Human body electromagnetic exposure assessment method for wireless charging system based on K-GRU

Abstract

The invention is suitable for the technical field of wireless electric energy transmission of electric vehicles, and provides a K-GRU-based human body electromagnetic exposure assessment method for a wireless charging system. In order to solve the electromagnetic exposure safety problem of an EV-WPT system on an adult male implanted with an intracranial artery stent under uncertainty factors, an EV-WPT system human body electromagnetic exposure safety uncertainty quantification proxy model method based on a K-GRU architecture is adopted to quantify the internal electromagnetic exposure dose. According to the method, the MLP structure is optimized through the KAN, the fitting effect is improved, the function self-learning ability and the network interpretability are activated, and the spatial relationship among the human body, the EV and the WPT system is comprehensively considered. Results show that the K-GRU proxy model can efficiently evaluate the electromagnetic exposure dose, the calculation time cost is only 5% of that of an MC method, a scientific basis is provided for EV-WPT system electromagnetic exposure safety evaluation and protection, and the method has important engineering application value.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless power transmission of electric vehicles, and in particular relates to a method for assessing human electromagnetic exposure of a wireless charging system based on K-GRU. Background Art

[0002] The rapid development of the electric vehicle (EV) industry has significantly reduced dependence on fossil fuels and environmental pollution, and promoted the green and low-carbon transformation of the economy and society. At the same time, the emergence of wireless power transfer (WPT) technology has broken away from the limitations of traditional wired charging wires and improved the safety, reliability and comfort of charging. WPT technology uses a wire-free contact method and relies on electromagnetic fields excited in space to achieve the transmission of electrical energy. The magnetically coupled resonant wireless power transfer (MCR-WPT) technology for electric vehicles is a key technology to solve the charging problem. In the long run, it is expected to realize the charging of smart cars while parking, greatly improving the convenience of charging.

[0003] As the EV market expands, the demand for high-power WPT systems is increasing to achieve fast charging. Since the energy transmission device of the WPT system is a loosely coupled structure, leakage electromagnetic fields will inevitably be generated during the energy transmission process, and the high-power WPT system will generate high-intensity leakage electromagnetic fields in the surrounding space. If the human body is exposed to such an electromagnetic environment for a long time, it may face electromagnetic safety threats. Especially for people with metal medical implants in their bodies, such as patients with intracranial arterial stents, these implants will change the distribution of electromagnetic fields and may aggravate the impact of leakage electromagnetic fields on the human body, thereby causing harm to the human body. In addition, the human safety issues in the electromagnetic environment of the EV-WPT system are also affected by a variety of uncertain factors, such as the system manufacturing process, coil offset caused by the driver's parking, and the human exposure position. These factors make the traditional analysis of human electromagnetic exposure based on a fixed environment without variables less comprehensive.

[0004] To ensure the safety of EV-WPT systems, their leakage electromagnetic fields must comply with international standards, such as the limits set by the International Commission on Non-Ionizing Radiation Protection (ICNIRP), the Institute of Electrical and Electronics Engineering (IEEE), and the Society of Automotive Engineers (SAE). Currently, WPT systems usually operate at a frequency of 85kHz. The ICNIRP guidelines point out that the safety assessment of human electromagnetic exposure below 100kHz mainly uses the induced electric field strength as the evaluation target.

[0005] Based on experimental ethics, it is impossible to measure the electromagnetic radiation inside the human body by using humans as experimental subjects through experiments. Therefore, researchers usually use methods such as theoretical calculations or simulation software to determine the electromagnetic radiation distribution of EV-WPT systems. However, most existing studies are based on the analysis of human electromagnetic exposure under a fixed environment without variable states, only considering the working conditions of WPT systems under ideal conditions and ignoring the impacts brought by various uncertainty factors. The traditional Monte Carlo simulation method is an uncertainty quantification method that relies on multiple simulation results, with high computational costs and difficulty in achieving efficient uncertainty quantification. In recent years, the surrogate model-assisted analysis method based on deep learning architectures has received extensive attention, obtaining relatively reliable model results with a small number of training samples. Among them, the Gate Recurrent Unit (GRU) has been widely used in data fitting tasks. In 2024, a network called KAN was proposed by HIT, further improving the generalization ability and interpretability of the model through a learnable activation function, overcoming the limitations of traditional networks relying on sigmoid to achieve a single activation function. In addition, there are currently evaluation methods based on sparse chaotic polynomials and their improvements, but they have the problem of "curse of dimensionality" under high-dimensional inputs and insufficient universality. The traditional BP neural network has an overly simplified network structure, resulting in unsatisfactory learning effects and insufficient comprehensiveness of data fitting. Therefore, the present invention proposes a method for evaluating human electromagnetic exposure in a wireless charging system based on K-GRU. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for evaluating human electromagnetic exposure in a wireless charging system based on K-GRU, aiming to solve the problems proposed in the above background technology.

