Electric vehicle wireless charging system shielding structure and optimization method thereof

Abstract

The invention relates to the technical field of wireless electric energy transmission, and particularly discloses an electric vehicle wireless charging system shielding structure and an optimization method thereof.The shielding structure is a hollow aluminum plate arranged around a transmitting end, and the optimization method comprises the steps that a simulation model of an EV-WPT system with the shielding structure is established, acquiring data pairs of multiple groups of input variables and output variables to form a training set; on the basis of the training set, a Gaussian process regression method is adopted, and an agent model with optimization variables as input and the magnetic induction intensity and the transmission efficiency as output is established; taking the agent model as an objective function of a multi-objective optimization algorithm, and performing optimization calculation in a value range of an optimization variable to obtain an optimization variable value; the hollow aluminum plate is used as a passive shielding structure, when energy is transmitted between a transmitting end and a receiving end of the system, the magnetic induction intensity of a leakage magnetic field around the system can be weakened, and the optimization calculation efficiency of the passive shielding structure can be remarkably improved by taking the proxy model as a target function.

Description

Technical Field

[0001] This invention relates to the field of wireless power transmission technology, specifically a shielding structure for a wireless charging system for electric vehicles and its optimization method. Background Technology

[0002] Wireless Power Transfer (WPT) technology, primarily based on the principle of electromagnetic radiation coupling, utilizes an open mechanical structure to transmit energy, avoiding direct contact between the user and the charging device's mechanical structure. It offers advantages such as safety, convenience, and space saving, and is therefore widely used in industrial production, transportation, and biomedical fields. Against the backdrop of the electrification and intelligentization of the automotive industry, wireless power transfer technology has broad development prospects in the field of electric vehicles (EVs). However, the open energy transfer structure inevitably generates leakage magnetic fields around the WPT system. This not only affects energy transfer efficiency but may also lead to serious electromagnetic compatibility issues, impacting the normal operation of surrounding electrical and electronic equipment and even negatively affecting human electromagnetic safety. These problems severely limit the application and development of WPT technology; therefore, leakage magnetic field shielding technology for WPT systems is of significant research importance.

[0003] Existing shielding methods include passive shielding, active shielding, and resonant reactance shielding. Active shielding typically involves adding an extra reverse shielding coil around the WPT system to weaken the leakage magnetic field by generating an electromagnetic field in the opposite direction to the original coil. However, active shielding requires additional coils and more complex control algorithms to weaken the WPT system's leakage magnetic field, adding to the system's cost and size, and may even render it unsuitable for the original application scenario. Resonant reactance shielding also requires an extra coil as the main structure for shielding the WPT system's leakage magnetic field. Its principle is based on Faraday's law of electromagnetic induction; when the electromagnetic energy of the leakage magnetic field passes through the shielding coil, it induces a current in the shielding coil, generating an electromagnetic field opposite to the leakage magnetic field. However, compared to active shielding, resonant reactance shielding does not require an additional current source. However, resonant reactance shielding only provides shielding in a localized area and requires multiple units to weaken the leakage magnetic field around the WPT system, increasing system cost. Passive shielding can be mainly divided into magnetic shielding using high-permeability materials and conductive shielding using metallic materials. Some researchers have also used metamaterials as shielding structures for WPT systems; however, metamaterial methods only work at specific frequencies, and their shielding effectiveness deteriorates rapidly with changes in magnetic field frequency. Shielding utilizes high-permeability materials (such as ferrite) placed at both ends of the magnetic coupling mechanism of the WPT system to attract the surrounding electromagnetic energy to the vicinity of the coil, thereby increasing the self-inductance and mutual inductance of the coil and enhancing energy transmission to weaken the leakage magnetic field around the WPT system. Conductive shielding, on the other hand, is based on the eddy current effect and utilizes the reverse electromagnetic field induced when the electromagnetic field passes through a metal material (such as aluminum or copper) to achieve the shielding effect.

[0004] Compared to other shielding methods, passive shielding offers better shielding performance and lower manufacturing costs; therefore, this invention chooses passive shielding as the research object. However, passive shielding results in a larger size and mass, which affects transmission efficiency and makes the WPT system very bulky.

