are optimized, the modeling accuracy problem in electromagnetic performance optimization is solved, and the motor performance is improved.

CN119026464AInactive Publication Date: 2024-11-26CHONGQING JINKANG POWER NEW ENERGY CO LTD

Patent Information

Application Number
CN202411063873.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the optimization of electromagnetic performance of drive motors for new energy vehicles, the modeling accuracy of machine learning algorithms still has room for improvement, especially in terms of motor base speed, maximum output power and efficiency.

Method used

Simulation data was obtained through experimental design. A surrogate model between structural parameters and performance objectives was established using machine learning algorithms. The GWO optimization algorithm and multi-objective optimization algorithm were combined to optimize the GWO-SVR surrogate model between design variables and optimization objectives. The solution was obtained with the goal of maximizing motor base speed, maximum output power and efficiency.

Benefits of technology

It increases the motor's base speed and maximum output power, improves the motor's external characteristics and efficiency, and enhances the precision and efficiency of motor development.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026464A_ABST
    Figure CN119026464A_ABST
Patent Text Reader

Abstract

The invention discloses a vehicle motor electromagnetic performance optimization method and system, electronic equipment and a storage medium, and the method comprises the steps: obtaining simulation experiment data through experiment design, enabling the simulation experiment data to comprise motor data, training the simulation experiment data through a machine learning algorithm, and carrying out the training of the motor data, establishing an agent model between the structure parameters and the performance targets; a grey wolf optimization algorithm GWO is utilized to optimize the proxy model of the structure parameters and the performance targets, and a GWO-SVR proxy model between the design variables and the optimization targets is established; and solving the GWO-SVR proxy model between the design variables and optimization targets by using a multi-target optimization algorithm by taking the structure parameters as decision vectors and maximizing the motor base speed, the motor maximum output power and the motor efficiency as optimization targets to obtain an optimal solution. The base speed of the motor and the maximum output power of the motor can be improved, so that the external characteristics and efficiency of the motor are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle technology, and more specifically, to methods, systems, electronic devices, and storage media for optimizing the electromagnetic performance of vehicle motors. Background Art

[0002] Electricity, as an environmentally friendly secondary energy source, can replace fossil fuels in traditional internal combustion engines and drive electric motors to output mechanical energy. Vigorously developing new energy vehicles is an important way to solve energy and environmental problems and a key issue in national development planning.

[0003] Intense competition in the drive motor market for new energy vehicles is placing higher demands on the powertrain motor's power density, efficiency, and lightweight design. In the development of actual motor electromagnetic performance, explicit representation of input and output is difficult for practical engineering problems with multiple input variables, nonlinearity, and challenging mechanistic model derivation. With the advancement of computer computing performance, data-driven applications are gaining momentum across various industries, leveraging machine learning algorithms to establish proxy models for input and output. While machine learning algorithms offer good modeling accuracy for research objectives such as motor efficiency η, power density, and external motor characteristics, they still require significant improvement in certain areas. Summary of the Invention

[0004] In view of this, the present invention provides a method, system, electronic device and storage medium for optimizing the electromagnetic performance of a vehicle motor, in order to improve the accuracy of existing modeling.

[0005] In a first aspect, this application provides a method for optimizing the electromagnetic performance of a vehicle motor, comprising:

[0006] Simulation data is obtained through experimental design, including motor data. Machine learning algorithms are used to train the simulation data to establish a proxy model between structural parameters and performance targets.

[0007] The Grey Wolf Optimization Algorithm (GWO) is used to optimize the surrogate model of structural parameters and performance objectives, and a GWO-SVR surrogate model is established between design variables and optimization objectives.

[0008] Using structural parameters as decision vectors and maximizing motor base speed, maximum motor output power, and motor efficiency as optimization objectives, a multi-objective optimization algorithm is used to solve the GWO-SVR surrogate model between design variables and optimization objectives to obtain the optimal solution.

[0009] Secondly, this application provides a vehicle motor electromagnetic performance optimization system, comprising:

[0010] The module for establishing a proxy model between structural parameters and performance targets is used to obtain simulation experimental data through experimental design. The simulation experimental data includes motor data. Machine learning algorithms are used to train the simulation experimental data to establish a proxy model between structural parameters and performance targets.

[0011] The GWO-SVR surrogate model building module is used to optimize the surrogate model of structural parameters and performance objectives using the Grey Wolf Optimization Algorithm (GWO), and to establish a GWO-SVR surrogate model between design variables and optimization objectives.

[0012] The calculation module is used to solve the GWO-SVR surrogate model between the design variables and the optimization objectives by using structural parameters as decision vectors and maximizing the motor base speed, maximum motor output power, and motor efficiency. The optimal solution is obtained by using a multi-objective optimization algorithm to solve the GWO-SVR surrogate model between the design variables and the optimization objectives.

[0013] Thirdly, this application provides an electronic device, which includes:

[0014] Memory, used to store programs; and

[0015] A processor is used to execute the vehicle motor electromagnetic performance optimization method of any one of the present application by calling a program stored in memory.

[0016] Fourthly, this application provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, which are used to cause a computer to execute the vehicle motor electromagnetic performance optimization method according to any one of the claims of this application.

[0017] Compared with the prior art, the vehicle motor electromagnetic performance optimization method, system, electronic device, and storage medium provided by the present invention achieve at least the following beneficial effects:

[0018] This invention provides a method, system, electronic device, and storage medium for optimizing the electromagnetic performance of a vehicle motor. The method includes: acquiring simulation experimental data (including motor data) through experimental design; training the simulation experimental data using a machine learning algorithm to establish a surrogate model between structural parameters and performance targets; optimizing the surrogate model using the Grey Wolf Optimization (GWO) algorithm to establish a GWO-SVR surrogate model between design variables and optimization targets; using structural parameters as decision vectors and maximizing the motor base speed, maximum motor output power, and motor efficiency as optimization targets, solving the GWO-SVR surrogate model between design variables and optimization targets using a multi-objective optimization algorithm to obtain the optimal solution. This approach improves the motor base speed and maximum motor output power, thereby enhancing the motor's external characteristics and efficiency. Based on the GWO-SVR surrogate model between design variables and optimization targets, the method determines the motor design parameters, implements optimal constraints on performance targets and material usage, and quickly provides the optimal design scheme, improving the development efficiency and accuracy of the motor.