[0007] The purpose of the present invention is achieved through the following technical solutions:

[0008] A method for evaluating human electromagnetic exposure in a wireless charging system based on K-GRU includes the following steps:

[0009] Step 1: Establish an equivalent circuit model of the MCR-WPT system;

[0010] Based on the MCR-WPT system, an S-S type compensation circuit structure is adopted to reduce reactive power;

[0011] Step 2: Construct a K-GRU surrogate model;

[0012] Adopt a GRU network as the basic architecture and introduce a gating mechanism; the GRU network structure includes an update gate and a reset gate;

[0013] Introduce KAN, improve the activation function from a fixed form (such as sigmoid) to a learnable B-spline function; use KAN to decompose a high-dimensional function into a combination of univariate functions;

[0014] Step 3: Train the K-GRU surrogate model;

[0015] Train the K-GRU surrogate model to obtain a surrogate model for quantifying the safety uncertainty of human electromagnetic exposure in the EV-WPT system;

[0016] Step 4: Uncertainty quantification analysis;

[0017] Use the Monte Carlo method to call the surrogate model to obtain N MC samples and their corresponding output responses, and calculate the statistical characteristic parameters of the internal exposure dose of the human body.

[0018] Furthermore, in the S-S type compensation circuit, U s is the input high-frequency voltage source, I1 and I2 are the currents at the transmitting end and the receiving end respectively, R S1 is the equivalent resistance of the transmitting coil T x , C S1 is the series compensation capacitor of the transmitting coil T x , L1 and L2 are the self-inductances of the coils of the transmitting coil T x and the receiving coil R x respectively, M is the mutual inductance between the transmitting coil T x and the receiving coil R x , C S2 is the series compensation capacitor of the receiving coil R x , R S2 is the equivalent resistance of the receiving coil R x , R L is the load resistance;

[0019] According to the Kirchhoff voltage law of the circuit, the loop voltage equivalent formula is obtained:

[0020]

[0021] where Z is the system impedance matrix, w is the resonant angular frequency, and j is the imaginary part of the complex number; when the WPT system undergoes series resonance, the following equation holds:

[0022]

[0023] Therefore, Equation 3 exists:

[0024]

[0025] The S-S compensation circuit is in a resonant state to ensure the transmission efficiency of the WPT system. The circuit parameters of the transmitter and receiver are shown in Equations 4 and 5 respectively;

[0026] Resonant condition of the transmitter:

[0027]

[0028] Resonant condition of the receiver:

[0029]

[0030] where w is the resonant angular frequency, with a value of 2πf, and f is the resonant frequency.

[0031] Furthermore, the calculation process of the GRU is as follows:

[0032] Input initialization: Assume the input sequence is x = [x1, x2,..., x t , where x t is the output at the t-th step; initialize the hidden state h0;

[0033] Calculate the reset gate: The calculation formula of the reset gate r t is as follows:

[0034] Equation 6: r t = σ(W r · [h t-1 , x t + b r );

[0035] where σ is the sigmoid activation function, W r is the weight matrix of the reset gate, b r is the bias term, and [h t-1 , x t represents the vector obtained by concatenating the hidden state of the previous time step and the current input;

[0036] Calculate the candidate hidden state: Use the reset gate r t to control the influence of the hidden state h t-1 of the previous time step. The formula is as follows:

[0037]

[0038] where, is the candidate hidden state, tanh is the hyperbolic tangent function, is the weight matrix of the candidate hidden state, is the bias term;

[0039] Calculate the update gate: The calculation formula of the update gate z t is as follows:

[0040] Equation 8: z t = σ(W z ·[h t-1 , xt] + b z );

[0041] where W z is the weight matrix of the update gate, and b z is the bias term;

[0042] Calculate the current hidden state: Use the update gate z t to combine the hidden state h t-1 at the previous time step and the current hidden state h t , and the formula is as follows:

[0043]

[0044] where 1 - z t and z t respectively represent the proportions of retaining the state at the previous moment and adopting the candidate hidden state determined by the update gate.

[0045] Furthermore, the KAN achieves high-dimensional decomposition by parameterizing each one-dimensional function as a B-spline function. The p-order B-spline is recursively defined by the de Boor-cox formula as follows:

[0046]

[0047] where N i,1 (x) is the recursive definition formula of the 1st-order B-spline, representing the i-th 1st-degree B-spline; u is called the knot vector, u i is called the knot, u i+1 is the (i + 1)-th knot; the semi-open interval [u i , u i+1 is the i-th knot interval; p is the degree of the basis function, and the i-th p-th B-spline is written as N i,p (u); u i+p is the (i + p)-th knot, N i,p-1 (u) is the i-th (p - 1)-th B-spline, u i+p+1 is the (i + p + 1)-th knot, and N i+1,p-1 (u) is the (i + 1)-th (p - 1)-th B-spline.

[0048] Furthermore, the specific process of step 4 is as follows:

[0049] Start iterative calculation of the reset gate r t , the update gate z t , the current hidden state ht and the candidate hidden state

[0050] For each time step t from 1 to N, perform the following operations:

[0051] a. Obtain the input data x at the current time step t ;

[0052] b. Calculate the reset gate r t ;

[0053] c. Calculate the candidate hidden state

[0054] d. Calculate the update gate z t ;

[0055] e. Save the hidden state h at the current time step t for the next iteration, h t-1 = h t h t is the final hidden state, containing the information of the sequence;

[0056] Calculate using the trained K-GRU surrogate model the corresponding output response

