Wireless power transmission system coil parameter prediction method, device and medium

Abstract

The invention discloses a wireless electric energy transmission system coil parameter prediction method and device and a medium, and the method comprises the steps: constructing a wireless electric energy transmission coil simulation model with magnetic shielding, and constructing a first data set and a second data set through simulation; calculating parameters of the wireless electric energy transmission coil under the condition of the shielding magnetic core, and obtaining an analytic calculation data set as a third data set; dividing the first data set and the second data set into a training set and a verification set respectively, and adopting the third data set as supplement of the training set to ensure balance of data sources; setting different simulation examples, and obtaining a test data set; constructing and training a twin neural network; two sub-networks of the twin neural network share one part of parameters, and the two parts of networks are jointly trained; and performing coil parameter prediction by using a prediction model combination mode to obtain a prediction result. Through the twinning neural network model, the method achieves the prediction of the parameters of the wireless power transmission coil, and meets the requirements of complex design parameter optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless power transmission, and particularly to a method, device, and medium for predicting coil parameters of a wireless power transmission system. Background Art

[0002] Wireless power transfer (WPT) systems are widely used in multiple fields, especially in the wireless charging technology of electric vehicles. The core component of an IPT (inductive power transfer) system is a magnetic coupler, which consists of a transmitting coil and a receiving coil. They need to resonate at the operating frequency to effectively transfer power. However, the design of the magnetic coupler is very complex. If the design is improper, it may lead to a decrease in efficiency and an increase in leakage magnetic flux. To improve the coupling efficiency and the transmission efficiency of electromagnetic energy, it is crucial to design a suitable geometric structure of the magnetic coupler. These designs are usually restricted by the application scenarios. For example, parameters such as the inner and outer diameters of the coil, the selection of materials, and the coil diameter need to balance electrical constraints and space limitations.

[0003] Traditional coil design methods, such as traditional analytical methods, usually assume that the coil geometry is relatively simple, such as circular or rectangular. However, in practical applications, the coil may include magnetic shielding, magnetic cores, complex curves, or multi-layer material structures. These complex geometries and material distributions will significantly affect the magnetic field distribution and energy transfer efficiency, and traditional analytical methods cannot accurately handle these situations. Especially in the presence of magnetic shielding materials and magnetic cores, the magnetic field will be distorted, affecting the coupling coefficient and other key parameters. Due to the complexity of the above factors, key parameters such as the self-inductance, mutual inductance, and coupling coefficient of the magnetic coupler cannot be accurately described by a closed-form analytical solution in the case of magnetic shielding or magnetic cores. Traditional analytical methods usually can only provide simplified approximate solutions, which limits their application in complex systems. Under the influence of the non-uniformity of the magnetic field domain and complex geometries, it becomes extremely difficult to solve the analytical equations of these parameters, and thus cannot meet the requirements of high-precision design. For example, the finite element method (FEM), although it can provide accurate results, is time-consuming and inefficient in calculation because it needs to simulate the complex geometries and magnetic field distributions in detail. Especially in scenarios with multiple design parameters and high complexity, the calculation time of the finite element method may be very long, which makes the design process extremely slow. Therefore, traditional analytical calculation methods and the finite element method are not suitable for dealing with the design of wireless power transmission systems in complex environments.

[0004] In recent years, researchers have proposed data-driven methods based on machine learning to accelerate the design process. By training a neural network using a finite element simulation dataset, a model of a magnetic coupler with various coil structures can be established. This method can quickly predict the performance of the magnetic coupler, such as self-inductance, mutual inductance, coupling coefficient, etc., and then optimize the coil design. Although this method can significantly shorten the design cycle and improve the design efficiency compared with the traditional design method based on physical equations, the pure data-driven method is highly dependent on the dataset and requires a large amount of high-quality data for training.

[0005] In summary, the analytical calculation is not sensitive enough to some system parameters and cannot accurately reflect the characteristics of system parameters; the finite element simulation is time-consuming and difficult to meet the high-efficiency requirements of multi-design parameter optimization, and the dataset acquisition cost is high when the design parameter range is large; the traditional machine learning method uses a neural network model to predict parameters through data-driven, but a single neural network method often has a high dependence on a large-scale high-quality dataset and is difficult to meet the accuracy requirements when the data is limited. Summary of the Invention

[0006] To solve at least one of the technical problems existing in the prior art to a certain extent, the object of the present invention is to provide a method, device and medium for predicting coil parameters of a wireless power transmission system based on a twin neural network.

[0007] The first technical solution adopted by the present invention is as follows:

[0008] A method for predicting coil parameters of a wireless power transmission system includes the following steps:

[0009] Construct a simulation model of a wireless power transmission coil with magnetic shielding, and obtain a simulation dataset as the first dataset by performing parameter scans on different coil size parameters; obtain a simulation dataset as the second dataset by performing parameter scans at different positions and under different core parameters;

[0010] Calculate the coil parameters of the wireless power transmission coil with a shielded core, and obtain an analytical calculation dataset as the third dataset;

[0011] Divide the first dataset and the second dataset into training sets and validation sets respectively, use the third dataset as a supplement to the training set to ensure the balance of data sources; set different simulation cases to obtain a test dataset for evaluating the performance of the model;

[0012] Construct and train a twin neural network: the first neural network is used to predict the electromagnetic parameters based on the coil size, and the second neural network is used to predict the difference in electromagnetic parameters based on the changes in position and core parameters; train the first neural network according to the first training set and the third dataset, and train the second neural network according to the second training set and the third dataset;

[0013] Two sub - networks of the twin neural network share some parameters, and the two parts of the network are jointly trained through a loss function; a weighted loss function is adopted to balance the training effects of the two sub - networks and adjust the weights of the analytical calculation dataset and the simulation dataset.