[0019] Of course, any product implementing the present invention does not necessarily need to achieve all of the above technical effects at the same time.

[0020] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0022] Figure 1 This is a flowchart illustrating a method for optimizing the electromagnetic performance of a vehicle motor provided by the present invention.

[0023] Figure 2 is a sample distribution of motor design variables provided by the present invention; Figure 2.1 This is a schematic diagram of the structure of an induction motor stator provided by the present invention; Figure 2.2 This is a schematic structural diagram of an induction motor rotor structure provided by the present invention;

[0024] Figure 3(a) shows the base speed n of the motor provided by the present invention. b The R-Square values ​​of each proxy model are used as evaluation metrics; Figure 3(b) shows the maximum output power P of a motor provided by this invention. max The magnitude of the R-Square value of each proxy model; Figure 3(c) shows the magnitude of the R-Square value of each proxy model for motor efficiency η provided by the present invention;

[0025] Figure 4(a) is a curve diagram of a motor efficiency η and a test sample point provided by the present invention; Figure 4(b) is a training result diagram of an optimized target motor efficiency η and a test sample point using GP provided by the present invention; Figure 4(c) is a training result diagram of an optimized target motor efficiency η and a test sample point using BP provided by the present invention; Figure 4(d) is a training result diagram of an optimized target motor efficiency η and a test sample point using RF provided by the present invention; Figure 4(e) is a training result diagram of an optimized target motor efficiency η and a test sample point using XGBoost provided by the present invention; Figure 4(f) is a training result diagram of an optimized target motor efficiency η and a test sample point using RBF provided by the present invention; Figure 4(g) is a training result diagram of an optimized target motor efficiency η and a test sample point using Kriging provided by the present invention; Figure 4(h) is a training result diagram of an optimized target motor efficiency η and a test sample point using RSM provided by the present invention; Figure 4(j) is a training result diagram of an optimized target motor efficiency η and a test sample point using SVR provided by the present invention;

[0026] Figure 5 This is a flowchart illustrating a GWO-SVR model provided by the present invention;

[0027] Figure 6 This is a schematic diagram of the NSGA-II algorithm provided by the present invention;

[0028] Figure 7 This invention provides a Pareto solution set for different optimization algorithms (CMOEA-MS, ToP, and NSGA-II).

[0029] Figure 8 This invention provides a graph showing the external characteristic curves of a motor before and after optimization.

[0030] Figure 9(a) is a schematic diagram of the stator slot structure before and after optimization provided by the present invention; Figure 9(b) is a schematic diagram of the rotor slot structure before and after optimization provided by the present invention.

[0031] Figure 10 The present invention provides a motor base speed n b Test experiment diagram;

[0032] Figure 11 This invention provides a maximum output power P of a motor. max Test experiment diagram;

[0033] Figure 12 This is a schematic diagram of the structure of a vehicle motor electromagnetic performance optimization system provided by the present invention. Detailed Implementation

[0034] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention.

[0035] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0036] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0037] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0038] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0039] See also Figures 1-11 As shown, Figure 1 This is a flowchart illustrating a method for optimizing the electromagnetic performance of a vehicle motor provided by the present invention. Figure 2 is a sample distribution of motor design variables provided by the present invention; Figure 2.1 This is a schematic diagram of the structure of an induction motor stator provided by the present invention; Figure 2.2 FIG3 (a) is a schematic diagram of a motor base speed n provided by the present invention. b The R-Square values ​​of each proxy model are used as evaluation metrics; Figure 3(b) shows the maximum output power P of a motor provided by this invention. maxFigure 3(c) shows the R-Square values ​​of the evaluation metrics for each proxy model for motor efficiency η provided by this invention; Figure 4(a) shows the curve of motor efficiency η versus test sample points provided by this invention; Figure 4(b) shows the training results of optimizing target motor efficiency η versus test sample points using GP provided by this invention; Figure 4(c) shows the training results of optimizing target motor efficiency η versus test sample points using BP provided by this invention; Figure 4(d) shows the training results of optimizing target motor efficiency η versus test sample points using RF provided by this invention. Figure 4(e) is a training result diagram of the optimized target motor efficiency η and test sample points using XGBoost provided by the present invention; Figure 4(f) is a training result diagram of the optimized target motor efficiency η and test sample points using RBF provided by the present invention; Figure 4(g) is a training result diagram of the optimized target motor efficiency η and test sample points using Kriging provided by the present invention; Figure 4(h) is a training result diagram of the optimized target motor efficiency η and test sample points using RSM provided by the present invention; Figure 4(j) is a training result diagram of the optimized target motor efficiency η and test sample points using SVR provided by the present invention. Figure 5 This is a flowchart illustrating a GWO-SVR model provided by the present invention; Figure 6 This is a schematic diagram of the NSGA-II algorithm provided by the present invention; Figure 7 This invention provides a Pareto solution set for different optimization algorithms (CMOEA-MS, ToP, and NSGA-II). Figure 8 Figure 9(a) is a schematic diagram of the stator slot structure before and after optimization provided by the present invention; Figure 9(b) is a schematic diagram of the rotor slot structure before and after optimization provided by the present invention. Figure 10 The present invention provides a motor base speed n b Test experiment diagram; Figure 11 This invention provides a maximum output power P of a motor. max Test experiment diagram; This embodiment provides a method for optimizing the electromagnetic performance of a vehicle motor, including:

[0040] S1. Obtain simulation experimental data through experimental design. The simulation experimental data includes motor data. Use machine learning algorithms to train the simulation experimental data and establish a proxy model between structural parameters and performance targets.

[0041] Specifically, simulation experimental data is obtained through experimental design, wherein the simulation experimental data includes motor data, and the motor data includes design parameters and performance parameters, wherein the design parameters include but are not limited to motor stator structure size, motor rotor structure size, magnetic steel data and winding data; the performance parameters include but are not limited to motor torque, motor efficiency, motor input current, motor maximum output power and motor base speed.

[0042] Specifically, the Latin Hypercube (LHS) method is used for experimental design. The Latin Hypercube experimental design has the advantages of a small number of sampling times and effectively avoids repeated sampling. It is suitable for optimization designs with many design variables and a large sampling space. The Latin Hypercube LHS method is used to generate 100 candidate design samples. The sample distribution is as follows: Figure 2 As shown, the sample points are evenly distributed within the design range. The corresponding target values ​​were obtained using the parametric simulation module of ANSYS software. Some sample point data and target values ​​are shown in Table 1. When selecting the design space of the design variables, the allowable range of actual processing was taken into consideration.