[0057] Calculate Y MC the probability distribution function f y (y) to obtain the statistical feature parameters.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] To evaluate the electromagnetic exposure safety of the human body with implanted medical devices in the head (such as intracranial artery stents) under the influence of uncertain factors in the EV-WPT system, the present invention focuses on adult males with implanted intracranial artery stents around the 11kW transmission power EV-WPT system, and quantitatively analyzes the electromagnetic exposure dose inside them. The method of using the EV-WPT system human electromagnetic exposure safety uncertainty quantification surrogate model based on the K-GRU architecture is adopted to achieve accurate equivalence in the evaluation under different exposure scenarios. This method optimizes the MLP structure with KAN on the basis of the traditional GRU method, significantly improving the fitting effect of the model, the self-learning ability of the activation function, and the interpretability of the network, while ensuring the accurate approximation ability of the high-dimensional function. In addition, the present invention comprehensively considers the spatial position relationship among the human body, EV, and WPT system, and conducts uncertainty quantification analysis on the electric field strength dose inside the human body. The results show that the K-GRU surrogate model can quickly and efficiently evaluate the electromagnetic exposure dose inside adult males with implanted intracranial artery stents around the EV-WPT system. Under the same scenario, the computational time cost of the K-GRU method is only 5% of that of the MC method, significantly improving the computational efficiency. The present invention provides a scientific basis and reasonable suggestions for the electromagnetic exposure safety evaluation and protection of the EV-WPT system for the human body, and has important engineering application value. Description of the Drawings

[0060] Figure 1 It is the flowchart of the method of the present invention.

[0061] Figure 2 It is the schematic diagram of the MCR-WPT system for electric vehicles.

[0062] Figure 3 It is the S-S compensation circuit model of the MCR-WPT system.

[0063] Figure 4 It is the GRU network structure.

[0064] Figure 5 It is the EV-WPT system.

[0065] Figure 6 It is the WPT system.

[0066] Figure 7 It is the human body model.

[0067] Figure 8 It is the brain tissue model and its mesh dissection diagram.

[0068] Figure 9 It is the detailed diagram of the intracranial artery stent model.

[0069] Figure 10 It is the schematic diagram of the human body in the vehicle.

[0070] Figure 11 Schematic diagram of the human body on the side of the vehicle; among them, (a) is the front view of the human body on the side of the vehicle, and (b) is the top view of the human body on the side of the vehicle.

[0071] Figure 12 Schematic diagram of the human body at the rear of the vehicle; among them, (a) is the front view of the human body at the rear of the vehicle, and (b) is the top view of the human body at the rear of the vehicle.

[0072] Figure 13 Probability density function of E in the human body when the human body is in the driver's seat inside the vehicle max ; among them, (a) is E in the brain of a human body without an intracranial artery stent implanted inside the vehicle max , (b) is E inside the entire human body without an intracranial artery stent implanted inside the vehicle max , (c) is E in the brain of a human body with an intracranial artery stent implanted inside the vehicle max , (d) is E inside the entire human body with an intracranial artery stent implanted inside the vehicle max .

[0073] Figure 14 Probability density function of E in the human body when the human body is on the side of the vehicle max ; among them, (a) is E in the brain of a human body without an intracranial artery stent on the side of the vehicle max , (b) is E inside the entire human body without an intracranial artery stent on the side of the vehicle max , (c) is E in the brain of a human body with an intracranial artery stent on the side of the vehicle max , (d) is E inside the entire human body with an intracranial artery stent on the side of the vehicle max .

[0074] Figure 15 Probability density function of E in the human body when the human body is at the rear of the vehicle max ; among them, (a) is E in the brain of a human body without an intracranial artery stent at the rear of the vehicle max , (b) is E inside the entire human body without an intracranial artery stent at the rear of the vehicle max , (c) is E in the brain of a human body with an intracranial artery stent at the rear of the vehicle max , (d) is E inside the entire human body with an intracranial artery stent at the rear of the vehicle max . Detailed implementation manner

[0075] For a clearer understanding of the technical features, objectives, and beneficial effects of the present invention, the technical solution of the present invention will be described in detail below, but it should not be construed as a limitation on the scope of implementation of the present invention.

[0076] The present invention provides a method for evaluating human body electromagnetic exposure of a wireless charging system based on K-GRU, as Figure 1As shown, it includes the following steps:

[0077] Step 1: Establish an equivalent circuit model of the MCR-WPT system;

[0078] The magnetically coupled resonant wireless power transfer (MCR-WPT) system is currently the most mature charging technology applied to electric vehicles. Based on Faraday's law of electromagnetic induction and the theory of near-field strong coupling, it realizes efficient energy transfer through the mutual coupling magnetic field between the transmitting coil and the receiving coil, with the electromagnetic field as the medium. The MCR-WPT system includes structures such as a wireless charging power supply for high-frequency inversion, a transmitting end, a receiving end, and a load. The schematic diagram is as Figure 2 shown. Its working principle is: The industrial-frequency alternating current input from the power grid is transformed into direct current through the rectifier circuit at the transmitting end, and the direct current is inverted into alternating current with a higher frequency through the high-frequency inverter circuit. The power frequency output by the inverter circuit is the resonant frequency of the transmitting resonant circuit. At this time, the transmitting end circuit works in the resonant state under the action of the transmitting end compensation circuit and presents a pure resistive state. According to Faraday's law of electromagnetic induction, the transmitting coil T x excites a high-frequency electromagnetic field with a frequency of the resonant frequency into space, and is inductively received by the receiving coil R x At this time, under the action of the receiving end compensation circuit, the receiving end circuit works in the resonant state and also presents a pure resistive state. The receiving end rectifier circuit and the voltage regulation (DC-DC) circuit transform the high-frequency power supply into the direct current required by the load. The transmitting end circuit and the receiving end circuit work at the resonant frequency, and the electric energy is transmitted through the coupling of the high-frequency electromagnetic field generated by resonance.