[0014] The coil parameter prediction is carried out by using the combined prediction model to obtain the final prediction result; the model prediction result is compared with the actual parameters of the test dataset, and the error is calculated to evaluate the model performance.

[0015] Further, calculating the wireless power transfer coil parameters in the case of a shielded magnetic core includes:

[0016] Combining the mirror method and the planar coil parameter calculation method to calculate the wireless power transfer coil parameters in the case of a shielded magnetic core.

[0017] Further, the process of calculating the coil parameters by the mirror method is as follows:

[0018] 1) Calculate the single - turn coil parameters without a magnetic core:

[0019] According to the Neumann equation, the mutual inductance M between concentric single - turn coils can be expressed as:

[0020]

[0021]

[0022] By introducing the parameter γ:

[0023]

[0024] The mutual inductance of concentric single - turn coils is obtained:

[0025]

[0026] For the single - turn coil parameters in the case of offset, there are:

[0027]

[0028] Introduce the parameters γ, α, β, and δ:

[0029]

[0030] According to the Neumann equation, the mutual inductance of the single - turn coil in the case of offset is:

[0031]

[0032] Among them, a is the radius of the transmitting coil, b is the radius of the receiving coil, d is the transmission distance between the transmitting coil and the receiving coil, x is the lateral offset between the transmitting coil and the receiving coil; μ 0 is the permeability of free space; is the angle between the current element of the transmitting coil and the initial position, is the angle between the current element of the receiving coil and the initial position; R 12 is the distance between the current elements in the transmitting coil and the receiving coil; l 1 is the current element of the transmitting coil, l 2 is the current element of the receiving coil;

[0033] 2) Calculate the parameters of the multi-turn coil without a magnetic core

[0034] The radius of each turn of the transmitting coil is expressed as:

[0035] R i = R oi + N i × (a i + R wi )

[0036] The radius of each turn of the receiving coil is expressed as:

[0037] R j = R oj + N j × (a j + R wj )

[0038] In the formula, R oi is the inner diameter of the transmitting coil, N i is the number of turns of the transmitting coil, a i is the wire diameter of the transmitting coil, R wi is the turn spacing between each turn of the transmitting coil;

[0039] Similarly, R oj is the inner diameter of the receiving coil, N j is the number of turns of the receiving coil, a j is the wire diameter of the receiving coil, R wj is the turn spacing between each turn of the receiving coil;

[0040] The mutual inductance between the coils with different numbers of turns is expressed as:

[0041]

[0042] Calculate the mutual inductance between each turn of the transmitting coil and the receiving coil, and add them up to obtain the mutual inductance parameter of the multi-turn coil without a magnetic core:

[0043]

[0044] Wherein, N i is the number of turns of the transmitting coil, and N j is the number of turns of the receiving coil;

[0045] After adding the magnetic shielding layer, the source current will form an infinite number of sets of image currents through the alternating reflections of the two magnetic planes. They have equal amplitudes and are located on the planes of z = ±d, ±3d, ±5d... respectively; the mutual inductance parameters of the multi-turn coil with a shielded magnetic core are obtained:

[0046]

[0047] Wherein, n is the number of mirror image times;

[0048] Calculate the self-inductance parameter of the multi-turn coil with a shielded magnetic core:

[0049] Its self-inductance is expressed as:

[0050] L = L self + M self + M mirror

[0051] Wherein, L self is the self-inductance of each turn of the coil, and M self is the mutual inductance between each turn of the coil, and M mirror is the mutual inductance between the image coil generated by the image current and the source coil;

[0052] It is further expressed as:

[0053]

[0054] Wherein, is the mutual inductance between the i a th turn and the i b th turn of the coil; M ij is the mutual inductance between the image coil and the i

[0055] th turn and the j

[0056] Further, the resistance parameter of the coil is expressed as:

[0057]

[0058] Wherein, l i is the length of the i

[0059] Further, the input of the first neural network is the coil size parameters, and the output is each coil parameter; among them, the coil size parameters include the number of turns, inner diameter, outer diameter, wire diameter, and turn spacing, and each coil parameter includes mutual inductance M, self-inductance L 1 , L 2 , internal resistance R 1 , R 2 , coupling coefficient ε;

[0060] The input of the second neural network includes position parameters, different core sizes, and core thicknesses, and the output is the difference between each coil parameter and the coil in the constant case; among them, the differences include mutual inductance ΔM, self-inductance ΔL 1 , ΔL 2 , internal resistance ΔR 1 , ΔR 2 , coupling coefficient Δε.

[0061] Further, the structure of the first sub-neural network includes a fully connected layer, an activation function, and an output layer;

[0062] The fully connected layer contains 3 hidden layers, each composed of 128 neurons; the activation function uses ReLU (Rectified Linear Unit) as the activation function to improve the non-linear learning ability of the model; the output layer is 6 neurons, corresponding to the following coil parameters respectively: mutual inductance M, self-inductance L 1 , L 2 , internal resistance R 1 , R 2 , coupling coefficient ε.