[0043] Table 1 Some sample point data and target values

[0044]

[0045] Combined with Table 1.1 and Figure 2.1 As shown, the above motor stator structure dimensions can be the induction motor stator structure dimensions, such as the induction motor stator slot dimensions, combined with Table 1.2 and Figure 2.2 As shown, the above-mentioned motor rotor structural dimensions may be the induction motor rotor structural dimensions, such as the induction motor rotor slot dimensions.

[0046] Table 1.1 Induction motor stator slot dimensions

[0047]

[0048]

[0049] Table 1.2 Induction motor rotor slot dimensions

[0050]

[0051] The simulation experimental data are trained using machine learning algorithms to establish a proxy model between structural parameters and performance targets. The machine learning algorithms include genetic programming, error back propagation method, random forest algorithm, distributed gradient boosting library algorithm, radial basis function algorithm, kriging interpolation, response surface analysis method or support vector regression algorithm.

[0052] Specifically, for some practical engineering problems with numerous input design variables, nonlinearity, and difficult mechanistic model derivation, it is difficult to obtain explicit expressions of inputs and outputs. With the improvement of computer computing performance, data-driven approaches are gradually being applied in various industries. The main idea is to obtain a black-box model of inputs and outputs based on a large number of input variables x and outputs y through algorithms such as machine learning and deep learning. That is, to establish a surrogate model between inputs and outputs, such as a surrogate model between structural parameters and performance targets.

[0053] It's important to note that machine learning is a type of artificial intelligence that, given a dataset, can automatically learn patterns and use the learned results to predict or make decisions. It relies on mathematical and statistical algorithms to build models that allow computers to learn autonomously without explicit programming. Deep learning, on the other hand, is a method within machine learning that utilizes multi-layered neural networks to simulate and solve complex problems. Its core idea is to progressively improve the model's ability to represent and predict data through hierarchical learning and feature extraction. Key technologies in deep learning include backpropagation, convolutional neural networks, recurrent neural networks, and survival-versus-adversarial networks. Deep learning has a wide range of applications, such as image recognition, speech recognition, natural language processing, and artificial intelligence.

[0054] Genetic programming GP, error back propagation method BP, random forest algorithm RF, distributed gradient boosting library algorithm XGBoost, radial basis function algorithm RBF, Kriging interpolation Kriging, response surface analysis method RSM, and support vector regression algorithm SVR are used to establish a proxy model between structural parameters and performance objectives.

[0055] 1. Genetic Programming (GP)

[0056] GP is a supervised machine learning method and a modeling method based on Bayesian statistical principles. Bayesian theory regards unknown deterministic functions as the realization of random processes. Before obtaining data, the uncertainty of the variable is described by the prior distribution. After obtaining data, the uncertainty is corrected by the prior distribution. If the input x = [x1, x2, ..., x i Let f(x) be an i-dimensional variable and f(x) be the corresponding response function. Then the general model of f(x) can be expressed as:

[0057] f(x)~GP(m(x),V(x,x)),

[0058] In the formula, m(x) is the mean function, where m(x) = B T h(x); h(·) is a vector of predefined regression functions, usually h T (x)=(1,x T ), V is an unknown regression coefficient matrix with respect to h(·), where V(x,x) represents the covariance function and T represents the transpose form.

[0059] 2. Backpropagation (BP) method

[0060] The Backpropagation (BP) neural network algorithm is a multi-layer feedforward network trained using the backpropagation error algorithm, and it is one of the most widely used neural network models. BP networks can be used to learn and store large amounts of input-output model mappings without requiring explicit mathematical equations describing these mappings. Its learning rule uses the steepest descent method, adjusting the network's weights and thresholds through backpropagation to achieve the minimum sum of squared errors. During the modeling process, the output y of neuron j... j The calculation formula is as follows:

[0061]

[0062] Where θ j Let be the bias value, and sgn(·) be the activation function.

[0063] 3. Random Forest Algorithm (RF)

[0064] The RF regression algorithm is an ensemble learning algorithm that combines a large number of regression trees and is one of the representative machine learning algorithms. A regression tree represents a set of conditions or restrictions that are organized hierarchically and applied sequentially from the root to the leaves of the tree. Its main process includes:

[0065] a) Randomly sample and replace m samples from the original training data. Perform a total of n samplings to obtain n training sets.

[0066] b. For the obtained n training sets, train n decision tree models respectively.

[0067] c. For a single decision tree model, split it according to information gain and select the best feature.

[0068] d. The generated decision trees form a RF. After pruning and other operations, the final prediction value is determined by the outputs of the multiple decision trees.

[0069] 4. Distributed gradient boosting library algorithm XGBoost

[0070] The XGBoost model has been widely used in the construction of surrogate models due to its excellent learning efficiency and high training speed, and has received widespread attention. XGBoost constructs multiple decision trees. When constructing the i-th tree, it needs to fit the residuals generated by the regression of the previous i-1 trees. Each time a new tree is constructed, it needs to traverse the previous trees and select the tree that minimizes the objective function. We can express the model as:

[0071]

[0072] Where, For the prediction result, f k (x i Let f(·) be the model of the k-th tree in the previous iteration, Γ be the set of regression trees, and each f(·) be a function in Γ, x i , k, f k K represents the number of base models. Based on the above principles, the objective function can be constructed in the following form:

[0073]

[0074] Where, the first half of the target Let y be the loss function. i is the true value, Here, is the predicted value, and 'l' is the loss function algorithm used to describe the model's fit. The latter half... Ω is a regularization term used to prevent the model from overfitting and becoming too complex.

[0075] 5. Radial Basis Function Algorithm (RBF)

[0076] RBF has the advantages of strong approximation ability and fast learning convergence speed. Given an input sample set x = (x1, x2, ..., x... n By training the RBF model, the corresponding response value f(x) = (f(x1), f(x2), ..., f(x3)) can be obtained. n The RBF model is expressed as follows:

[0077]

[0078] Where, ω i The weighting coefficient for each sample, is a nonlinear radial basis function, and the training data point x i for The center of the Gaussian radial basis function is denoted as , where n is the number of sample points, and · represents the norm. A commonly used Gaussian radial basis function can be expressed as:

[0079]

[0080] 6. Kriging interpolation

[0081] The Kriging surrogate model is an interpolation model that minimizes the estimated variance. It consists of a global model and a local bias, and its descriptive form is as follows:

[0082]

[0083] Where, is the global approximate model of the design space, and Z(x) is the local deviation term.