[0005] In existing electromagnetic shielding methods for WPT systems, simulation software is typically used to establish a system simulation model. Multi-objective optimization algorithms are then used to perform optimization calculations with the simulation model as the objective function. However, due to the large number of grids in the simulation model, the single calculation time of the model is long. Furthermore, multi-objective optimization algorithms require multiple populations to perform iterative calculations and call the simulation model for a large amount of computation. Therefore, this method has the significant drawback of high optimization computation cost. Summary of the Invention

[0006] The purpose of this invention is to provide a shielding structure for a wireless charging system for electric vehicles and its optimization method, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A shielding structure for a wireless charging system for electric vehicles, wherein the shielding structure is a hollow aluminum plate disposed around the transmitter.

[0009] The thickness of the hollow aluminum plate is consistent with the inherent thickness of the transmitter aluminum plate; the structural parameters of the hollow aluminum plate include: the length l and width h of the hollow part, and the extension length k and the extension length j of the aluminum plate part in the length and width directions.

[0010] As a further embodiment of the present invention, the structural parameters l, h, k, and j are determined by an optimization method based on machine learning.

[0011] The present invention also provides an optimization method for the shielding structure of a wireless charging system for electric vehicles, the optimization method comprising:

[0012] The geometric parameters to be optimized of the shielding structure are determined as optimization variables, and the optimization variables include at least the length, width, length extension, and width extension of the hollow part;

[0013] A simulation model of an EV-WPT system with a shielded structure is established. Within the preset value range of the optimization variables, the simulation model of the EV-WPT system with a shielded structure is run to obtain multiple sets of data pairs of input and output variables to form a training set. The input variables are different combinations of the optimization variables, and the output variables include the magnetic induction intensity and system transmission efficiency at specified observation points around the EV-WPT system.

[0014] Based on the training set, a surrogate model is established using the Gaussian process regression method, with the optimization variables as input and the magnetic induction intensity and transmission efficiency as output.

[0015] The surrogate model is used as the objective function of the multi-objective optimization algorithm, and optimization calculations are performed within the range of values ​​of the optimization variables to obtain the values ​​of the optimization variables.

[0016] As a further embodiment of the present invention, the designated observation point is located at a predetermined horizontal position around the EV-WPT system.

[0017] As a further embodiment of the present invention, the Gaussian process regression method employs the Matérn3 / 2 kernel function.

[0018] As a further aspect of the present invention, the hyperparameters in the Gaussian process regression model are determined by maximizing the log-marginal likelihood function of the training samples.

[0019] As a further embodiment of the present invention, the multi-objective optimization algorithm is a non-dominated sorting genetic algorithm.

[0020] Compared with the prior art, the beneficial effects of the present invention are: it proposes a hollow aluminum plate as a passive shielding structure for the EV-WPT system, which can ensure that the transmission performance of the system is not affected to the greatest extent when transmitting energy between the system's transmitter and receiver, while weakening the magnetic induction intensity of the leakage magnetic field around the system.

[0021] The proposed hollow aluminum plate structure has the advantages of low material cost, low processing difficulty, small space occupation, and flexible and convenient placement.

[0022] This paper proposes to use Gaussian process regression, a machine learning method, to establish a surrogate model for passive shielding structures. This method requires only a small training set to establish a high-accuracy surrogate model. Furthermore, compared to existing multi-objective optimization algorithms that directly use the simulation model as the objective function, using the surrogate model as the objective function can significantly improve the optimization computational efficiency of passive shielding structures. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention.

[0024] Figure 1 The simulation model diagram of the EV-WPT system provided in the embodiment of the present invention is shown.

[0025] Figure 2 This is a schematic diagram of the observation points provided in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of a passive shielding structure provided in an embodiment of the present invention.

[0027] Figure 4 A schematic diagram of the passive shielding structure parameters provided in an embodiment of the present invention.

[0028] Figure 5 A flowchart illustrating the optimization method provided in an embodiment of the present invention.