[0063] Further, during the training process, the loss function of the first neural network is:

[0064]

[0065] In the formula, M i , L 1i , L 2i , R 1i , R 2i and ε i are the true values of the i-th sample, where M represents the mutual inductance parameter, L 1 represents the self-inductance parameter of the transmitting coil, L 2 represents the self-inductance parameter of the receiving coil, R 1 represents the internal resistance parameter of the transmitting coil, R 2 represents the internal resistance parameter of the receiving coil, and ε represents the coupling coefficient; and are the predicted values of this sample, and N is the total number of samples.

[0066] Further, during the training process, the loss function of the second neural network is as follows:

[0067]

[0068] In the formula, ΔM i , ΔL 1i , ΔL 2i , ΔR 1i , ΔR 2i and Δε i are the true values of the i-th sample, and are the predicted values of the sample, and n is the total number of samples.

[0069] The second technical solution adopted by the present invention is:

[0070] An electronic device, the electronic device includes a processor and a memory, and at least one instruction, at least one program, a code set or an instruction set is stored in the memory. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the above-mentioned method for predicting coil parameters of a wireless power transmission system.

[0071] The third technical solution adopted by the present invention is:

[0072] A computer-readable storage medium, in which at least one instruction, at least one program, a code set or an instruction set is stored. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the above-mentioned method for predicting coil parameters of a wireless power transmission system.

[0073] The fourth technical solution adopted by the present invention is:

[0074] A computer program product or a computer program, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above-mentioned method for predicting coil parameters of a wireless power transmission system.

[0075] The beneficial effects of the present invention are as follows: By combining analytical calculations and finite element simulation data, the present invention constructs a twin neural network model for coil size parameters, position parameters, and magnetic core parameters, overcoming the deficiencies of analytical calculations in terms of insufficient sensitivity to system parameters, long time-consuming for finite element simulations, and the dependence of a single neural network on a large-scale high-quality data set, thereby realizing efficient and high-precision prediction of wireless power transmission coil parameters and meeting the requirements of complex design parameter optimization. Brief Description of the Drawings

[0076] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following provides an introduction to the drawings of the relevant technical solutions in the embodiments of the present invention or the prior art. It should be understood that the drawings in the following introduction are only for conveniently and clearly presenting some embodiments of the technical solutions of the present invention. For those skilled in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0077] Figure 1 is a flowchart of the steps of a method for predicting coil parameters of a wireless power transmission system in an embodiment of the present invention;

[0078] Figure 2 is a schematic diagram of a single-turn coil without a magnetic core in an embodiment of the present invention;

[0079] Figure 3 is a schematic diagram of a multi-turn coil with a magnetic core in an embodiment of the present invention;

[0080] Figure 4 is a schematic diagram of calculating coil parameters by the mirror image method in an embodiment of the present invention. Detailed Description of the Embodiments

[0081] The following details the embodiments of the present invention. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention. For the step numbers in the following embodiments, they are only set for convenience of description and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0082] In the description of the present invention, it should be understood that for the orientation description, such as the orientation or positional relationship indicated by up, down, front, back, left, right, etc., is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0083] In the description of the present invention, "several" means one or more, "multiple" means more than two, "greater than", "less than", "exceeding", etc. are understood as not including the corresponding number, and "above", "below", "within", etc. are understood as including the corresponding number. If "first" and "second" are described, they are only used to distinguish technical features and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or the sequence relationship of the indicated technical features.

[0084] In the description of the present invention, unless otherwise clearly defined, terms such as "set", "install", "connect", etc. shall be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific content of the technical solution.

[0085] In view of the existing technical problems, the present invention proposes a method for predicting coil parameters of a wireless power transmission system based on a siamese neural network. This method mainly uses the idea of a siamese neural network to divide the task of predicting the parameters of a wireless power transmission coil into two subtasks, and processes them using two neural networks respectively. Specifically, the first neural network processes the parameters related to the coil size, and the second neural network processes the parameters related to the position offset, core size and thickness. Through the collaborative work of the two networks, the present invention can accurately predict the coil parameters under different conditions. The siamese neural network can improve the prediction ability and generalization ability of the model in complex scenarios by sharing weights and learning the similarities and differences between inputs.

[0086] Embodiment 1

[0087] As Figure 1 shown, this embodiment provides a method for predicting coil parameters of a wireless power transmission system based on a siamese neural network, including the following steps:

[0088] S1. Use finite element simulation software to construct a simulation model of a wireless power transmission coil with magnetic shielding. By performing parameter scans on different coil size parameters, a simulation data set of battery parameters is obtained as the first data set; by performing parameter scans at different positions and core parameters, a simulation data set of battery parameters is obtained as the second data set.

[0089] Specifically, first, a simulation model of a wireless power transfer coil with magnetic shielding is constructed using finite element simulation software. By varying different coil size parameters (such as number of turns, inner diameter, outer diameter, wire diameter, turn spacing), and adopting a parameter scanning method at no offset and a constant transmission distance, coil parameters (mutual inductance M, self-inductance L1, L2, internal resistance R1, R2, and coupling coefficient ε) in different situations are obtained, and a simulation data set is acquired, which is called the first data set. By varying different position parameters (position offsets and angular offsets in the x, y, and z directions) and different magnetic core sizes and magnetic core thicknesses, a simulation data set of battery parameters is obtained through the above parameter scanning method, which is called the second data set.

[0090] It should be noted that in addition to using finite element simulation software for simulation, other electromagnetic simulation software such as COMSOL and HFSS can also be used to construct a simulation model of the wireless power transfer coil, and different simulation methods (such as time-domain simulation, frequency-domain simulation, etc.) can be adopted to obtain an electromagnetic parameter data set.