[0084] The local deviation term Z(x) is a term with zero mean and variance σ. 2 , a random process with non-zero covariance. Its covariance matrix and related functions are as follows:

[0085] Cov[Z(x i ),Z(x j )]=σ 2 R[r(x i ,x j )],

[0086]

[0087] Where r is the correlation matrix and R is the correlation function. Variance σ 2 The least squares estimate is as follows:

[0088]

[0089] 7. Response Surface Methodology (RSM)

[0090] RSM uses regression equations to fit the functional relationship between variables and response values. Usually, a first-order or second-order response surface model in the form of a linear or quadratic polynomial is used for approximation. The basic form of a first-order response surface model is:

[0091]

[0092] The basic form of the second-order response surface model is as follows:

[0093]

[0094] Where, f RSM (x) is the response value, x i Let β be the i-th component of the n-dimensional variable x, β0, β i , β ij Let be a parameter, and ε be the error.

[0095] 8. Support Vector Regression (SVR) Algorithm

[0096] SVR is a supervised machine learning algorithm based on support vector machine, which can build a high-precision proxy model based on limited training sample data and has significant advantages in solving small sample problems. The SVR method can be expressed as:

[0097]

[0098] Where, Let C be the weight, C be the penalty factor, and ξ be the loss function, f(x) i () represents a nonlinear mapping. Introducing Lagrange multipliers and a kernel function, after minimizing the objective function, SVR can be expressed as:

[0099]

[0100] Where, α i and K(x) is a Lagrange multiplier, b is the intercept, and K(x) is a Lagrange multiplier. i Let x be the kernel function. The commonly used Gaussian kernel function can be expressed by the following formula.

[0101]

[0102] In the formula, x is the center of the kernel function, and γ is the parameter of the kernel function.

[0103] Based on the sample data generated by the experimental design method, surrogate models for the target to be optimized were constructed using the eight methods mentioned above. This section uses Mean Absolute Percentage Error (MAPE), Median Absolute Deviation (MAD), Root Mean Square Error (RMSE), and R-squared as indicators to evaluate the accuracy of the models, thereby measuring the applicability of different surrogate models to the problem under study. The specific expressions of the above model evaluation indicators are shown in the following formulas:

[0104]

[0105]

[0106]

[0107]

[0108] In the formula, f is the predicted value. The true value is n, the number of samples in the test set is n, med refers to the median, and mean refers to the mean.

[0109] In summary, a proxy model for structural parameters and performance targets is established through a data-driven approach.

[0110] Table 2 shows the evaluation metrics of MAPE, MAD, RMSE, and R-squared when using different surrogate modeling methods. Lower values ​​for MAPE, MAD, and RMSE indicate more accurate predictions of structural parameters and performance objectives by the surrogate model; conversely, a value closer to 1 for R-squared indicates higher prediction accuracy. The bolded metrics in the table represent the best performance across the various models. For the three optimization objectives, the XGBoost method showed the most satisfactory results in both MAPE and R-squared metrics.

[0111] Table 2 Structural parameters and performance target surrogate model evaluation index values ​​of different surrogate model methods

[0112]

[0113] The R-square metric is better able to reflect the model's fit and prediction accuracy, and is most commonly used in surrogate models and related research on structural parameters and performance objectives. Figures 3(a)-3(c) It shows the R-square value of the model when the three objectives to be optimized are modeled using different surrogate model methods. Figures 3(a)-3(c) As can be seen, for the three objectives, the R-squared value is closer to 1 when modeling using the XGBoost method, which indicates that the XGBoost surrogate model has higher accuracy for the three objectives studied.

[0114] Figures 4(a)-4(j) is the prediction accuracy of target three (motor efficiency η) on the test set of each agent model, from Figures 4(a)-4(j) It is not difficult to see that, for the research objective of motor efficiency η, although XGBoost has the best modeling accuracy among existing methods, there is still room for improvement in some aspects, and other corresponding methods (such as the Grey Wolf Optimization Algorithm) need to be adopted to improve the model accuracy.

[0115] S2 uses the Grey Wolf Optimization algorithm GWO to optimize the surrogate model of structural parameters and performance targets, and establishes a GWO-SVR surrogate model between design variables and optimization targets;

[0116] Specifically, the collected motor data is preprocessed to establish a sample set, which is divided into a training set and a test set. Optionally, the training set accounts for 80% of the total motor data, and the test set accounts for 20%. Of course, the proportions of the training set and the test set to the total motor data can be adjusted according to the actual situation. This embodiment does not impose specific limitations on this. The motor parameters of GWO are initialized, and the penalty factor and kernel parameters of the support vector regression (SVR) model are optimized to obtain the GWO-SVR surrogate model between the design variables and the optimization objective. The accuracy of the GWO-SVR surrogate model between the design variables and the optimization objective is verified through the test set.

[0117] Specifically, based on the established surrogate model of structural parameters and performance objectives, the Gray Wolf Optimization (GWO) algorithm is used to optimize the SVR model, establishing a GWO-SVR surrogate model between design variables and optimization objectives. GWO is a swarm intelligence heuristic optimization algorithm that mimics the leadership hierarchy and hunting mechanisms of gray wolves in nature to achieve optimization. Four types of gray wolves (α, β, δ, ω) are responsible for group decision-making, assisting in decision-making, scouting and hunting, and coordinating internal relationships. In the optimization process of the surrogate model of structural parameters and performance objectives, the distance between the gray wolves and prey is first determined, and then the positions of the gray wolves are updated to surround the target.

[0118]

[0119] Where D is the distance between the wolf and the prey, t is the current iteration number, X(t) and X p X(t) represents the positions of the prey and the gray wolf at iteration number t, respectively. X(t+1) represents the updated position of the gray wolf at iteration number t+1. C and A are vectors that control the behavior of the gray wolf.