[0029] Figure 6 A comparison chart of magnetic induction intensity data at point A1 between the proxy model and the simulation model provided in this embodiment of the invention.

[0030] Figure 7 A comparison chart of magnetic induction intensity data at point A2 between the proxy model and the simulation model provided in this embodiment of the invention.

[0031] Figure 8 A comparison chart of transmission efficiency data between the proxy model and the simulation model provided in this embodiment of the invention. Detailed Implementation

[0032] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0033] In this embodiment of the invention, a shielding structure for a wireless charging system for electric vehicles is provided, wherein the shielding structure is a hollow aluminum plate disposed around the transmitting end;

[0034] The thickness of the hollow aluminum plate is consistent with the inherent thickness of the transmitter aluminum plate; the structural parameters of the hollow aluminum plate include: the length l and width h of the hollow part, and the extension length k and the extension length j of the aluminum plate part in the length and width directions.

[0035] In this embodiment, the passive shielding structure is applied to Electric Vehicles Wireless Power Transfer (EV-WPT) systems. Taking an 11kW EV-WPT system as an example, this invention utilizes the multiphysics simulation software Comsol to establish a simulation model of the EV-WPT system, such as... Figure 1 As shown.

[0036] Figure 1 The EV-WPT system operates at a frequency of 85kHz. Its transmitter consists of a transmitting coil, a transmitting ferrite core, and an aluminum plate. The transmitting coil is wound with 9 turns of double-stranded wire, each turn being 2500 turns of Litz wire. The transmitter's dimensions are 0.76m x 0.6m. The receiver consists of a receiving coil and a receiving ferrite core. The receiving coil is wound with 9 turns of double-stranded wire, each turn being 1500 turns of Litz wire. The receiver's dimensions are 0.5m x 0.5m. The mechanical air gap between the transmitter and receiver is 0.15m. Using the system's center position as the coordinate origin, two observation points are set horizontally around the system, such as... Figure 2 As shown.

[0037] The coordinates of the observation points are A1 (0.85, 0, 0.08) and A2 (0, 0.75, 0.08). The leakage magnetic field around the system is observed using magnetic flux density as the observation target. Simulation calculations show that the magnetic flux density of A1 and A2 are 23.9 μT and 28.1 μT, respectively, with a transmission efficiency of 93.2%. To further reduce the magnetic flux density around the system without affecting the transmission efficiency and while maintaining a low cost, this invention designs a hollow aluminum plate around the system's transmitting end as an additional passive shielding structure, such as... Figure 3 and Figure 4 As shown.

[0038] The hollow aluminum plate in this invention has the same thickness as the transmitting end aluminum plate. Design parameters include the length *l* and width *h* of the central control section, and the length extensions *k* and *j* of the aluminum plate section. Because the aluminum plate has a hollow structure, it achieves shielding while maximizing the system's transmission performance. Compared to existing active shielding, passive shielding, and resonant reactance shielding, the passive shielding design proposed in this invention has advantages such as low cost, simple processing, minimal impact on transmission performance, and flexible placement.

[0039] The present invention also provides an optimization method for the shielding structure of a wireless charging system for electric vehicles, the optimization method comprising:

[0040] The geometric parameters to be optimized of the shielding structure are determined as optimization variables, and the optimization variables include at least the length, width, length extension, and width extension of the hollow part;

[0041] A simulation model of an EV-WPT system with a shielded structure is established. Within the preset value range of the optimization variables, the simulation model of the EV-WPT system with a shielded structure is run to obtain multiple sets of data pairs of input and output variables to form a training set. The input variables are different combinations of the optimization variables, and the output variables include the magnetic induction intensity and system transmission efficiency at specified observation points around the EV-WPT system.

[0042] Based on the training set, a surrogate model is established using the Gaussian process regression method, with the optimization variables as input and the magnetic induction intensity and transmission efficiency as output.

[0043] The surrogate model is used as the objective function of the multi-objective optimization algorithm, and optimization calculations are performed within the range of values ​​of the optimization variables to obtain the values ​​of the optimization variables.