[0091] S2. Combine the mirror image method with the planar coil parameter calculation method to calculate the parameters of the wireless power transfer coil in the case of a shielded magnetic core, and obtain an analytical calculation data set as the third data set.

[0092] In one embodiment, by combining the mirror image method with the planar coil parameter calculation method, which is an analytical calculation method, the parameters of the wireless power transfer coil in the case of a shielded magnetic core are calculated. Coil parameters (mutual inductance M, self-inductance L1, L2, internal resistance R1, R2, and coupling coefficient ε) are calculated through coil parameters (number of turns, wire diameter, initial radius, turn spacing, transmission distance, offset distance), and an analytical calculation data set is obtained, which is called the third data set.

[0093] It should be noted that when obtaining the simulation data set, it is not limited to considering basic coil size parameters such as the number of turns, inner and outer diameters, wire diameter, and turn spacing. According to actual application requirements, it can also be extended to other parameters that affect the performance of wireless power transfer, such as coil material properties and wire types.

[0094] S3. Divide the first data set and the second data set into a training set and a validation set respectively, use the third data set as a supplement to the training set to ensure the balance of the data source; set different simulation ratios to obtain a test data set for evaluating the performance of the model.

[0095] As an alternative implementation, divide the first data set into a training set and a validation set according to a ratio of 8:2, and call them the first training set and the first validation set respectively. Similarly, divide the second data set into a training set and a validation set according to a ratio of 8:2, and call them the second training set and the second validation set respectively.

[0096] Then, set multiple simulation examples with coil size parameters, position parameters, and magnetic core parameters different from those of the first dataset and the second dataset in the finite element simulation software, and obtain a test dataset for evaluating the performance of the model.

[0097] S4. Construct and train a siamese neural network: The first neural network is used to predict electromagnetic parameters based on coil size, and the second neural network is used to predict the difference in electromagnetic parameters based on changes in position and magnetic core parameters; train the first neural network with the first training set and part of the analytical dataset, and train the second neural network with the second training set and part of the analytical dataset.

[0098] Exemplarily, the input of the first neural network is coil size parameters (number of turns, inner diameter, outer diameter, wire diameter, turn spacing), and the output is (mutual inductance M, self-inductance L1, L2, internal resistance R1, R2, and coupling coefficient ε); the input of the second neural network is position parameters (position offsets and angular offsets in the x, y, and z directions), different magnetic core sizes, and magnetic core thicknesses, and the output is the difference (mutual inductance ΔM, self-inductance ΔL1, ΔL2, internal resistance ΔR1, ΔR2, and coupling coefficient Δε) between the coil parameters (mutual inductance M, self-inductance L1, L2, internal resistance R1, R2, and coupling coefficient ε) and the coil in the constant case (i.e., no offset, constant transmission distance, constant magnetic core size and thickness). Train the first sub-neural network with the first training set and part of the analytical calculation dataset (i.e., the third dataset), and train the second sub-neural network with the second training set and part of the third dataset.

[0099] It should be noted that in addition to simulation data and analytical calculation datasets, experimental data (such as actual test measurement data) can also be introduced, and data from different sources can be fused to enhance the diversity and robustness of the training set.

[0100] S5. Two sub-networks of the siamese neural network (i.e., the first neural network and the second neural network) share a part of the parameters, and jointly train the two parts of the network through a loss function; use a weighted loss function to balance the training effects of the two sub-networks and adjust the weights of the analytical calculation dataset and the simulation dataset.

[0101] The two sub-neural networks share a part of the weights to ensure that they learn similar features in the same feature space, so as to be able to make predictions for different inputs through the shared parameters and structures. Through weight sharing, the neural network can perform effective mapping and reasoning under similar physical features.

[0102] During the training process, the weight ratio of the analytical calculation dataset to the simulation dataset can be adaptively adjusted through the verification effect of the model on the validation set, ensuring that the contributions of different data sources to the model training are balanced.

[0103] Through the above-mentioned training method and weight adjustment method, a first prediction model and a second prediction model are obtained respectively. The first prediction model can predict the coil parameters of a wireless power transmission system under different coil size parameters; the second prediction model can predict the difference between the coil parameters and the initial coil state under different position parameters and core parameters.

[0104] S6. Use the combined method of the first prediction model and the second prediction model to predict the coil parameters, and obtain the final prediction result; compare the model prediction result with the actual parameters of the test data set, calculate the error, and evaluate the model performance.

[0105] As an optional implementation manner, by adding the initial value of the coil parameters predicted by the first prediction model (i.e., the first neural network) and the difference of the coil parameters predicted by the second prediction model (i.e., the second neural network), the prediction of the coil parameters under different coil size parameters, different offset conditions, and core parameters can be realized.

[0106] Finally, compare the prediction result obtained by superimposing the first prediction model and the second prediction model with the test data set, and the error between the prediction result of the prediction method proposed by the present invention and the actual parameters can be obtained.

[0107] The following is a detailed explanation in conjunction with the accompanying drawings and specific embodiments.

[0108] (1) Calculating coil parameters by the mirror image method

[0109] In this embodiment, the mirror image method is adopted as an analytical calculation method to calculate the parameters of the wireless power transmission coil, as a supplementary data set for neural network training. The mirror image method usually assumes that the core is infinite, thus simplifying the electromagnetic field distribution and the interaction between coils. Specifically, the mirror image method regards the coil as a filament and the core as a material with an infinite volume, ignoring the influence of the finite size of the core and the wire diameter parameters on the electromagnetic parameters. Although this assumption simplifies the problem theoretically, in practical applications, the core is usually finite, and the change of the wire diameter will also affect the accuracy of the electromagnetic parameters. Therefore, there are certain deviations in the coil parameters calculated by the mirror image method. Especially when calculating key electromagnetic parameters such as self-inductance, mutual inductance, and coupling coefficient, the finite size of the core will significantly affect the results.