[0120] Wolf α leads wolves β and δ to capture prey. The position of the gray wolf will also change in real time according to the target location. The distance between wolves α, β, δ and the prey and the position of wolves α, β, δ can be determined by the following formula:

[0121] D k =|C i X k (t)-X(t)|,

[0122] X i =X k -A i ·D k ,

[0123] Where k = α, β, δ, i = 1, 2, 3, and finally the position of wolf ω can be calculated by the position of wolf α, β, δ. The calculation formula is as follows:

[0124] X(t+1)=(X1+X2+X3)3,

[0125] The accuracy of the SVR model is related to the penalty factor C and the kernel parameter γ of the kernel function. The penalty factor C and the kernel parameter γ of the SVR are optimized using GWO. The flow chart of the improved GWO-SVR surrogate model between the design variables and the optimization objectives is shown in the figure below. Figure 5 As shown.

[0126] from Figure 5 As can be seen, the motor data is first preprocessed, with the training and test sets accounting for 80% and 20%, respectively. The GWO parameters are then initialized, and the SVR model parameters C and γ are optimized. Based on the optimal parameters C and γ, a GWO-SVR surrogate model is established between the design variables and the optimization objective using the training set data. The accuracy of this model is then verified on the test set.

[0127] To further demonstrate the advantage of the GWO-SVR surrogate model in terms of modeling accuracy between design variables and optimization objectives, MAPE, MAD, RMSE, and R-squared are used as evaluation metrics to assess the model's accuracy. Table 3 shows the model evaluation results for GWO-SVR, SVR, and XGBoost methods. The comparison shows that the R-squared value of the GWO-SVR surrogate model is close to 1 for all three optimization objectives. Furthermore, the calculation results of MAPE, MAD, and RMSE also indicate that the accuracy of the GWO-SVR surrogate model is superior to that of the original SVR and XGBoost. Therefore, the GWO-SVR method is chosen to establish a surrogate model between design variables and optimization objectives, serving as preparation for subsequent optimization work.

[0128] Table 3. Model evaluation metrics for GWO-SVR, SVR, and XGBoost methods.

[0129]

[0130] Better motor external characteristics and efficiency can be achieved through the GWO-SVR proxy model between design variables and optimization objectives.

[0131] Specifically, take the optimization model of a high power density AC induction drive motor as an example:

[0132] ① Optimization variable determination

[0133] When the basic configuration parameters of the powertrain motor are determined, the main factors affecting the motor's external characteristic parameters are the stator and rotor structural dimensions. These dimensions significantly influence motor losses, thus affecting motor efficiency. Therefore, the structural dimensions affecting the motor's external characteristic parameters and efficiency are used as optimization variables, expressed as follows:

[0134] X = [Bs0, Hs0, Hs2, Bs1, Bs2].

[0135] ② Optimize the establishment of the objective function

[0136] a. Motor base speed n b Motor base speed n b It is the critical point of the speed between the constant torque area and the constant power area of ​​the motor. A larger base speed indicates that the powertrain motor has a larger constant torque area and a better speed regulation range. b As the optimization objective, it can be expressed as:

[0137] f nb = f(Bs0,Hs0,Hs2,Bs1,Bs2).

[0138] b. Maximum output power P of the motor max Maximum output power P of the motor max It is the characteristic parameter of the external characteristic curve of the powertrain motor, which increases the maximum output power P max It can increase the area of ​​constant power zone, thus improving the external characteristics of the motor. max As the optimization objective, it can be expressed as:

[0139] f Pmax = f(Bs0,Hs0,Hs2,Bs1,Bs2).

[0140] c. Motor efficiency η: For electric vehicle powertrain motors, higher motor efficiency η can improve energy utilization and effectively increase the driving range of electric vehicles. This section uses rated efficiency η as the optimization target, which can be expressed as:

[0141] f η = f(Bs0,Hs0,Hs2,Bs1,Bs2).

[0142] ③Constraints

[0143] Different stator and rotor slot structural parameters of powertrain motors affect the motor's external characteristics and efficiency. Considering the allowable processing capabilities of these parameters, they should be controlled within reasonable ranges. If any structural parameter (such as stator or rotor slot structural parameters) is too small, it increases processing difficulty; if any structural parameter (such as stator or rotor slot structural parameters) is too large, it may lead to slot (stator and rotor slot) interference. Therefore, the optimization variables should satisfy the following constraints:

[0144] Structural parameter constraints:

[0145] a. The stator slot width Bs0 of the induction motor shall be within the range of 2.7≤Bs0≤3.5, unit: mm;

[0146] b. The value range of the induction motor rotor slot height Hs0 is 0.25≤Hs0≤0.75, unit: mm;

[0147] c. The value range of the stator slot width Hs2 of the induction motor is 15≤Hs2≤20, unit: mm;

[0148] d. The value range of the stator slot width Bs1 of the induction motor is 2.0≤Bs1≤3.2, unit: mm;

[0149] e. The value range of the stator slot width Bs2 of the induction motor is 0.8≤Bs2≤1.8, unit: mm.

[0150] Motor starting performance constraints:

[0151] The choice of rotor slot shape has a conflicting impact on motor efficiency and starting torque. While closed slots improve motor efficiency, starting torque decreases. Insufficient starting torque in the powertrain motor can prevent electric vehicles from starting or extend starting times, leading to additional losses and power consumption.

[0152] To reduce losses, the powertrain motor uses a cast copper rotor with high conductivity. This causes the rotor to generate a large current and induced electromotive force (EMF). The induced EMF generates a magnetic field that cancels out the excitation magnetic field generated by the stator. To maintain a constant excitation magnetic field, the stator windings increase the current in the windings, thus increasing the starting current. For powertrain motors used in electric vehicles, the starting current must be within a safe range to prevent the windings from overheating, causing power loss and severe winding shock.

[0153] Therefore, while optimizing the external characteristics and efficiency of the powertrain motor, it is necessary to ensure that the starting torque and starting current are within the motor's feasible range to ensure the motor's starting performance. When establishing the GWO-SVR proxy model between the design variables and the optimization objective, the starting torque and starting current are used as constraints.

[0154] f. Starting torque T st To ensure the starting performance of the powertrain motor, the starting torque should be large enough to overcome static friction and load. To ensure the starting performance of the powertrain motor for electric vehicles, the starting torque multiple is usually between 1.7 and 2.2. To meet the minimum starting torque requirement, the starting torque multiple must be no less than 1.7, i.e., 212.5 Nm.