[0044] like Figure 5 As shown, in this embodiment, in order to clarify the optimal structural parameters of l, h, k and j, this invention addresses the problem of high time cost in existing shielding structure design methods by proposing to establish a proxy model of magnetic induction intensity at observation points of a passive shielding structure using the Gaussian process regression method in machine learning, and using the proxy model as the objective function of a multi-objective optimization algorithm.

[0045] Gaussian Process Regression (GPR) is a machine learning method that can rapidly construct surrogate models based on small-scale experimental designs. Taking advantage of this, this invention uses GPR to construct a surrogate model for optimizing the passive shielding structure of the EV-WPT system. The input variable of the surrogate model is the geometric parameters of the passive shielding structure, with a dimension of D. In this invention, the passive shielding structure has four parameters, hence the input variable dimension D is 4. The output variable is the magnetic flux density B at the observation point.A1 B A2 With respect to the transmission efficiency η, the calculation process of the finite element simulation software is regarded as a black box function f(x). From the perspective of function space, the Gaussian process can be expressed as:

[0046] ;

[0047] Where θ represents the hyperparameters of the covariance function, m(x) and Let f(x) be the mean function and covariance function of the stochastic process f(x), respectively.

[0048] The learning objectives of GPR are:

[0049] ;

[0050] in and The noise is the estimated GPR, and Here, n is the number of training sets, and the optimization objective of this invention is the magnetic flux density B of the WPT system. A1 B A2 For ease of calculation and interpretation, the transmission efficiency η will be expressed in terms of magnetic induction intensity B. A1 Let's take an example. It's worth noting that, to ensure consistency when training the proxy model, the input variables in the training set are from the same set, while the output variables are from their respective corresponding sets.

[0051] To simplify the calculation process, let there be n training sample points. Composition of B A1 The a priori form is Let the latent function constructed from m test sample points be... Then B A1 and The joint prior distribution is:

[0052] ;

[0053] in, , , These represent the training samples, the test samples, and the covariance matrix between the training and test samples, respectively. Based on Bayesian theory, the mean of the prediction distribution can be derived. and variance They are respectively:

[0054] ;

[0055] ;

[0056] in It is possible to obtain an approximate prediction of the magnetic flux density. The uncertainty of the prediction can be obtained. In the calculation of GPR, the selection of kernel function and hyperparameters is a key step. The following section introduces the solution of kernel function and hyperparameters.

[0057] The principle behind kernel functions is that, when the kernel function satisfies the Mercer condition, it maps points in a low-dimensional space to a high-dimensional feature space through vector inner products, effectively avoiding the problems of the "curse of dimensionality" and overfitting. Commonly used kernel functions include:

[0058] Square kernel (i.e., SE covariance):

[0059] ;

[0060] Matérn3 / 2 core:

[0061] ;

[0062] To meet the requirements for solution accuracy, the Matérn 3 / 2 kernel was used for regression analysis.

[0063] In the GPR kernel function, various parameters and noise can be changed; these parameters are collectively called hyperparameters. The optimal hyperparameters can be obtained by calculating the maximum value of the log-marginal likelihood function of the training samples.

[0064] ;

[0065] Based on the above, surrogate models for optimizing the passive shielding structure parameters can be established. These surrogate models will then be used as the objective function of a multi-objective optimization algorithm for calculation. Optionally, this invention employs the Non-Dominated Sorting Genetic Algorithm-II (NSGA-II) to optimize the passive shielding structure parameters, with l, h, k, and j as optimization variables. Since NSGA-II is a relatively mature algorithm in the field of optimization calculation, it will not be described in detail. Through the Gaussian process method and the NSGA-II algorithm, the passive shielding structure parameters can be obtained. The passive shielding structure design process proposed in this invention is as follows: Figure 5 As shown.