[0110] However, the advantage of the mirror method lies in its relatively simple calculation process, which can quickly obtain an analytical data set. Therefore, this invention adopts this method as a supplement to simulation calculation. Despite certain deviations, the analytical calculation data set can make up for the shortage of the finite element simulation data set in terms of data volume and effectively improve the generalization ability of the neural network. In the twin neural network method of this invention, by combining the analytical calculation data set with the finite element simulation data set, the accuracy and training efficiency of the model can be effectively improved, especially in the case of limited data sets.

[0111] The process of calculating the coil parameters by the mirror method adopted in the embodiment of this invention is as follows:

[0112] 1) First, calculate the parameters of a single-turn coil without a magnetic core

[0113] See Figure 2 , according to the Neumann equation, the mutual inductance M between concentric single-turn coils can be expressed as:

[0114]

[0115] Where:

[0116]

[0117] By introducing the parameter γ:

[0118]

[0119] The mutual inductance of concentric single-turn coils is obtained:

[0120]

[0121] For the parameters of a single-turn coil in the offset case, there are:

[0122]

[0123] Introduce the parameters γ 、 α 、 β and δ:

[0124]

[0125] According to the Neumann equation, the mutual inductance of a single-turn coil in the offset case is:

[0126]

[0127] 2) Then, calculate the parameters of a multi-turn coil without a magnetic core

[0128] See Figure 3 , the radius of each turn of the transmitting coil can be expressed as:

[0129] Ri = R oi + N i × (a i + R wi )

[0130] The radius of each turn of the same receiving coil can be expressed as:

[0131] R j = R oj + N j × (a j + R wj )

[0132] The mutual inductance between coils with different numbers of turns can be expressed as:

[0133]

[0134] Next, calculate the mutual inductance between each turn of the transmitting coil and the receiving coil, and add them up to obtain the mutual inductance parameter of the multi-turn coil without a magnetic core:

[0135]

[0136] See Figure 4 , after adding the magnetic shielding layer, the source current is alternately reflected by two magnetic planes, forming an infinite number of sets of image currents. They have equal amplitudes and are located on the planes of z = ±d, ±3d, ±5d... respectively. In fact, with multiple specular reflections, the image currents gradually move away from the plane of the receiving coil, and the influence on the coil mutual inductance gradually decreases. Only part of the image currents are taken for approximate calculation in theoretical calculation.

[0137] Obtain the mutual inductance parameter of the multi-turn coil with a shielded magnetic core:

[0138]

[0139] Next, calculate the self-inductance parameter of the multi-turn coil with a shielded magnetic core:

[0140] Its self-inductance can be expressed as:

[0141] L = L self + M self + M mirror

[0142] where L self is the self-inductance of each turn of the coil, M self is the mutual inductance between each turn of the coil, and M mirror is the mutual inductance between the image coil generated by the image current and the source coil.

[0143] It can be further expressed as:

[0144]

[0145] L i It can be calculated by the following formula

[0146] L i =μ 0 R(i){ln[8R i / g]-2}

[0147] Where g is the geometric mean distance of the cross-sectional area of ​​the coil. For a coil with a circular cross section and a diameter of R_a, its expression is as follows:

[0148]

[0149] ( is the mutual inductance between the iath and ibth coils) can be calculated by the following formula. According to the mutual inductance formula calculated previously, at this time, d and x are both equal to 0,

[0150] The mutual inductance formula can be simplified to:

[0151]

[0152] M ij (M ij is the mutual inductance between the mirror coil and the i-th and j-th turns of the original coil) can be calculated by the following formula. According to the mutual inductance formula calculated previously, x is equal to 0 at this time, and the mutual inductance formula can be simplified to:

[0153]

[0154] By combining the above equations, the self-inductance parameters of the multi-turn coil with a shielded magnetic core can be obtained.

[0155] The resistance parameter of the coil can be expressed as:

[0156]

[0157] The coil length l i and cross-sectional area A are shown below

[0158] l i =2π[(R o +(N i -1)(R_a+Rw)]

[0159]

[0160] Through the above parsing and calculation, the mutual inductance, self-inductance, and internal resistance parameters of the coil under the shielded magnetic core can be obtained. However, in the mirror image method, the coil is regarded as a filament, and the magnetic core is regarded as a material with an infinite volume, ignoring the influence of the finite size of the magnetic core and the wire diameter parameters on the electromagnetic parameters, resulting in its inability to effectively reflect the system parameters when the wire diameter and magnetic core parameters change. Therefore, it is used as a supplementary data set for the training of the neural network.

[0161] (2) Siamese neural network framework

[0162] The siamese neural network proposed in the embodiment of the present invention can be divided into two main parts: the first sub-neural network and the second sub-neural network. They enhance the generalization ability of the model by sharing some weights.

[0163] Specifically, the purpose of the first sub-neural network is to predict the electromagnetic parameters (mutual inductance M, self-inductance L 1 、L 2 , internal resistance R 1 、R 2 , coupling coefficient ε) based on the coil size parameters (number of turns, inner diameter, outer diameter, wire diameter, turn pitch). The input layer receives the coil size parameters such as the number of turns, inner diameter, outer diameter, wire diameter, and turn pitch.

