Wireless charging electromagnetic shielding optimization design method based on machine learning

Abstract

The invention relates to the technical field of wireless electric energy transmission, and provides a wireless charging electromagnetic shielding optimization design method based on machine learning, which can quickly predict the shielding performance and greatly improve the design efficiency by constructing a machine learning agent model. A multi-objective optimization algorithm is adopted to comprehensively consider the shielding efficiency, the energy transmission efficiency and the material cost, and diversified actual engineering requirements are met; a closed-loop optimization mechanism of simulation verification and dynamic adjustment is introduced, and it is guaranteed that the final design reaches an expected target; data distribution is optimized through Latin hypercube sampling, a high-quality training database is constructed in combination with data cleaning and normalization processing, and the model generalization ability and prediction precision are improved; optimization of various shielding structure parameters is supported, adaptability is high, and the method can be widely applied to wireless charging scenes; the test frequency is remarkably reduced, the design period is shortened, the development cost is reduced, and meanwhile the shielding performance, transmission efficiency and economical efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless power transmission, and in particular to a wireless charging electromagnetic shielding optimization design method based on machine learning. Background Art

[0002] Wireless Power Transfer (WPT) is an energy transmission technology based on electromagnetic field coupling, which is widely used in electric vehicles, consumer electronics, industrial automation, medical equipment and other fields. It has the advantages of non-contact transmission, convenience, flexibility, safety and reliability. However, in practical applications, wireless charging systems usually operate in the medium and high frequency range (such as kHz to MHz), and the electromagnetic radiation generated by them may cause the following problems.

[0003] The first is electromagnetic interference (EMI). The high-frequency electromagnetic field in the wireless charging system can easily interfere with surrounding electronic equipment and affect their normal operation. The second is a hidden danger to human health. Long-term exposure to a strong electromagnetic radiation environment may have an adverse effect on human health. The third is reduced energy transmission efficiency. During the wireless charging process, electromagnetic leakage not only interferes with the outside world, but may also cause energy loss in external metal objects, affecting transmission efficiency.

[0004] In order to deal with the above problems, electromagnetic shielding design is widely used in wireless charging systems. However, traditional shielding design methods mainly rely on finite element simulation or experimental verification. Although relatively accurate results can be obtained, there are the following problems. First, the calculation cost is high. The traditional method requires the use of finite element software for analysis, which consumes a lot of computing resources and time for complex shielding structures and multi-parameter designs. Second, there is a lack of flexibility. Design schemes often require repeated adjustment of parameters and re-simulation or experiments, which makes it difficult to quickly adapt to different application requirements. Third, the optimization efficiency is low. Multi-objective optimization problems (such as the trade-off between shielding efficiency, manufacturing cost, and material utilization) are difficult to implement in traditional methods and usually require manual intervention.

[0005] Based on this, the present invention provides a wireless charging electromagnetic shielding optimization design method based on machine learning to improve design efficiency, reduce development costs, and provide technical support for the promotion and application of wireless charging technology. Summary of the invention

[0006] The purpose of this invention is to address the deficiencies in the prior art and to provide a wireless charging electromagnetic shielding optimization design method based on machine learning to improve design efficiency, reduce development costs, and provide technical support for the promotion and application of wireless charging technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A wireless charging electromagnetic shielding optimization design method based on machine learning includes the following steps:

[0009] Step 1: Data preparation: Collect shielding structure design parameters, obtain corresponding electromagnetic shielding performance data through simulation calculation and experimental test, and build a training sample database;

[0010] Step 2: Constructing a machine learning proxy model: Using shielding structure design parameters as independent variables and corresponding electromagnetic shielding performance data as dependent variables, construct a machine learning proxy model;

[0011] Step 3: Model training: Based on the training sample database obtained in step 1, the machine learning proxy model in step 2 is trained, and the model prediction performance is verified by cross-validation and independent test sets to obtain a trained machine learning proxy model;

[0012] Step 4: Construct a multi-objective optimization function. Calculate the electromagnetic shielding performance data based on the machine learning agent model trained in step 3, and construct a multi-objective optimization function in combination with the constraints of the actual project.