[0079] The magnetically coupled resonant WPT system realizes efficient energy transfer through the coupling of the transmitting coil T x and the receiving coil R x Since in actual use, the coil has an internal resistance, and the loss of the internal resistance cannot be eliminated. Therefore, if the transmission efficiency of the WPT system is to be improved, it is necessary to ensure that the system works in the resonant state, which is generally achieved by using a reactive power compensation circuit composed of capacitors to reduce the reactive power of the system and thus improve the efficiency. In the design process of the WPT system, the primary side adopts a series topology structure. Compared with the parallel structure, it can effectively increase the coupling degree between the primary side and the secondary side of the circuit. Therefore, for the topology structure of the compensation circuit, the present invention selects a series-series (S-S) type compensation circuit structure with high-efficiency resonant energy transfer characteristics, Figure 3 shows the linear equivalent circuit diagram of the magnetically coupled resonant WPT system containing the S-S compensation circuit. The compensation capacitor of the S-S compensation structure is independent of the mutual inductance of the WPT system coil and the load, and is suitable for the situation where there is movement at the transmitting end and the receiving end, and has the characteristics of high-efficiency resonant energy transfer.

[0080] In the S-S compensation topology network circuit, Us is the input high-frequency voltage source, I1 and I2 are the currents at the transmitting end and the receiving end respectively, and R S1 is the equivalent resistance of the transmitting coil T x , C S1 is the series compensation capacitor of the transmitting coil T x , L1 and L2 are the self-inductances of the transmitting coil T x and the receiving coil R x respectively, M is the mutual inductance between the transmitting coil T x and the receiving coil R x , C S2 is the series compensation capacitor of the receiving coil R x , R S2 is the equivalent resistance of the receiving coil R x , R L is the load resistance.

[0081] According to the Kirchhoff voltage law of the circuit, the equivalent formula of the loop voltage can be obtained:

[0082]

[0083] where Z is the system impedance matrix, w is the resonant angular frequency, and j is the imaginary part of the complex number. When the WPT system undergoes series resonance, the following equation holds:

[0084]

[0085] Therefore, there is Equation 3:

[0086]

[0087] To ensure the transmission efficiency of the WPT system, the S-S compensation circuit should be in a resonant state. The circuit parameters of the transmitting end and the receiving end are shown in Equations 4 and 5 respectively.

[0088] Resonant condition at the transmitting end:

[0089]

[0090] Resonant condition at the receiving end:

[0091]

[0092] where w is the resonant angular frequency, with a value of 2πf, and f is the resonant frequency.

[0093] It can be seen from the above formula that the mutual inductance between the coils will affect the working state of the WPT system, and the self-inductance of the coupled coils and the mutual inductance between the coils will change under the influence of uncertain factors. Therefore, it is necessary to analyze the safety of human electromagnetic exposure when the WPT system operates in a non-ideal state.

[0094] Step 2: Construct a K-GRU surrogate model; specifically including:

[0095] 1. The architecture of the surrogate model for electromagnetic exposure safety analysis based on GRU;

[0096] During the ideal working state of the EV-WPT system, the electromagnetic exposure doses received by the driver and passengers are fixed. However, during the actual use of the EV-WPT system, it is almost impossible for the WPT system to work in a completely ideal state. For example, in the WPT system, the relative distance between the transmitter and receiver coils is susceptible to various dynamic factors. Among them, the working state of the vehicle chassis height and the changes in the chassis structure will cause significant changes in the coil spacing. In addition, from the perspective of ergonomics, subjective behavioral factors such as the position distribution and posture adjustment of the occupants in the vehicle will also cause fluctuations in the electromagnetic coupling distance between the human body and the WPT system. The interaction of these complex variables makes the system working environment exhibit significant uncertainty characteristics. Therefore, the electromagnetic exposure dose of the human body also changes with the influence of uncertainty. Currently, in order to improve the safety, reliability, comfort, and charging speed of charging, the power of the WPT system usually reaches several kilowatts or even dozens of kilowatts, which makes the electromagnetic leakage caused by uncertainty factors non-negligible, and it will inevitably increase the risk of non-ionizing radiation to which the human body is exposed around the EV-WPT system.

[0097] Considering that some people need to have medical implants in their bodies due to the treatment of certain diseases. These implants are usually made of metal alloy materials and have the ability to change the distribution of electromagnetic fields (EMF) and attract electromagnetic fields. For example, the intracranial artery stent is a mesh-shaped ball-expandable stent made of medical-grade 316L stainless steel metal material. Its permanent implantation in intracranial blood vessels can significantly improve hemodynamics, prevent thrombosis, and promote the repair and reconstruction of vascular walls. It is widely used in the treatment of symptomatic intracranial atherosclerotic disease (ICAD) and aneurysms. When metal implants in the human body are exposed to the leakage electromagnetic field of the EV-WPT system, the distribution of electromagnetic energy may be changed, especially at the edge of the tip of the precision metal implant, which is more likely to aggravate the impact of the leakage electromagnetic field of the EV-WPT system on the human body, thereby causing harm to the human body. Considering many uncertain factors such as the manufacturing process of the components of the WPT system of electric vehicles, the coil offset caused by the driver's parking, and the exposure position of the human body, it is obvious that there will be significant changes and impacts on the safety of the human body exposed to the electromagnetic environment of the WPT system of electric vehicles. In response to the above problems, uncertainty quantification (UQ) method will play an important role in realizing the safety assessment of human electromagnetic exposure and providing reasonable suggestions and guidance for further human protection.