[0155] g. Starting current I st The starting current must be within a safe range to avoid additional power loss and severe winding shock; limit the starting current multiple to 210A.

[0156] In summary, the GWO-SVR proxy model between design variables and optimization objectives is expressed as follows:

[0157] find x=[Bs0,Hs0,Hs2,Bs1,Bs2]

[0158] maxF(x)=(max n b ,maxP max ,maxη)

[0159]

[0160] Where Bs0 is the stator slot width of the induction motor, Hs0 is the stator and rotor slot height of the induction motor, Hs2 is the stator and rotor slot height of the induction motor, Bs1 is the upper slot width of the stator and rotor of the induction motor, Bs2 is the lower slot width of the stator and rotor of the induction motor, and T st I is the motor torque. st Input current to the motor.

[0161] S4 uses structural parameters as decision vectors and maximizes the motor base speed, maximum motor output power, and motor efficiency as optimization objectives. It uses a multi-objective optimization algorithm to solve the GWO-SVR surrogate model between design variables and optimization objectives to obtain the optimal solution.

[0162] Specifically, the motor design parameters and constraint range are determined, and an initial population of motor design parameters is generated based on these parameters and constraints. Virtual fitness values ​​of individuals in the initial population are calculated using a non-dominated sorting method. According to the definition of non-dominated sorting, for multiple objectives seeking minimization, the crowding coefficient id in the same dominance layer is calculated. For each level of non-dominated solution layer, individuals are sorted in ascending order based on their objective values. For each individual, the perimeter of the cuboid with its nearest neighbor as a vertex is calculated as the crowding coefficient, yielding the relationship curves for maximizing motor base speed, maximum motor output power, and motor efficiency. The initial motor design parameters are compared with the motor design parameters optimized by the multi-objective optimization algorithm to obtain the optimal solution.

[0163] Specifically, the optimization model was solved using three multi-objective optimization algorithms in the PlatEMO toolbox: the multi-stage constrained multi-objective evolutionary algorithm (CMOEA-MS), ToP, and the non-dominated sorting genetic algorithm II (NSGA-II). Matlab programming was used to optimize and solve the established model.

[0164] CMOEA-MS is a two-stage evolutionary algorithm that adjusts the fitness evaluation strategy during the evolution process to adaptively balance objective optimization and constraint satisfaction. ToP is a two-stage algorithm that transforms multi-objective problems into single-objective problems. NSGA-II proposes a fast non-dominated sorting strategy based on NSGA and adds an elite mechanism. Its algorithm flow is as follows: Figure 6 As shown.

[0165] continue Figure 6 As shown, the crowding coefficient in the same dominant layer is calculated as id. For each level of non-dominated solution layer, the target values ​​of the individuals are sorted in ascending order. For each individual, the perimeter of the cuboid with the nearest neighbor as the vertex is calculated as the crowding coefficient. The relationship curves of the maximum motor base speed, the maximum motor output power, and the motor efficiency are obtained, including:

[0166] First, the new population Q generated in generation t... t With parent generation P t Combined to form R t The population size is 2N;

[0167] Then, R t Perform a non-dominated sort to generate a series of non-dominated sets Z. i And calculate the crowding degree; since both offspring and parent individuals are contained in R t In the non-dominated set Z1 after non-dominated sorting, the individuals contained in it are R. t The best one is Z, so let's start with Z. i Insert a new parent population P t+1 In the middle, if P t+1 If the value is less than N, then continue to P. t+1 Fill the next level non-dominated set Z2 in the middle until Z is added. k When the population size exceeds N, the Z i Individuals in the cluster are compared using the crowding degree operator, taking the top {num(Z)} k )-(num(P t+1 A population of individuals is formed by selecting N individuals and then using genetic operators (selection, crossover, mutation) to generate a new offspring population Q. t+1 .

[0168] After determining the motor design parameters, the NSGA-II is used to iteratively optimize the GWO-SVR surrogate model between the design variables and the optimization objective.

[0169] In the optimization solution, the optimal individuals were set to 40, and the number of iterations was 1000. The three algorithms yielded 40 Pareto solution sets, as follows: Figure 7 As shown.

[0170] Because the three objectives are Pareto-dependent, it is impossible to obtain all three optimal solutions simultaneously. Therefore, a method is needed to find the "optimal solution" that meets the optimization requirements. Here, the Utopian point method is chosen for multi-criteria decision-making. The optimization objective is to maximize the basic velocity n. b Maximum output power P max And considering the motor efficiency η, according to the principle of the Utopian point method, the further the Utopian point (the "optimal" design point) is from the origin, the better. A set of solutions from the middle parts of the Pareto fronts of each of the three optimization methods (CMOEA-MS, ToP, and NSGA-II) are selected to form candidate solutions. The candidate solutions for different optimization algorithms are shown in Table 4.

[0171] Meanwhile, to verify the reliability of the optimization results, the actual target values ​​were obtained by simulating the stator and rotor slot parameters of the motor using three sets of candidate schemes, and then compared with the corresponding optimized values. Table 4 shows that the error of the target parameter values ​​of NSGA-II is relatively smaller compared to CMOEA-MS and ToP; therefore, the solution of this algorithm was selected for result analysis.

[0172] The initial motor design parameters were compared with the optimized scheme (motor design parameters) obtained from the NSGA-II algorithm, and the results are shown in Table 5. As can be seen from the data in Table 5, the motor's external characteristic parameters and efficiency were improved. Design variables Bs1, Bs2, and Hs2 increased by 22.27%, 5.04%, and 3.23%, respectively, while Bs0 and Hs0 decreased by 5.55% and 48%, respectively. The optimized motor base speed n... b and the maximum output power P of the motor max The speed and power outputs increased by 40.18 rpm and 0.53 kW respectively, resulting in a 0.69% increase in motor efficiency.