[0013] Step 5: Optimization solution and verification;

[0014] Step 5.1, optimization solution; after normalizing the shielding structure design parameters to be optimized, input the machine learning agent model in step 3, and based on the multi-objective optimization function in step 4, use the multi-objective optimization algorithm to search for the Pareto optimal solution set, and determine the corresponding optimal shielding structure design parameters from the Pareto optimal solution set;

[0015] Step 5.2, optimization verification and adjustment; verify the actual effect of the optimization scheme. If the actual effect of the optimization scheme fails to meet the design requirements, return to the optimization process, adjust the relevant design parameters, and optimize again based on the optimization methods of steps 3 to 5.1 until the simulation and test results meet the expected goals.

[0016] Preferably, the step 1 specifically includes the following steps:

[0017] Step 1.1, obtain shielding structure design parameters, and obtain corresponding electromagnetic shielding performance data through simulation calculation and experimental test;

[0018] Step 1.2: Improve data quality by optimizing sample distribution through Latin hypercube sampling;

[0019] Step 1.3, data cleaning: check the outliers or noise in the sample data and remove the data points with large simulation errors;

[0020] Step 1.4, normalization processing: normalize all sample data;

[0021] Step 1.5, data annotation: generating corresponding electromagnetic shielding performance data for each sample data;

[0022] Step 1.6, data set division: divide the sample data into training set, validation set and test set in a ratio of 7:2:1.

[0023] Preferably, the shielding structure design parameters include at least: relative size parameters of the coil length, width and height; relative size parameters of the magnetic core length, width and height; relative size parameters of the shielding aluminum plate length, width and thickness; the electromagnetic shielding performance data include at least: shielding efficiency, energy transmission efficiency, and shielding material cost.

[0024] Preferably, in step 2, Gaussian process regression and neural network are used to construct a machine learning agent model.

[0025] Preferably, the step 4 specifically includes the following steps:

[0026] Step 4.1, determine the optimization objectives and constraints;

[0027] The optimization objectives include: ① Maximization of shielding efficiency, the specific formula is:

[0028]

[0029] Where SE is the shielding efficiency, H 无屏蔽 is the electric field strength without shielding structure, H 有屏蔽 is the electric field strength when there is a shielding structure;

[0030] ②Maximize energy transmission efficiency, which can be expressed as:

[0031]

[0032] Where η is the energy transfer efficiency, P 输入 is the system input power; P 输出 is the effective output power of the receiving coil;

[0033] ③ Minimize material cost, the calculation formula is:

[0034] C=ρ 铝 ·V 铝 ·P 铝 +ρ 磁芯 ·V 磁芯 ·P 磁芯 ;

[0035] Where C is the material cost, ρ 铝V is the density of the shielding aluminum plate material; 铝 is the volume of the shielding aluminum plate; P 铝 is the unit price of the shielding aluminum plate material; ρ 磁芯 is the density of the core material; V 磁芯 is the volume of the magnetic core; P 磁芯 is the unit price of the core material;

[0036] Constraints include: Shielding efficiency constraint, shielding efficiency must meet the minimum standard: SE ≥ SE min , the degree of violation is expressed as: Penalty SE =max(0,SE min -SE); Energy transmission efficiency constraint, energy transmission efficiency must meet the minimum standard: η≥ηmin, and the degree of violation is expressed as: Penalty η =max(0,η min -η); Geometric dimensions and process constraints. The geometric dimensions must meet the design space and processing technology requirements. The degree of violation is expressed as:

[0037] Step 4.2, construct a multi-objective optimization function;

[0038] The objective function adopts a weighted form to transform multiple objectives into a single objective for optimization. The objective function is defined as:

[0039] F=w1·f1+w2·f2+w3·f3+λ·Penalty;

[0040] Where: F is the comprehensive optimization objective function; f1, f2, f3 are the shielding efficiency objective function, energy transmission efficiency objective function and material cost objective function respectively, f1 = -SE, f2 = -η, f3 = C; w1, w2, w3 are the weight coefficients of shielding efficiency, energy transmission efficiency and material cost respectively, satisfying w1 + w2 + w3 = 1; λ is the penalty function coefficient, which is used to adjust the influence of violation of constraint conditions on the objective function; Penalty is the penalty function term of constraint conditions, which is used to indicate the degree of violation of constraint conditions.

[0041] Penalty=Penalty SE +Penalty η +Penalty 尺寸 .

[0042] Preferably, the normalization process in step 1.4 is based on the coil size, and the magnetic core size and the shielding aluminum plate size are standardized as relative parameters.

[0043] Preferably, in step 5.1, the actual effect of the optimization scheme is verified by combining simulation analysis and physical testing.

[0044] Preferably, the multi-objective optimization algorithm is a genetic algorithm or a particle swarm algorithm.

[0045] The present invention discloses a wireless charging electromagnetic shielding optimization design method based on machine learning, which has the following beneficial effects.

[0046] The present invention constructs a machine learning agent model to quickly predict shielding performance and greatly improve design efficiency; adopts a multi-objective optimization algorithm to comprehensively consider shielding efficiency, energy transmission efficiency and material cost to meet diverse practical engineering needs; introduces a closed-loop optimization mechanism of simulation verification and dynamic adjustment to ensure that the final design achieves the expected goals; optimizes data distribution through Latin hypercube sampling and combines data cleaning and normalization processing to build a high-quality training database to improve model generalization ability and prediction accuracy; supports the optimization of multiple shielding structure parameters, has strong adaptability, and can be widely used in wireless charging scenarios; significantly reduces the number of experiments, shortens the design cycle, reduces development costs, and at the same time improves shielding performance, transmission efficiency and economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the overall framework of the wireless charging electromagnetic shielding optimization method of the present invention.

[0048] Figure 2 This is the distribution of training sample data sets in the embodiment of the present invention.

[0049] Figure 3 It is a schematic diagram comparing the prediction results and simulation results of the machine learning agent model of the present invention.

[0050] Figure 4 The present invention uses genetic algorithm optimization and adopts Pareto frontier to display the optimal optimization result diagram.

[0051] Figure 5 It is a display diagram of the optimization effect of the magnetic field shielding structure of the present invention, wherein a is the magnetic field distribution diagram before optimization, b is the traditional finite element optimization magnetic field distribution diagram, and c is the magnetic field distribution after optimization by the optimization method proposed in the present invention.

[0052] Figure 6 This is a schematic diagram comparing the optimization speeds of the method proposed in the present invention and traditional finite element simulation. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0054] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the various technical features mentioned in the embodiments can be combined in any way to form a corresponding implementable technical solution.

[0055] Unless otherwise defined, the technical terms used in this document have the same meanings as those generally understood by those skilled in the art to which this application belongs; the use of relevant terms in this document is only for describing specific embodiments and is not intended to limit this application.

[0056] In the description of this application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" logical relationship.

[0057] Unless otherwise expressly specified or limited, in the description of the embodiments of the present application, the terms such as "install", "connect", "connect", "fix", "set", etc. used should be understood in a broad sense. For example, the "connection" can be a fixed connection, a detachable connection, or an integrated setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For technicians in the technical field to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0058] Example

[0059] Please refer to Figure 1 , a wireless charging electromagnetic shielding optimization design method based on machine learning, comprising the following steps:

[0060] Step 1: Data preparation: Collect shielding structure design parameters, obtain corresponding electromagnetic shielding performance data through simulation calculation and experimental test, and build a training sample database;

[0061] Preferably, in this embodiment, step 1 specifically includes the following steps:

[0062] Step 1.1. Obtain the design parameters of the shielding structure, and obtain the corresponding electromagnetic shielding performance data through simulation calculations and experimental tests; it should be understood that the design parameters of the existing shielding structure and its corresponding electromagnetic shielding performance data can also be directly collected. The data range needs to cover common wireless charging system specifications, while ensuring a certain degree of randomness to improve the representativeness of the data and the generalization ability of the model. Use finite element simulation tools (such as ANSYS, COMSOL, etc.) to simulate and analyze the electromagnetic shielding structure under different parameter combinations to obtain key performance indicators, such as:

[0063] Shielding Effectiveness (SE): Measures the effectiveness of the shielding structure in suppressing electromagnetic leakage.

[0064] Energy transfer efficiency: The impact of optimized shielding design on the efficiency of wireless charging systems.

[0065] Shielding material costs: Estimate material consumption for designs of different sizes.

[0066] Experimental verification: For typical parameter combinations, sample data is obtained through actual shielding structure tests;

[0067] Step 1.2, optimize the sample distribution through Latin hypercube sampling to improve data quality; generate representative sample points in the multidimensional space through the uniformly distributed parameter sampling method to ensure the uniformity of the sample distribution and reduce the computational overhead; as an option, parameter sensitivity analysis and active learning (Active Learning) can also be used in this embodiment to improve data quality; parameter sensitivity analysis: analyze the impact of different parameters on the shielding performance, give priority to generating parameter samples that have a greater impact on the performance, and reduce redundant calculations; active learning (Active Learning): during the optimization process, dynamically select sample points with higher information value for simulation or testing, gradually improve the sample data set, and improve the prediction accuracy of the proxy model.

[0068] Step 1.3: Data cleaning: Check the outliers or noise in the sample data, remove the data points with large simulation errors, and ensure data quality;

[0069] Step 1.4, normalization processing; normalize all sample data; take the coil size as the benchmark, standardize the core size and shielding aluminum plate size into relative parameters (such as relative length, relative thickness), which is convenient for the learning and prediction of the proxy model. It can improve the versatility and applicability of the shielding optimization process and ensure that wireless charging devices of different structures and sizes can efficiently complete the shielding design. The normalization method and specific formula are as follows:

[0070] First, define the normalization coefficient

[0071]

[0072] Among them, Lmax is the maximum length of the coil length, width, height, and further, kn can be used to obtain the dimensionless representation of the system after normalization. The dimensionless representations Lcoil_n, Wcoil_n, and Hcoil corresponding to Lcoil, Wcoil, and Hcoil are obtained. The normalized size coefficients KLc, KWc, and KHc of the length, width, and height of the shielded magnetic core are obtained, as well as the normalized coefficients KLs, Kws, and KHs of the length, width, and height of the shielded aluminum plate.

[0073] L coil_n =k n ·L coil , W coil_n =k n ·W coil , H coil_n =k n ·H coil

[0074]

[0075] Step 1.5, data annotation: Generate corresponding electromagnetic shielding performance data for each sample data, including shielding efficiency, energy transmission efficiency and material cost, etc.;

[0076] Step 1.6, data set division: divide the sample data into training set, validation set and test set in a ratio of 7:2:1.

[0077] In this embodiment, the shielding structure design parameters include at least: relative size parameters of the coil length, width and height; relative size parameters of the magnetic core length, width and height; relative size parameters of the shielding aluminum plate length, width and thickness; the electromagnetic shielding performance data includes at least: shielding efficiency, energy transmission efficiency, and shielding material cost.