[0098] Regarding the above-mentioned issues related to the uncertainty quantification of human electromagnetic exposure in the EV-WPT system, considering the complexity of the EV-WPT system and the human anatomical model, the Monte Carlo method, which has high computational accuracy but also high computational cost, is obviously not applicable. In recent years, uncertainty quantification based on the surrogate model method has been widely applied to uncertainty quantification analysis. Among them, the most widely used is the method related to polynomial chaos expansion (PCE). The surrogate model method based on chaotic polynomials can obtain computational accuracy comparable to that of machine learning-related methods with only fewer training sets. Although the above research has achieved the uncertainty quantification of the WPT working efficiency, due to the limitation of the PCE method driven by probability, the research results have certain limitations. In addition to PCE and its improved methods, uncertainty quantification achieved through deep learning methods has also been widely applied to solve engineering problems in recent years. Among them, GRU can obtain an accurate surrogate model in a data-driven manner with fewer training samples when solving regression problems, and can effectively predict the model, which is fully applicable to the uncertainty quantification problem of EV-WPT human electromagnetic exposure assessment. In addition, Kolmogorov-Arnold Networks (KAN), as a new type of network unit, can improve the fitting effect of traditional MLP units and enhance the accuracy and interpretability of the network. Therefore, in view of the deficiencies in existing research, we fully consider the uncertain factors of human exposure to the EV-WPT system, take the human body with an intracranial artery stent as the research object, and use the GRU architecture (K-GRU) improved based on KAN to quantify the uncertainty of human electromagnetic exposure safety, and accurately and effectively evaluate the human electromagnetic exposure safety of the EV-WPT system.

[0099] GRU is one of the variants of the Recurrent Neural Network (RNN). Due to the fixed gradient transmission method of RNN, it is very prone to the problems of gradient vanishing or gradient explosion, especially when dealing with high-dimensional and long-sequence data, the effect is very poor. To solve the problems of gradient vanishing and explosion in RNN, GRU introduces a gating mechanism, including an update gate and a reset gate, which are used to control the flow of information. Different from LSTM, the structure of GRU is relatively simple, only containing two gates: the update gate and the reset gate. The reset gate determines how to combine the new input information with the previous memory, and the update gate defines the amount of the previous memory saved to the current step. These two gating vectors determine which information can finally be used as the output of the gated recurrent unit. The special feature of these two gating mechanisms is that they can preserve the information in the long-term sequence and will not be removed because they are not relevant to the prediction. The simplification of this structure enables GRU to improve the computational efficiency while maintaining the effect. Because of the simplicity of its network, it is more suitable for constructing high-dimensional and relatively complex networks. The GRU network structure is as Figure 4 shown, and the current input is x t , where σ is the sigmoid activation function, which changes the data into a value within the range of 0-1. The hidden state passed down from the previous node is h t-1 , and this hidden state contains the relevant information of the previous node. According to x t and h t-1 , GRU will obtain the output y t of the current step and the hidden state h t passed to the next node. represents the candidate hidden state. The calculation process of GRU is as follows:

[0100] (1) Input initialization;

[0101] Assume the input sequence is x = [x1, x2,..., x t , where x t is the output of the t-th step. Initialize the hidden state h0, usually as a zero vector or randomly initialized.

[0102] (2) Calculate the reset gate;

[0103] The reset gate r t determines the influence degree of the hidden state h t-1 of the previous time step on the current hidden state h t . The calculation formula of the reset gate is as follows:

[0104] Equation 6: r t = σ(W r · [h t-1 , xt +b r );

[0105] Among them, σ is the sigmoid activation function, W r is the weight matrix of the reset gate, b r is the bias term, [h t-1 , x t represents the vector obtained by concatenating the hidden state at the previous time step and the current input.

[0106] (3) Calculate the candidate hidden state;

[0107] Use the reset gate r t to control the influence of the hidden state h t-1 at the previous time step. The formula is as follows:

[0108]

[0109] Among them, is the candidate hidden state, tanh is the hyperbolic tangent function, is the weight matrix of the candidate hidden state, is the bias term.

[0110] (4) Calculate the update gate;

[0111] The role of the update gate z t is to determine how much information from the previous time step should be retained in the hidden state at the current time step. Its output value ranges between 0 and 1. The larger the value, the more past information is retained, and the smaller the value, the more it depends on the information of the current input. The calculation formula of the update gate is as follows:

[0112]

[0113] Among them, W z is the weight matrix of the update gate, b z is the bias term.

[0114] (5) Calculate the current hidden state;

[0115] Use the update gate z t to combine the hidden state h t-1 at the previous time step and the current hidden state h t , as shown in Equation 9:

[0116]

[0117] Among them, 1 - z t and z t respectively represent the proportions of retaining the state at the previous moment and adopting the candidate hidden state determined by the update gate.

[0118] Through its unique gating mechanism and linear combination method, the GRU controls the influence degree of the hidden state at the previous moment on the current moment through the update gate, thus avoiding the vanishing gradient; and controls the influence degree of the hidden state at the previous moment on the candidate hidden state at the current moment through the reset gate, thus avoiding the exploding gradient. Therefore, it effectively alleviates the problems of vanishing gradient and exploding gradient in traditional RNNs, enabling the gradient to be effectively transmitted during backpropagation, so that the human electromagnetic exposure safety uncertainty quantification model of the EV-WPT system under high-dimensional conditions can better capture long-term dependencies.