[0173] Table 4 Optimal solutions for different optimization algorithms

[0174]

[0175] Table 5 Design parameter values ​​before and after optimization

[0176]

[0177] External characteristic curves before and after optimization are as follows: Figure 8As shown in Figure 9, the stator and rotor slot shapes before and after optimization are as follows. Figures 9(a) and 9(b) show that the stator slot width Bs0 narrows, the rotor slot width increases, and the slot depth Hs2 decreases. This is because the reduced stator slot width Bs0 lowers the harmonic magnetic field in the air gap and the resulting power loss, thus improving the motor's external characteristics. The increased upper slot width Bs1 and lower slot width Bs2 by 0.06 mm increases the upper slot area, reducing rotor resistance and leakage reactance to some extent, decreasing copper losses during steady-state operation, and improving the motor's external characteristics and efficiency. However, this slot width design reduces starting torque and starting current. Simultaneously, the increased rotor slot depth Hs2 by 0.58 mm reduces rotor resistance, improving motor efficiency, but this is detrimental to reducing rotor starting current.

[0178] Finally, the optimal solution is verified as follows:

[0179] Based on the optimization results, experimental verification was conducted. By testing and comparing the characteristic parameters of the motor's external characteristic curve and measuring the efficiency at the rated operating point, the improved external characteristics and efficiency of the powertrain motor were verified.

[0180] (1) External characteristics optimization experimental verification

[0181] The powertrain motor's external characteristic curve includes a constant torque region and a constant power region, therefore, the motor's base speed n is... b and the maximum output power of the motor P max Test experiments were conducted to verify the effectiveness of stator and rotor parameter optimization in improving the constant torque and constant power regions.

[0182] ① Motor base speed

[0183] motor base speed n b It is the maximum speed that the powertrain motor can reach when outputting rated torque. The load torque of the powertrain motor is adjusted to the rated torque Trated through the bench calibration software. When the motor speed reaches the maximum, the speed at this time is the motor base speed n b The experimental measurement curve is as follows: Figure 10 As shown. Figure 10 It can be seen that the motor speed gradually stabilizes at 6538.32 rpm in 5.5 s, and the optimized value of the motor base speed is 6540.18 rpm. The error between the experimental data and the optimized value is 0.028%, which is within a reasonable range, indicating that the optimization result has high reliability.

[0184] ② Maximum output power of the motor

[0185] When the powertrain motor load is constant, the motor output power is directly proportional to the motor speed and increases with increasing speed. When the motor speed reaches the base speed n...max At this point, the motor output power is the maximum output power. The motor output power curve is shown in Figure 11. From Figure 11 As can be seen, the motor output power gradually stabilized at 85.58kW in 5.5s, and the optimized maximum output power of the motor was 85.60kW. The error between the experimental data and the optimized value was 0.023%, which is within a reasonable range, indicating that the optimization results have high reliability.

[0186] (2) Efficiency optimization experiment verification

[0187] This section verifies the effect of motor stator and rotor slot dimensions on motor efficiency optimization by measuring the motor efficiency under rated operating conditions and comparing it with the motor efficiency before structural optimization. The motor efficiency measurement process is as follows: the motor speed is set to 6500 r / min, the load torque of the powertrain motor is adjusted to the rated torque using bench calibration software, and the motor efficiency at rated motor speed and output torque is measured using a power analyzer.

[0188] The powertrain motor is a dynamic system, and its operating conditions are not absolutely steady-state. To ensure the accuracy of motor efficiency measurement, the motor efficiency was continuously measured at its rated operating conditions using a power analyzer, and the average value was taken as the motor efficiency under rated conditions. Furthermore, using the above method, the motor efficiency at different speeds and torques was measured sequentially. When the powertrain motor operates under rated conditions, the motor efficiency reaches a maximum value of 92.12%, and the optimized efficiency value is 92.174%. The error between the experimental data and the optimized value is 0.0058%, which is within a reasonable range, indicating that the optimization results have high reliability and verifying the reliability of the algorithm.

[0189] The rationality and applicability of the vehicle motor electromagnetic performance optimization method in this application are verified by comparing the optimized simulation results with the actual performance measurements of product engineering samples (such as motor stator slots and motor rotor slots).

[0190] As can be seen from the above embodiments, the vehicle motor electromagnetic performance optimization method provided by the present invention achieves at least the following beneficial effects:

[0191] The vehicle motor electromagnetic performance optimization method provided by this invention obtains simulation experimental data through experimental design, including motor data. Machine learning algorithms are used to train the simulation experimental data to establish a surrogate model between structural parameters and performance objectives. The Grey Wolf Optimization (GWO) algorithm is then used to optimize the surrogate model, establishing a GWO-SVR surrogate model between design variables and optimization objectives. Using structural parameters as the decision vector and maximizing the motor base speed, maximum output power, and efficiency as optimization objectives, a multi-objective optimization algorithm is used to solve the GWO-SVR surrogate model between design variables and optimization objectives to obtain the optimal solution. This method improves the motor base speed and maximum output power, thereby enhancing the motor's external characteristics and efficiency. Based on the GWO-SVR surrogate model between design variables and optimization objectives, motor design parameters are determined, and optimal constraints are applied to performance objectives and material usage, quickly providing the optimal design scheme and improving motor development efficiency and accuracy.

[0192] Example 2

[0193] This embodiment provides a vehicle motor electromagnetic performance optimization system, including:

[0194] The proxy model establishment module 100 between structural parameters and performance targets is used to obtain simulation experimental data through experimental design. The simulation experimental data includes motor data. The simulation experimental data is trained using machine learning algorithms to establish a proxy model between structural parameters and performance targets.

[0195] The GWO-SVR surrogate model building module 200 is used to optimize the surrogate model of structural parameters and performance objectives using the Grey Wolf Optimization Algorithm (GWO), and to establish a GWO-SVR surrogate model between design variables and optimization objectives.

[0196] The calculation module 300 is used to solve the GWO-SVR surrogate model between the design variables and the optimization objectives, with structural parameters as the decision vector and maximizing the motor base speed, maximum motor output power, and motor efficiency as the optimization objectives, using a multi-objective optimization algorithm to obtain the optimal solution.

[0197] Example 3

[0198] An electronic device, comprising:

[0199] Memory, used to store programs; and

[0200] The processor is used to execute any of the vehicle motor electromagnetic performance optimization methods provided in this application by calling a program stored in memory.

[0201] Example 4

[0202] A computer-readable storage medium storing computer instructions for causing a computer to execute any of the vehicle motor electromagnetic performance optimization methods provided in this application.