[0164] Specifically, the structure of the first sub-neural network mainly includes a fully connected layer, an activation function, an output layer, a loss function, and training data.

[0165] Specifically, the fully connected layer of the first sub-neural network contains 3 hidden layers, each composed of 128 neurons; the activation function uses ReLU (Rectified Linear Unit) as the activation function to improve the non-linear learning ability of the model; the output layer has 6 neurons, corresponding to the following coil parameters (mutual inductance M, self-inductance L 1 、L 2 , internal resistance R 1 、R 2 , coupling coefficient ε); the input data is the first data set (coil size parameters and their corresponding electromagnetic parameters) and part of the third data set (i.e., the parsing and calculation data set), which is divided into a training set and a validation set according to a ratio of 8:2 for training and validation.

[0166] Specifically, for the first sub-network, the loss function is:

[0167]

[0168] Where, M i 、L 1i 、L 2i 、R 1i 、R 2i and ε iis the true value of the i-th sample, and is the predicted value of this sample, and N is the total number of samples.

[0169] Specifically, the task of the second sub-neural network is to predict the coil parameter changes (differences: mutual inductance ΔM, self-inductance ΔL 1 , ΔL 2 , internal resistance ΔR 1 , ΔR 2 , coupling coefficient Δε) caused by the position parameters (offsets in the xyz directions and angular offset) and core parameters (core size and thickness).

[0170] Similarly, the structure of the second sub-neural network mainly includes a fully connected layer, an activation function, an output layer, a loss function, and training data.

[0171] Specifically, the fully connected layer of the second sub-neural network contains 3 hidden layers, each composed of 128 neurons; the ReLU is used as the activation function; the output layer has 6 neurons, corresponding to the following coil parameters (mutual inductance M, self-inductances L1, L2, internal resistances R1, R2, coupling coefficient ε) respectively; the input data is the second data set (position parameters and core parameters) and part of the third data set (i.e., the analytical calculation data set), which is divided into a training set and a validation set according to a ratio of 8:2 for training and validation.

[0172] Specifically, for the second sub-network, the loss function is:

[0173]

[0174] where, ΔM i , ΔL 1i , ΔL 2i , ΔR 1i , ΔR 2i and Δε i are the true values of the i-th sample, and are the predicted values of this sample, and N is the total number of samples.

[0175] During the training process of the siamese neural network, it is necessary to perform a weighted sum of the losses of the two sub-networks to obtain the total loss function. Since these two sub-networks handle different tasks (one is to predict electromagnetic parameters and the other is to predict differences), it is necessary to reasonably combine the losses of both.

[0176] The total loss function is:

[0177] L loss = λ 1 gL loss1 + λ 2gL loss2

[0178] where L loss1 is the loss function of the first sub-network, which is responsible for predicting the electromagnetic parameters under the coil size. L loss2 is the loss function of the second sub-network, which is responsible for predicting the difference in electromagnetic parameters under the position offset and core parameters. λ 1 and λ 2 are hyperparameters used to control the importance of the two loss terms and adjust the weights.

[0179] By minimizing the total loss function L loss , the weights of the neural network are adjusted so that the first sub-network can better predict the electromagnetic parameters under the coil size, and the second sub-network can better predict the difference in electromagnetic parameters under the offset and core changes. During the training process, the total loss function is calculated using the validation set, and methods such as cross-validation are combined to adjust the hyperparameters λ 1 and λ 2 , thereby balancing the training contributions of the two sub-networks.

[0180] During the actual training process, in order to better integrate the contributions of the simulation data set and the analytical data set, an adaptive weighting mechanism is introduced. By dynamically adjusting the values of λ 1 and λ 2 , the two sub-networks can adjust their weights according to their performance on the validation set during training, thus ensuring the balance between data sources.

[0181] It should be noted that the core idea of the siamese neural network is that through weight sharing, the two sub-networks can use similar feature extraction methods during the learning process.

[0182] Specifically, the first sub-neural network and the second sub-neural network share the same weights in the first few layers of the network. This means that the two sub-networks will learn similar features, thus generating consistent output features under different input conditions. Through the shared network layers, the two sub-networks can extract features with similar physical meanings from the input data. This sharing ability enables the network to learn general features across data sources during training, thereby improving the generalization ability of the model. Although the network layers are shared, the output layers of each sub-network are independent and are responsible for the predictions under the coil size parameters and the position offset and core parameters respectively. This design enables each sub-network to be finely tuned for specific tasks while also leveraging the shared feature extraction ability to improve the overall performance.

[0183] Further, during the training process, the parameters of the network are optimized using gradient descent (the Adam optimizer is adopted in the present invention). During the training process, the first sub-neural network and the second sub-neural network are respectively trained using corresponding training sets, and the network weights are dynamically adjusted according to the effects of the validation sets.

[0184] Further, during training, the contributions of the simulation data set and the analytical data set are balanced by weighting. During the network training process, according to the performance of the model on the validation set, the weights of different data sets are adaptively adjusted, so as to ensure the balance of the contributions of different data sources to the model.

[0185] Further, after the training is completed, the first prediction model can predict the electromagnetic parameters according to different coil size parameters. The second prediction model can predict the difference in electromagnetic parameters according to different offsets and core parameters. Finally, by adding the output results of the two models, the complete electromagnetic parameters of the coil can be obtained.

[0186] Finally, the test data set is used to evaluate the difference between the prediction results and the actual simulation data, and the prediction accuracy of the model is verified by calculating the error value.