[0078] Step 2: Constructing a machine learning proxy model: Using shielding structure design parameters as independent variables and corresponding electromagnetic shielding performance data as dependent variables, construct a machine learning proxy model;

[0079] It should be noted that, according to the requirements of wireless charging electromagnetic shielding optimization design, such as nonlinearity, multi-objective and high-dimensional parameter space, the present invention selects a suitable proxy model as the core tool for optimization. The following are several commonly used proxy models and their applicable scenarios:

[0080] a) Gaussian Process Regression (GPR): GPR is suitable for small sample and high precision scenarios. It can accurately model the nonlinear relationship between design parameters and shielding performance, and provide uncertainty information of the predicted values ​​to support dynamic sampling and optimization processes.

[0081] b) Neural Network (NN): NN can process large sample data and high-dimensional parameter space, and supports the construction of multi-layer structures to capture complex nonlinear relationships. Especially in multi-objective optimization, it can directly predict multiple performance indicators such as shielding efficiency and material cost through multi-output structures.

[0082] c) Random Forest Regression (RFR): RFR is suitable for processing nonlinear and heterogeneous data, has a faster model training speed and higher interpretability, and can be used for fast modeling in medium and low dimensional parameter spaces.

[0083] d) Support Vector Regression (SVR): SVR is more suitable for data from small and medium-sized samples and can effectively handle nonlinear regression tasks, especially showing good modeling capabilities in low-dimensional parameter space.

[0084] By comprehensively analyzing the characteristics and applicable scope of the above-mentioned proxy models, the present invention gives priority to Gaussian process regression and neural networks to take into account the high-precision modeling requirements and the flexibility of multi-objective optimization in small sample scenarios. The choice of model is adjusted according to the specific application scenarios and optimization requirements of wireless charging electromagnetic shielding design to ensure the best balance between modeling efficiency and prediction performance. The input and output of the model are the basis for the construction of the proxy model. The input includes normalized shielding structure design parameters and optional environmental parameters, and the output covers key indicators such as shielding performance, energy transmission efficiency and material cost. Through multi-objective output, the model can comprehensively evaluate the impact of design parameters on performance and provide accurate prediction support for optimization. The specific selection of input and output variables in this method is as follows.

[0085] Input variables:

[0086] The standardized shielding structure design parameters include: core size (normalized length, width, height), shielding aluminum plate size (normalized length, width, thickness), coil size (normalized length, width, height). As an option, environmental parameters (such as operating frequency, coil distance, coil type) can also be added.

[0087] Output variables:

[0088] Shielding performance indicators, including: Shielding Effectiveness (SE): the attenuation of the electromagnetic field, Energy transmission efficiency: the impact of shielding design on energy transmission loss, Material cost: the amount of material required for the corresponding design, Models that support multi-objective output directly predict multiple performance indicators

[0089] Step 3: Model training: Based on the training sample database obtained in step 1, the machine learning proxy model in step 2 is trained, and the model prediction performance is verified by cross-validation and independent test sets to obtain a trained machine learning proxy model;

[0090] It should be noted that during the training process, the training set is used to fit the model, and the hyperparameters are adjusted through cross-validation, grid search, or Bayesian optimization to improve the model performance.

[0091] Finally, the prediction accuracy and generalization ability of the model were evaluated by the validation set, and the mean square error (MSE) was used to measure the deviation between the predicted value and the true value, the determination coefficient (R 2 ) Evaluate the model's ability to explain the target variable and the prediction accuracy. The model's prediction value is close to the target value, and other indicators quantify the model effect to ensure that the model can accurately reflect the characteristics of the target variable. After the model is initially trained, it is necessary to optimize and improve the model;

[0092] Model optimization and improvement mainly include three aspects: dynamic update, multi-model integration and model lightweight.

[0093] In dynamic updating, active learning technology is combined to dynamically generate new samples according to the prediction uncertainty of the proxy model, improve the training set, and improve the model's adaptability to new samples through incremental training.

[0094] In terms of multi-model integration, multiple models (such as a combination of GPR and NN) are used to fuse the prediction results through weighted averaging or voting mechanisms to improve the overall prediction performance.