[0119] 2. Improved method based on KAN;

[0120] In the network structure of the GRU, whether it is the update gate or the reset gate, their historical state information and current state information are input into the sigmoid activation function after linear transformation. The role of the sigmoid function is to compress the output after linear transformation into the interval (0,1), thus providing a probability value for each element. This probability value determines the importance of the current input and historical state when updating the current hidden state, achieving fine control of information flow. However, the sigmoid activation function is not learnable, that is, no matter what data, linear or non-linear, it is processed in the same way. To address this problem, a new type of network structure, KAN, is proposed. KAN is based on the Kolmogorov-Arnold (K-A) representation theorem, which proves that any continuous function f(x1,...,x n ) can be represented as a nested combination of a finite number of univariate functions:

[0121]

[0122] where f(x) is a multivariate continuous function, x is an n-dimensional vector, that is, the set of input variables; n represents the number of input variables; p and q are indices representing the number of functions and the variables on which each function acts; and Φ q are both univariate functions. The K-A theorem shows through the above formula that a high-dimensional function can be reduced to learning one-dimensional functions of polynomial order, but these one-dimensional functions are not necessarily easy-to-learn smooth functions. Therefore, KAN realizes high-dimensional decomposition by parameterizing each one-dimensional function as a B-spline (B-spline) function. The p-order B-spline is recursively defined by the de Boor-cox formula as follows:

[0123]

[0124] where N i,1B-spline of order 1 is recursively defined as (x), representing the i-th first-order B-spline; u is called the knot vector, and u i is called a knot, and u i+1 is the (i + 1)-th knot; the semi-open interval [u i , u i+1 ) is the i-th knot interval; p is the degree of the basis function, and the i-th p-th order B-spline is written as N i,p (u); u i+p is the (i + p)-th knot, N i,p-1 (u) is the i-th (p - 1)-th order B-spline, u i+p+1 is the (i + p + 1)-th knot, N i+1,p-1 (u) is the (i + 1)-th (p - 1)-th order B-spline.

[0125] After high-dimensional decomposition is achieved through B-spline, each local B-spline basis function has learnable coefficients. In the entire KAN network structure, different from the traditional Multi-Layer Perceptron (MLP) in the GRU network, in the traditional MLP link, there is only a fixed activation function on the neurons. In the MLP, the connection between neurons is usually a real value representing the weight, and the neuron itself is equipped with a non-linear activation function, usually the sigmoid function. Therefore, the calculation process of the MLP is to first weight the weight input and then introduce non-linearity through the activation function. After the present invention replaces the MLP with KAN, a K-GRU proxy model applied to the uncertainty quantification of human electromagnetic exposure safety assessment in the present invention is formed. Regarding the weights of the network, since the B-Spline is learnable, the activation function is thus learnable, and the nodes of KAN simply sum the incoming signals without applying any non-linearity. Utilizing the better interpretability of KAN, it realizes the self-learning of the activation function while moving the activation function to the "edge", parameterizes it as a B-spline function, realizes the smoothing processing of the data, can not only learn features but also optimize these learned features with high precision, thereby obtaining a smooth function approximating the data, enhancing the representation ability, ensuring the accurate approximation ability of the high-dimensional function, and also decomposing the multi-dimensional function into a combination of single-variable functions, simplifying the computational complexity.

[0126] Step 3: Train the K-GRU proxy model;

[0127] Small-sample data with input variables being system parameters and uncertainty factors and output variable being the human electromagnetic exposure dose is obtained through simulation using finite element simulation software. The K-GRU proxy model is trained with the above small-sample data to obtain an efficient and accurate proxy model for uncertainty quantification of human electromagnetic exposure safety assessment of the EV-WPT system.

[0128] Step 4: Uncertainty quantification analysis;

[0129] Using the Monte Carlo (MC) method to call the surrogate model can obtain N MC samples and the corresponding output responses Based on these samples, the statistical characteristic parameters of the human body electromagnetic exposure dose can be directly obtained, including the probability density function PDF of E at different human body parts under different exposure scenarios, and the mean value and over-limit probability of E at different human body parts obtained based on the PDF data. max of max mean value and over-limit probability.

[0130] The pseudo-code of the K-GRU surrogate model established by the present invention is as follows:

[0131]

[0132] The following describes the specific implementation of the present invention in detail with specific embodiments.

[0133] Embodiment 1: Human body electromagnetic exposure assessment of EV-WPT system based on K-GRU;

[0134] 1. EV-WPT system simulation model;

[0135] The EV model adopted by the present invention is as Figure 5 shown, with dimensions of 4.78m × 2.03m × 1.136m (length × width × height). The model accurately considers the geometric shape and material properties of the vehicle body. The main material of the vehicle body is aluminum, the tires are rubber, and the windows and rear engine cover are made of glass. Although these details increase the complexity of the simulation calculation, they improve the accuracy of the model.