[0187] In summary, the method of the present invention has at least the following advantages and beneficial effects compared with the prior art:

[0188] (1) Due to the adoption of the siamese neural network method, the dependence on large-scale data sets is significantly reduced

[0189] By splitting the parameter prediction task of the wireless power transfer coil into two neural networks: the first neural network focuses on predicting the initial system parameters according to the coil size parameters, and the second neural network focuses on predicting the difference in system parameters under offset conditions according to the position offset and core parameters. In this way, the present invention effectively reduces the demand for huge data sets, and can still ensure high prediction accuracy especially in the case of insufficient data.

[0190] (2) Due to the combination of analytical calculation and simulation data, the generalization ability and accuracy of the prediction model are improved

[0191] The present invention combines analytical calculation and finite element simulation data. The analytical calculation method provides preliminary training data for the neural network, and at the same time the finite element simulation provides more accurate system parameters. The fusion of these data sources enables the model to learn the characteristics of the coil parameters from different perspectives, avoids the overfitting problem that may be brought by a single data source, and improves the accuracy and robustness of the prediction results.

[0192] (3) By adaptively adjusting the weights of the analytical calculation and simulation data sets, the adaptability of the model is improved

[0193] During the training process, the present invention can adaptively adjust the weight ratio of the parsing calculation dataset and the simulation dataset according to the training effect, ensuring the balanced contribution of different data sources to model training. This dynamic adjustment mechanism enables the model to better adapt to different design scenarios and still ensure a high prediction accuracy even when the data is insufficient or the simulation calculation cannot cover all cases.

[0194] Embodiment 2

[0195] An embodiment of the present invention further provides an electronic device, which includes a processor and a memory. At least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement Figure 1 a method for predicting coil parameters of a wireless power transmission system as shown.

[0196] It can be understood that the memory may include a random access memory (RAM) and may also include a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the above method embodiments, etc.; the data storage area may store data created according to the use of the server, etc.

[0197] The processor may include one or more processing cores. The processor uses various interfaces and circuits to connect various parts within the entire server. By running or executing instructions, programs, code sets, or instruction sets stored in the memory, and by invoking data stored in the memory, the processor performs various functions of the server and processes data. Optionally, the processor may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor may integrate one or a combination of several of a central processing unit (CPU) and a modem, etc. Among them, the CPU mainly processes the operating system and application programs, etc.; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor and may be implemented separately through a single chip.

[0198] Since this electronic device is the electronic device corresponding to a method for predicting coil parameters of a wireless power transmission system according to an embodiment of the present invention, and the principle by which this electronic device solves problems is similar to that of this method, the implementation of this electronic device can refer to the implementation process of the above method embodiment, and repeated parts will not be elaborated.

[0199] Embodiment 3

[0200] An embodiment of the present invention further provides a computer-readable storage medium, in which at least one instruction, at least one segment of program, code set, or instruction set is stored, and the at least one instruction, the at least one segment of program, the code set, or the instruction set is loaded and executed by a processor to implement Figure 1 a method for predicting coil parameters of a wireless power transmission system as shown.

[0201] Those of ordinary skill in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and this program can be stored in a computer-readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disc memories, tape memories, or any other computer-readable medium capable of carrying or storing data.

[0202] Since this storage medium is the storage medium corresponding to a method for predicting coil parameters of a wireless power transmission system according to an embodiment of the present invention, and the principle of solving problems by this storage medium is similar to that of this method, the implementation of this storage medium can refer to the implementation process of the above method embodiment, and repeated parts will not be elaborated.

[0203] Embodiment 4

[0204] In some possible implementation manners, various aspects of the method according to an embodiment of the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a computer device, the program code is used to cause the computer device to execute the steps of a method for predicting coil parameters of a wireless power transmission system according to various exemplary implementation manners described above in this specification. Among them, the executable computer program code or "code" for executing each embodiment can be written in a high-level programming language such as C, C++, C#, Smalltalk, Java, JavaScript, Visual Basic, structured query language (for example, Transact-SQL), Perl, or written in various other programming languages.

[0205] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0206] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0207] The above embodiments are only for illustrating the technical concept and features of the present invention, and the purpose is to enable those of ordinary skill in the art to understand the content of the present invention and implement it accordingly, and it cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the essence of the content of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting coil parameters of a wireless power transmission system, characterized in that: The steps include: Construct a simulation model of a wireless power transfer coil with magnetic shielding. By performing parameter scans on different coil size parameters, obtain a simulation data set as the first data set. By performing parameter scans at different positions and with different magnetic core parameters, obtain a simulation data set as the second data set. Calculate the parameters of the wireless power transfer coil with a shielded magnetic core, and obtain an analytical calculation data set as the third data set. Divide the first data set and the second data set into training sets and validation sets respectively, and use the third data set as a supplement to the training set to ensure the balance of the data sources. Set different simulation cases to obtain a test data set. Construct and train a siamese neural network: The first neural network is used to predict the electromagnetic parameters based on the coil size, and the second neural network is used to predict the difference in electromagnetic parameters based on the changes in position and magnetic core parameters. Train the first neural network according to the first training set and the third data set, and train the second neural network according to the second training set and the third data set. The two sub-networks of the siamese neural network share some parameters, and jointly train the two parts of the network through a loss function. Use the method of combining prediction models to predict the coil parameters and obtain the final prediction results. Compare the model prediction results with the actual parameters of the test data set, calculate the error, and evaluate the performance of the model.