[0095] To meet the needs of real-time prediction, model lightweighting reduces the computational complexity by pruning or quantizing the neural network model, and introduces sparse technology (such as induced point sparse approximation) to the Gaussian process regression model to reduce computing resource consumption.

[0096] After the model is optimized and improved, the model is verified by verifying the performance of the proxy model on the test set and comparing the results of the proxy model with those of the traditional simulation method, such as Figure 3 As shown, the prediction accuracy can meet the design requirements.

[0097] Step 4: Construct a multi-objective optimization function. Calculate the electromagnetic shielding performance data based on the machine learning agent model trained in step 3, and construct a multi-objective optimization function in combination with the constraints of the actual project.

[0098] Preferably, in this embodiment, step 4 specifically includes the following steps:

[0099] Step 4.1, determine the optimization objectives and constraints;

[0100] The optimization objectives include: ① Maximization of shielding efficiency, the specific formula is:

[0101]

[0102] Where SE is the shielding efficiency, H 无屏蔽 is the electric field strength without shielding structure, H 有屏蔽 is the electric field strength when there is a shielding structure;

[0103] ②Maximize energy transmission efficiency, which can be expressed as:

[0104]

[0105] Where η is the energy transfer efficiency, P 输入 is the system input power; P 输出 is the effective output power of the receiving coil;

[0106] ③ Minimize material cost, the calculation formula is:

[0107] C=ρ 铝 ·V 铝 ·P 铝 +ρ 磁芯 ·V 磁芯 ·P 磁芯 ;

[0108] Where C is the material cost, ρ 铝 V is the density of the shielding aluminum plate material (in kg / m3); 铝 is the volume of the shielding aluminum plate (in m3); P 铝 is the unit price of the shielding aluminum plate material (in yuan / kg);

[0109] ρ 磁芯 is the density of the core material (in kg / m3); V 磁芯 is the volume of the core (in m3);

[0110] P 磁芯 is the unit price of the core material (in yuan / kg);

[0111] Constraints include: shielding efficiency constraints. The shielding design must meet the electromagnetic radiation limit standards (such as ICNIRP standards). According to the limit standards, the minimum shielding performance index at different frequencies is obtained. The shielding efficiency must meet the minimum standard: SE ≥ SE min , the degree of violation is expressed as: Penalty SE =max(0,SE min -SE); Energy transmission efficiency constraint, energy transmission efficiency must meet the minimum standard: η≥ηmin, and the degree of violation is expressed as: Penalty η=max(0,η min -η); Geometric dimensions and process constraints. The geometric dimensions must meet the design space and processing technology requirements. The degree of violation is expressed as:

[0112]

[0113] Step 4.2, construct a multi-objective optimization function;

[0114] The objective function adopts a weighted form to transform multiple objectives into a single objective for optimization. The objective function is defined as:

[0115] F=w1·f1+w2·f2+w3·f3+λ·Penalty;

[0116] Where: F is the comprehensive optimization objective function; f1, f2, f3 are the shielding efficiency objective function, energy transmission efficiency objective function and material cost objective function respectively, f1 = -SE, f2 = -η, f3 = C; w1, w2, w3 are the weight coefficients of shielding efficiency, energy transmission efficiency and material cost respectively, satisfying w1 + w2 + w3 = 1; λ is the penalty function coefficient, which is used to adjust the influence of violation of constraint conditions on the objective function; Penalty is the penalty function term of constraint conditions, which is used to indicate the degree of violation of constraint conditions.