[0066] Simulation example: Figure 1 Taking the EV-WPT system as an example, a passive shielding structure is established according to the method proposed in this invention. The optimization ranges of the passive shielding structure parameters l, h, k, and j are shown in Table 1:

[0067] Table 1. Optimization range of parameters l, h, k, and j

[0068]

[0069] A simulation model of the EV-WPT system with a passive shielding structure is run within the optimization range, and the corresponding input and output variables are compared with the magnetic induction intensity B at the observation point. A1 B A2 The transmission efficiency η forms a training set, requiring 400 sets in this invention. A proxy model of the passive shielding structure is established using the GPR method. To demonstrate the accuracy of the proxy model, 20 sets of input variable data are selected, and the calculation results of the output variables in the proxy model and simulation software are compared. Figures 6 to 8 As shown.

[0070] observe Figures 6 to 8 As can be seen, the computational accuracy of the surrogate model is basically consistent with that of the simulation model, and it can effectively represent the computational results of the simulation model. It can be used for multi-objective optimization design of passive shielding structures. The NSGA-II algorithm was used to perform optimization calculations within the parameter optimization range. The population size was set to 50, and the number of iterations was 500. The calculated passive shielding structure parameter results were presented in the form of a Pareto front, and the results required for this invention were selected from them, as shown in Table 2.

[0071] Table 2 Calculation results of passive shielding structure parameters optimization

[0072]

[0073] Substituting the optimization results of the passive shielding structure into the simulation model, the magnetic induction intensities at observation points A1 and A2 are calculated to be 19.8 μT and 22.9 μT, respectively, with the leakage magnetic field strength reduced by 20.9% and 18.5%. Comparing the computational costs of the surrogate model and existing methods, when using the surrogate model as the objective function, 400 training sets are required, meaning the simulation model needs to be calculated 400 times. The surrogate model uses an analytical form as the objective function of the multi-objective optimization algorithm, thus requiring lower time costs when performing multi-objective optimization calculations. In contrast, when using the simulation model as the objective function, with a population size of 50 and 500 iterations, it is equivalent to running the simulation model 25,000 times. Therefore, using the passive shielding structure surrogate model for multi-objective optimization calculations can save approximately 98.4% of the computational cost, significantly increasing the optimization computational efficiency of the passive shielding structure in the EV-WPT system.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An electric vehicle wireless charging system shielding structure, characterized by, The shielding structure is a hollow aluminum plate arranged around the transmitting end; The thickness of the hollow aluminum plate is consistent with the inherent thickness of the aluminum plate of the transmitting end; The structural parameters of the hollow aluminum plate include the length l and the width h of the hollow part, and the extension length k in the length direction and the extension length j in the width direction of the aluminum plate part.

2. The shielding structure for a wireless charging system of an electric vehicle according to claim 1, wherein The structural parameters l, h, k and j are determined by an optimization method based on machine learning.

3. An optimization method of a shielding structure of an electric vehicle wireless charging system, the optimization method being used for the shielding structure of the electric vehicle wireless charging system according to claim 1 or 2, characterized in that, The optimization method includes: determining the to-be-optimized geometric parameters of the shielding structure as optimization variables, the optimization variables at least including the length, the width, the length extension and the width extension of the hollow part; establishing a simulation model of an EV-WPT system with a shielding structure, running the simulation model of the EV-WPT system with the shielding structure in a preset value range of the optimization variables, obtaining a plurality of groups of data pairs of input variables and output variables, and constituting a training set, wherein the input variables are different value combinations of the optimization variables, and the output variables include the magnetic induction intensity of a designated observation point around the EV-WPT system and the transmission efficiency of the system; based on the training set, adopting a Gaussian process regression method to establish a proxy model with the optimization variables as inputs and the magnetic induction intensity and the transmission efficiency as outputs; taking the proxy model as a target function of a multi-objective optimization algorithm, and performing optimization calculation in the value range of the optimization variables to obtain the values of the optimization variables.

4. The method of claim 3, wherein, The designated observation point is located at a preset horizontal position around the EV-WPT system.

5. The method of claim 3, wherein the method is characterized by: The Gaussian process regression method adopts a Matérn3 / 2 kernel function.

6. The method of claim 3, wherein the method is characterized by: The hyperparameters in the Gaussian process regression model are determined by maximizing the log marginal likelihood function of the training samples.

7. The method of claim 3, wherein the method further comprises: The multi-objective optimization algorithm is a non-dominated sorting genetic algorithm.

Images