[0136] The WPT system is as Figure 6 shown, which consists of a transmitting coil T x , a receiving coil R x , a transmitting end shielding layer and a receiving end shielding layer. d0 is the distance between the transmitting coil and the receiving coil. The ground clearance of most household cars on the market is 0.15 - 0.2m. Considering the actual use in daily life, d0 = 0.2m is taken in this embodiment. The number of turns of the transmitting coil is 9, wound in double turns, the coil cross-sectional area is 3×10 -6 m 2 , the adjacent turn spacing is 0.014m, and the overall size is 0.695 × 0.52m 2 ; the number of turns of the receiving coil is 9, wound in double turns, the coil cross-sectional area is 3×10 -6 m 2 , the adjacent turn spacing is 0.0075m, and the overall size is 0.38 × 0.38m 2; Transmitting coil T x and receiving coil R x are both made of copper. The shielding layer is composed of aluminum material and ferrite magnetic material. Due to the defect that the large-area high-permeability ferrite shielding material is fragile, the ferrite unit is designed in a small unit form. The gap between each independent ferrite unit is 0.001 m, and the size of the ferrite unit is 0.05×0.05 m 2 , and its thickness is 0.005 m. The size of the aluminum plate on the transmitting side is 0.8 m×0.65 m×0.005 m; the size of the aluminum plate on the receiving side is 0.5 m×0.5 m×0.005 m. The shielding layer structure realizes the diversion of the leaked magnetic field of WPT through the high-permeability ferrite material, and realizes electrical shielding through the eddy current effect of the aluminum plate, so as to achieve the effect of reducing the magnetic leakage in the non-working area. In addition, z0 and x0 are the displacement amounts of the transmitting coil T of the WPT system x in the transverse and longitudinal directions respectively. Since the receiving side of the WPT system is usually fixed on the ground in actual applications and the receiving end is usually embedded in the EV chassis, its position change is mainly caused by the position of the vehicle itself.

[0137] 2. Human body model;

[0138] Referring to the body parameters of adult males, this invention uses the multi-physics finite element simulation software Comsol to establish two human body models in standing and sitting postures. The models include brain tissues and are used to analyze the electromagnetic exposure situation after the implantation of intracranial artery stents, such as Figure 7 and Figure 8 shown. The height of the human body model is 1.8 m and the weight is 75 kg, which is within the height and weight range of some adult males. The research object of this invention is the EV-WPT system with a working frequency of 85 kHz. At a working frequency of 85 kHz, the conductivity of human tissues is 0.27 S / m, the relative permittivity is 5400, the thermal conductivity is 0.49 W / m / ℃, and the density is 1050 kg / m 3 ; the conductivity of the brain is 0.152 S / m, the relative permittivity is 3920, the thermal conductivity is 0.51 W / m / ℃, and the density is 1040 kg / m 3 .

[0139] According to the ICNIRP guidelines, the working frequency of the WPT system of this invention is 85 kHz (lower than 100 kHz). Therefore, the induced electric field strength (induced electric field) inside the human body is used as the evaluation index for the safety analysis of the human electromagnetic exposure of the EV-WPT system. The formula for calculating the limit value of the internal induced electric field strength for public exposure to 3 kHz - 10 MHz in the ICNIRP guidelines is shown in Equation 13:

[0140]

[0141] Among them, Elimit is the maximum value of the induced electric field strength inside the human body under public exposure in the ICNIRP guidelines, and f is the resonant frequency at which the WPT system operates.

[0142] 3. Intracranial artery stent model;

[0143] Intracranial artery stents are widely used to keep the lumen of cerebral arteries unobstructed and make the atherosclerotic blood vessel wall more stable. The intracranial artery stent established in the present invention is as Figure 9 shown. Its nominal length is 23 mm, nominal diameter is 3 mm, and nominal thickness is 0.2 mm. It is composed of 8 identical units connected together, and each unit is in the shape of a pseudo-sine wave, making the stent have excellent flexibility in the expanded state. The intracranial artery stent is made of medical-grade 316L stainless steel material with good biocompatibility and corrosion resistance, with a conductivity of 1.3×106 S / m and a relative permittivity of 1. Considering that the geometric parameters of the intracranial artery stent are small and the mechanical structure is relatively precise, when it is exposed to the leakage electromagnetic field of the EV-WPT system, the tip edge of the metal will enhance the induced electric field strength of the nearby human body. Therefore, in order to obtain more accurate numerical calculation results, the resolution of the mesh division of the intracranial artery stent in the present invention is 0.1×0.1×0.1 mm 3 .

[0144] 4. Exposure scenarios;

[0145] Considering the actual usage in life, since the metal body of an electric vehicle is not a uniform plane, when a human body is in different positions, even if the distance between the human body and the WPT system is the same, the electromagnetic exposure dose of the human body is different. Therefore, it is impossible to simply describe the positional relationship between the human body and the WPT system using a single distance. The positional relationship between the human body and the WPT system is mainly manifested as the human body standing at the tail, side of the EV, and sitting in the driver's seat in the vehicle. The present invention considers the following three exposure scenarios:

[0146] (1) The human body is inside the vehicle, as Figure 10 shown, used to simulate the scenario where the driver is waiting for charging in the driver's seat, and the driver should be in a stationary state.

[0147] (2) The human body is on the side of the vehicle, as shown in (a) of Figure 11 , used to simulate the scenario where the driver or passenger is waiting for charging on the side of the vehicle during the charging process of the vehicle. At this time, there are uncertainties in the lateral and longitudinal relative positions between the human body and the EV-WPT system and the orientation of the human body, as shown in (b) of Figure 11 .