[0117] Penalty=Penalty SE +Penalty η +Penalty 尺寸 . By combining multi-objective functions with constraints, the shielding efficiency, transmission efficiency and material cost are comprehensively optimized to achieve global design balance; penalty items of constraint conditions are dynamically introduced to effectively avoid designs that do not meet physical or process requirements; weights and penalty function coefficients are adjustable to adapt to different application scenarios and optimization requirements;

[0118] In this embodiment, the weights of shielding efficiency, energy transmission efficiency and material cost are set as follows:

[0119] Shielding efficiency weight w1 = 0.5

[0120] Energy transfer efficiency weight w2 = 0.3

[0121] Material cost weight w3 = 0.2

[0122] The penalty function coefficient λ = 1000. Then the comprehensive objective function is:

[0123] F=0.5·(-SE)+0.3·(-η)+0.2·C+1000·(Penalty SE +Penalty η +Penalty尺寸 );

[0124] Step 5: Optimization solution and verification;

[0125] Step 5.1, optimization solution; after normalizing the shielding structure design parameters to be optimized, input the machine learning agent model in step 3, and based on the multi-objective optimization function in step 4, adopt a multi-objective optimization algorithm. As a preferred embodiment, the present embodiment is as follows Figure 4 As shown, a multi-objective optimization algorithm such as a genetic algorithm or a particle swarm algorithm is used to search for a Pareto optimal solution set, and the corresponding optimal shielding structure design parameters are determined from the Pareto optimal solution set; as a preferred method, in step 5.1, the actual effect of the optimization scheme is verified by combining simulation analysis with physical testing;

[0126] Step 5.2, optimization verification and adjustment: verify the actual effect of the optimization scheme. If the actual effect of the optimization scheme fails to meet the design requirements, return to the optimization process, adjust the relevant design parameters, and optimize again based on the optimization methods of steps 3 to 5.1 until the simulation and test results meet the expected goals. Figure 5 This is a diagram showing the optimization effect of the magnetic field shielding structure of the present invention.

[0127] like Figure 6 As shown, the present invention constructs an efficient machine learning agent model to quickly predict shielding performance and greatly improve design efficiency; adopts a multi-objective optimization algorithm to comprehensively consider shielding efficiency, energy transmission efficiency and material cost to meet diverse practical engineering needs; introduces a closed-loop optimization mechanism of simulation verification and dynamic adjustment to ensure that the final design achieves the expected goals; significantly reduces the number of experiments, shortens the design cycle, reduces development costs, and at the same time improves shielding performance, transmission efficiency and economy.

[0128] The above are only preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The substitution may be a substitution of a part of the structure, device, method step, or a complete technical solution. The equivalent replacement or change according to the technical solution and the inventive concept of the present invention shall be covered within the protection scope of the present invention.

Claims

1. A wireless charging electromagnetic shielding optimization design method based on machine learning, characterized in that: The following steps are involved: Step 1: Data preparation; Collect shielding structure design parameters, obtain corresponding electromagnetic shielding performance data through simulation calculation and experimental testing, and build a training sample database; Step 2: Constructing a machine learning proxy model: Using shielding structure design parameters as independent variables and corresponding electromagnetic shielding performance data as dependent variables, construct a machine learning proxy model; Step 3: Model training: Based on the training sample database obtained in step 1, the machine learning proxy model in step 2 is trained, and the model prediction performance is verified by cross-validation and independent test sets to obtain a trained machine learning proxy model; Step 4: Construct a multi-objective optimization function. Calculate the electromagnetic shielding performance data based on the machine learning agent model trained in step 3, and construct a multi-objective optimization function in combination with the constraints of the actual project. Step 5: Optimization solution and verification; Step 5.1, optimization solution; After normalizing the shielding structure design parameters to be optimized, the machine learning agent model in step 3 is input, and based on the multi-objective optimization function in step 4, a multi-objective optimization algorithm is used to search for the Pareto optimal solution set, and the corresponding optimal shielding structure design parameters are determined from the Pareto optimal solution set; Step 5.2, optimization verification and adjustment; verify the actual effect of the optimization scheme. If the actual effect of the optimization scheme fails to meet the design requirements, return to the optimization process, adjust the relevant design parameters, and optimize again based on the optimization methods of steps 3 to 5.1 until the simulation and test results meet the expected goals.

2. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 1, characterized in that: The step 1 specifically includes the following steps: Step 1.1, obtain shielding structure design parameters, and obtain corresponding electromagnetic shielding performance data through simulation calculation and experimental test; Step 1.2: Improve data quality by optimizing sample distribution through Latin hypercube sampling; Step 1.3, data cleaning: check the outliers or noise in the sample data and remove the data points with large simulation errors; Step 1.4, normalization processing: normalize all sample data; Step 1.5, data annotation: generating corresponding electromagnetic shielding performance data for each sample data; Step 1.6, data set division: divide the sample data into training set, validation set and test set in a ratio of 7:2:

1.

3. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 1 or 2, characterized in that: The shielding structure design parameters include at least: relative size parameters of coil length, width and height; relative size parameters of magnetic core length, width and height; relative size parameters of shielding aluminum plate length, width and thickness; the electromagnetic shielding performance data include at least: shielding efficiency, energy transmission efficiency and shielding material cost.

4. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 1, characterized in that: In step 2, Gaussian process regression and neural network are used to construct a machine learning agent model.

5. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 1, characterized in that: The step 4 specifically comprises the following steps: Step 4.1, determine the optimization objectives and constraints; The optimization objectives include: ① Maximization of shielding efficiency, the specific formula is: Where SE is the shielding efficiency, H 无屏蔽 is the electric field strength without shielding structure, H 有屏蔽 is the electric field strength when there is a shielding structure; ②Maximize energy transmission efficiency, which can be expressed as: Where η is the energy transfer efficiency, P 输入 is the system input power; P 输出 is the effective output power of the receiving coil; ③ Minimize material cost, the calculation formula is: C6ρ 铝 ·V 铝 ·P 铝 +ρ 磁芯 ·V 磁芯 ·P 磁芯 100. Where C is the material cost, ρ 铝 V is the density of the shielding aluminum plate material; 铝 is the volume of the shielding aluminum plate; P 铝 is the unit price of the shielding aluminum plate material; ρ 磁芯 is the density of the core material; V 磁芯 is the volume of the magnetic core; P 磁芯 is the unit price of the core material; Constraints include: Shielding efficiency constraint, shielding efficiency must meet the minimum standard: SE ≥ SE min , the degree of violation is expressed as: Penalty SE =max(0,SE min -SE); Energy transmission efficiency constraint, energy transmission efficiency must meet the minimum standard: η≥ηmin, and the degree of violation is expressed as: Penalty η =max(0,η min -η); Geometric dimensions and process constraints. The geometric dimensions must meet the design space and processing technology requirements. The degree of violation is expressed as: Step 4.2, construct a multi-objective optimization function; The objective function adopts a weighted form to transform multiple objectives into a single objective for optimization. The objective function is defined as: F=w1·f1+w2·f2+w3·f3+λ·Penalty; Where: F is the comprehensive optimization objective function; f1, f2, f3 are shielding efficiency objective function, energy transmission efficiency objective function and material cost objective function, respectively, f1 = -SE, f2 = -η, f3 = C; w1, w2, w3 are the weight coefficients of shielding efficiency, energy transmission efficiency and material cost, respectively, satisfying w1 + w2 + w3 = 1; λ is the penalty function coefficient, which is used to adjust the degree of influence of violation of constraint conditions on the objective function; Penalty is the penalty function term of constraint conditions, which is used to indicate the degree of violation of constraint conditions, Penalty = Penalty SE +Penalty η +Penalty 尺寸 .

6. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 2, characterized in that: The normalization process in step 1.4 is based on the coil size, and the core size and the shielding aluminum plate size are standardized as relative parameters.

7. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 1, characterized in that: In step 5.1, the actual effect of the optimization scheme is verified by combining simulation analysis and physical testing.

8. The wireless charging electromagnetic shielding optimization design method based on machine learning as claimed in claim 1, characterized in that: The multi-objective optimization algorithm is a genetic algorithm or a particle swarm algorithm.

Images