[0203] Although some specific embodiments of the present invention have been described in detail by way of examples, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It should be understood by those skilled in the art that modifications may be made to the above embodiments without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A method for optimizing electromagnetic performance of a vehicle motor, characterized in that: include: Acquire simulation experiment data through experimental design, wherein the simulation experiment data includes motor data, and train the simulation experiment data using a machine learning algorithm to establish a proxy model between structural parameters and performance targets; The Grey Wolf Optimization (GWO) algorithm is used to optimize the proxy model of the structural parameters and performance targets, and a GWO-SVR proxy model between the design variables and the optimization targets is established; Taking the structural parameters as the decision vector and maximizing the motor base speed, the motor maximum output power and the motor efficiency as the optimization objectives, the GWO-SVR proxy model between the design variables and the optimization objectives is solved using a multi-objective optimization algorithm to obtain the optimal solution.

2. The vehicle motor electromagnetic performance optimization method according to claim 1, characterized in that: The method of optimizing the proxy model of the structural parameters and performance targets by using the Grey Wolf Optimization Algorithm GWO and establishing the GWO-SVR proxy model between the design variables and the optimization targets includes: Perform data preprocessing on the collected motor data, establish a sample set, and divide the sample set into a training set and a test set; Initializing the motor parameters of the GWO, and optimizing the penalty factor and kernel parameters of the support vector regression SVR model to obtain a GWO-SVR proxy model between the design variables and the optimization target; The accuracy of the GWO-SVR proxy model between the design variables and the optimization objectives is verified through a test set.

3. The vehicle motor electromagnetic performance optimization method according to claim 1, characterized in that: The structural parameters are used as decision vectors, and the maximization of the motor base speed, the motor maximum output power and the motor efficiency are used as optimization targets. The GWO-SVR proxy model between the design variables and the optimization targets is solved using a multi-objective optimization algorithm to obtain the optimal solution, which includes: Determining motor design parameters and constraint ranges, and generating an initial population of motor design parameters according to the motor design parameters and constraint ranges; The virtual fitness values ​​of individuals in the initial population are calculated by non-dominated sorting. According to the definition of non-dominated sorting, for multiple objectives to be minimized; Calculate the crowding coefficient in the same dominance layer as id. For each level of non-dominated solution layer, sort them in ascending order according to the target value of the individual. For each individual, calculate the perimeter of the cuboid with the nearest neighbor as the vertex as the crowding coefficient, and obtain the relationship curve between the maximum motor base speed, the maximum motor output power and the motor efficiency. The initial motor design parameters are compared with the motor design parameters optimized by the multi-objective optimization algorithm to obtain the optimal solution.

4. The vehicle motor electromagnetic performance optimization method according to claim 3, characterized in that: The crowding coefficient in the same dominating layer is calculated as id. For each level of non-dominated solution layer, the target values ​​of the individuals are sorted in ascending order. For each individual, the perimeter of the cuboid with the nearest neighbor as the vertex is calculated as the crowding coefficient. The relationship curve between the maximum motor base speed, the maximum motor output power and the motor efficiency is obtained, including: The new population Q generated in the tth generation t With the parent P t Combined to form R t , the population size is 2N; R t Perform non-dominated sorting to generate a series of non-dominated sets Z i And calculate the crowding degree; since both the offspring and the parent individuals are included in R t Then the individuals contained in the non-dominated set Z1 after non-dominated sorting are R t The best one, so put Z first i Add the new parent population P t+1 If P t+1 If the size is smaller than N, then continue to P t+1 Fill the next level of non-dominated set Z2 until Z is added k When the population size exceeds N, the Z i The individuals in the crowding comparison operator are used to take the first {num(Z k )-(num(P t+1 )-N)} individuals, so that the number of individuals reaches N, and then a new offspring population Q is generated through genetic operators t+1 .

5. The vehicle motor electromagnetic performance optimization method according to claim 1, characterized in that: The motor data includes design parameters and performance parameters, wherein the design parameters include but are not limited to motor stator structure dimensions, motor rotor structure dimensions, magnetic steel data and winding data; the performance parameters include but are not limited to motor torque, motor efficiency, motor input current, motor maximum output power and motor base speed.

6. The vehicle motor electromagnetic performance optimization method according to claim 1, characterized in that: The machine learning algorithm includes genetic programming, error back propagation method, random forest algorithm, distributed gradient boosting library algorithm, radial basis function algorithm, Kriging interpolation, response surface analysis method or support vector regression algorithm.

7. A vehicle motor electromagnetic performance optimization system, characterized in that: include: A proxy model establishment module between structural parameters and performance targets, used to obtain simulation experimental data through experimental design, the simulation experimental data including motor data, train the simulation experimental data using a machine learning algorithm, and establish a proxy model between structural parameters and performance targets; A GWO-SVR proxy model building module is used to optimize the proxy model of the structural parameters and performance targets using the Grey Wolf Optimization Algorithm GWO, and to establish a GWO-SVR proxy model between the design variables and the optimization targets; The calculation module is used to use the structural parameters as decision vectors, and to maximize the motor base speed, the motor maximum output power and the motor efficiency as optimization objectives, and to solve the GWO-SVR proxy model between the design variables and the optimization objectives using a multi-objective optimization algorithm to obtain the optimal solution.

8. An electronic device, characterized in that: The electronic device comprises: memory for storing programs; and A processor, configured to execute the vehicle motor electromagnetic performance optimization method as claimed in any one of claims 1 to 6 by calling the program stored in the memory.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the vehicle motor electromagnetic performance optimization method according to any one of claims 1-6.

Citation Information

Patent Citations

  • Agent-aided optimization design method and system for permanent magnet motor

    CN111985064A

  • Method for optimizing fuel economy and dynamic property of tractor based on multi-objective genetic algorithm

    CN114019799A

  • Method for optimizing external characteristics of induction motor based on proxy model

    CN115374664A

  • Integrated performance optimization method for material structure of induction motor

    CN116484665A

  • Machine learning-based disaster modeling and high-impact weather event forecasting

    US20220343221A1

Cited By

  • Outer rotor permanent magnet brushless direct current motor optimization design method, device, equipment and medium

    CN120951771A

  • Optimization design method, device and equipment of outer rotor permanent magnet brushless direct current motor and medium

    CN120951771B

  • Motor reverse design method and system based on proxy model and active learning

    CN122088294A

  • A motor reverse design method and system based on a proxy model and active learning

    CN122088294B

  • Bayesian active learning motor multi-objective optimization method based on physical prior guidance

    CN122549118A