2. A method for predicting coil parameters of a wireless power transmission system according to claim 1, characterized in that: The calculation of the parameters of the wireless power transfer coil with a shielded magnetic core includes: Combining the mirror method and the planar coil parameter calculation method to calculate the parameters of the wireless power transfer coil with a shielded magnetic core.

3. A method for predicting coil parameters of a wireless power transmission system according to claim 2, characterized in that: The process of calculating the coil parameters using the mirror method is as follows: 1) Calculate the parameters of a single-turn coil without a magnetic core: According to the Neumann equation, the mutual inductance M between concentric single-turn coils can be expressed as: By introducing the parameter γ: Obtain the mutual inductance of concentric single-turn coils: For the parameters of a single-turn coil in the offset case: Introduce the parameters γ, α, β, and δ: According to the Neumann equation, the mutual inductance of a single-turn coil in the offset case is: Where a is the radius of the transmitting coil, b is the radius of the receiving coil, d is the transmission distance between the transmitting coil and the receiving coil, x is the lateral offset between the transmitting coil and the receiving coil; μ0 is the vacuum magnetic permeability; is the angle between the transmitting coil current element and the initial position, is the angle between the receiving coil current element and the initial position; R 12 is the distance between the current elements in the transmitting coil and the receiving coil; l1 is the current element of the transmitting coil, and l2 is the current element of the receiving coil; 2) Calculate the parameters of a multi-turn coil without a magnetic core The radius of each turn of the transmitting coil is expressed as: R i =R oi +N i ×(a i +R wi ) The radius of each turn of the receiving coil is expressed as: R j =R oj +N j ×(a j +R wj ) In the formula, R oi is the inner diameter of the transmitting coil, N i is the number of turns of the transmitting coil, a i is the wire diameter of the transmitting coil, R wi is the turn spacing between each turn of the transmitting coil; Similarly, R oj is the inner diameter of the receiving coil, N j is the number of turns of the receiving coil, a j is the wire diameter of the receiving coil, R wj is the turn spacing between each turn of the receiving coil; The mutual inductance between different turns of the coil is expressed as: Calculate the mutual inductance between each turn of the transmitting coil and the receiving coil, and add them up to obtain the mutual inductance parameter of the multi-turn coil without a magnetic core: After adding the magnetic shielding layer, the source current is alternately reflected by two magnetic planes, and an infinite number of sets of image currents will be formed. They have equal amplitudes and are located on the planes of z = ±d, ±3d, ±5d... respectively. Obtain the mutual inductance parameter of the multi-turn coil with a shielded magnetic core: In the formula, n is the number of mirror reflections; Calculate the self-inductance parameter of the multi-turn coil with a shielded magnetic core: Its self-inductance is expressed as: L=L self +M self +M mirror Where, L self is the self-inductance of each coil turn, M self is the mutual inductance between each coil turn, M mirror is the mutual inductance between the image coil and the source coil generated by the image current; Further expressed as: In the formula, For coil i a Circle and i b Mutual inductance between loops; M ij is the mutual inductance between the mirror coil and the i-th and j-th turns of the original coil; By联立 the above formulas, obtain the self-inductance parameter of the multi-turn coil with a shielded magnetic core.

4. A method for predicting coil parameters of a wireless power transmission system according to claim 3, characterized in that: The resistance parameter of the coil is expressed as: In the formula, l i is the length of the i-th coil, A is the cross-sectional area of ​​the coil, and ρ is the resistivity.

5. The method for predicting coil parameters of a wireless power transmission system according to claim 1, characterized in that: The input of the first neural network is the coil size parameters, and the output is each coil parameter. Among them, the coil size parameters include the number of turns, inner diameter, outer diameter, wire diameter, turn spacing, and each coil parameter includes the mutual inductance M, self-inductances l1, L2, internal resistances R1, R2, and coupling coefficient ε; The input of the second neural network includes position parameters, different core sizes and core thicknesses, and the output is the difference between the parameters of each coil and the coil in a constant state; wherein the difference includes mutual inductance ΔM, self-inductance ΔL1, ΔL2, internal resistance ΔR1, ΔR2, and coupling coefficient Δε.

6. A method for predicting coil parameters of a wireless power transmission system according to claim 1, characterized in that: The structure of the first sub-neural network includes a fully connected layer, an activation function and an output layer; The fully connected layer includes 3 hidden layers, each of which consists of 128 neurons; the activation function uses ReLU as the activation function to improve the nonlinear learning ability of the model; the output layer is 6 neurons, corresponding to the following coil parameters: mutual inductance M, self-inductance L1, L2, internal resistance R1, R2, and coupling coefficient ε.

7. A method for predicting coil parameters of a wireless power transmission system according to claim 1, characterized in that: During the training process, the loss function of the first neural network is: Where M i , L 1i , L 2i , R 1i , R 2i and ε i is the true value of the i-th sample, where M represents the mutual inductance parameter, L1 represents the self-inductance parameter of the transmitting coil, L2 represents the self-inductance parameter of the receiving coil, R1 represents the internal resistance parameter of the transmitting coil, R2 represents the internal resistance parameter of the receiving coil, and ε represents the coupling coefficient; and is the predicted value of the sample, N is the total number of samples.

8. A method for predicting coil parameters of a wireless power transmission system according to claim 1, characterized in that: During the training process, the loss function of the second neural network is: In the formula, ΔM i , ΔL 1i , ΔL 2i , ΔR 1i , ΔR 2i and Δε i is the true value of the i-th sample, and is the predicted value of the sample, and N is the total number of samples.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 8.

Images