[0148] (3) The human body is at the rear of the vehicle, as Figure 12As shown in (a), it is used to simulate the scenario where a driver or passenger waits for charging at the rear of the vehicle during the vehicle charging process. At this time, there are also uncertainties in the lateral and longitudinal relative positions of the human body with respect to the EV-WPT system and the orientation of the human body. For example, Figure 12 As shown in (b).

[0149] The above three exposure scenarios cover the exposure scenarios of people during the actual use of the EV-WPT system. Considering that factors such as the uncertainties generated during the production process of the WPT system due to manufacturing, the uncertainties caused by the misalignment between the transmitting end and the receiving end, the uncertainties of the human body position, and the front orientation of the human body will also have a great impact on the electromagnetic exposure of the human body, the present invention conducts a quantitative study on the safety uncertainties of the electromagnetic exposure of the human body with an intracranial artery stent implant in the above three exposure scenarios. The present invention considers the uncertainty factors and their distribution types and parameters in actual life applications as shown in Table 1, where U is a uniform distribution and N is a normal distribution.

[0150] Table 1 Variables for the safety assessment of human electromagnetic exposure in the EV-WPT system

[0151] Uncertainty factors Probability distribution and its parameters Unit <![CDATA[Coil pitch d0]]> U(0.15,0.25) m <![CDATA[Tx lateral offset z0]]> U(-0.2,0.2) m <![CDATA[Tx longitudinal offset x0]]> U(-0.2,0.2) m <![CDATA[Cross-sectional area s0 of the Tx coil]]> <![CDATA[N(3e -6 ,1e -7 )]]> <![CDATA[m 2 > <![CDATA[Cross-sectional area s1 of the Rx coil]]> <![CDATA[N(3e -6 ,1e -7 )]]> <![CDATA[m 2 > Lateral displacement Zs of the human body on the side of the vehicle U(-0.5,0.5) m Longitudinal displacement Xs of the human body on the side of the vehicle U(0,0.5) m Lateral displacement Zb of the human body at the rear of the vehicle U(0,0.5) m Longitudinal displacement Xb of the human body at the rear of the vehicle U(-0.5,0.5) m Front-facing direction α of the human body U(0，360) °

[0152] 5. Uncertainty assessment based on the K-GRU surrogate model;

[0153] (1) Assessment of human electromagnetic exposure under fixed conditions;

[0154] Under the three set exposure scenarios, the model established by the present invention is used to evaluate the safety of human electromagnetic exposure. During the operation of the EV-WPT system, since the vehicle chassis and body structure are not uniform, the magnetic field distribution is also non-uniform. Among them, the magnetic field intensity around the WPT system is the largest, while the magnetic field intensity inside the vehicle is very small. Therefore, as the relative position of the human body changes, the electric field intensity inside the human body will also change accordingly.

[0155] Under ideal conditions, the WPT system is not affected by uncertainty factors. Based on this condition, the electromagnetic exposure safety of the human body inside and the intracranial artery stent is evaluated under different exposure scenarios, and a comparison is made with the human body without an intracranial artery stent implant. The operating frequency of the WPT system of the present invention is 85 kHz, and the E limit of the present invention is calculated by Equation 13 to be 11.475 V / m. Without considering uncertainties, the human electromagnetic exposure dose in the electromagnetic environment generated by the EV-WPT system varies greatly under different exposure conditions. Table 2 lists the maximum values E max of the induced electric fields at different parts inside the human body under various exposure conditions.

[0156] Table 2 Maximum values of the induced electric fields in the brain and human body under different exposure scenarios

[0157]

[0158] As can be seen from Table 2, when there is an intracranial artery stent implant in the human body sitting in the vehicle, E max will exceed the standard limit of ICNIRP 2010. Due to the shielding effect of the vehicle chassis, the induced electric field of the human body with an intracranial artery stent implanted inside the vehicle is significantly weaker than that outside the vehicle. The results show that the intracranial artery stent made of medical-grade 316L stainless steel material can significantly affect the distribution of electromagnetic fields in the human body. Compared with inside the vehicle, when the human body is in two scenarios outside the vehicle, it bears a greater induced electric field. This is because the human body outside lacks the shielding effect of the vehicle bottom compared to the human body inside the vehicle and is closer to the WPT. Therefore, the relative position between the intracranial artery stent implant and the WPT system and the shielding effect of the vehicle chassis are important factors affecting the degree of human electromagnetic exposure.

[0159] (2) Assessment of human electromagnetic exposure under the influence of uncertain factors;

[0160] Use the K-GRU surrogate model to evaluate the electromagnetic exposure dose in different exposure scenarios. The dimension of the input variables changes according to different exposure scenarios. When the human body is inside and outside the vehicle, the dimensions of the input variables are 5 and 8 respectively. Take the maximum value E max of the induced electric field intensity inside the whole human body and the brain as the model output, and compare it with the human body without an intracranial artery stent implanted in different exposure scenarios. At the same time, in order to verify the effectiveness of the K-GRU method of the present invention, the model results are compared with the 10,000-time Monte Carlo (MC) method, and the probability distribution function of the maximum value E max of the induced electric field of the brain and the human body is calculated as Figure 13 、 Figure 14 and Figure 15 shown. According to the ICNIRP 2010 guidelines, the present invention uses 11.475 V / m as the electromagnetic exposure safety standard limit, and the part exceeding the standard limit is rendered in red. The horizontal axis in the figure is the value of E max , and the vertical axis PDF is the probability density function of E max in different exposure scenarios.

[0161] In terms of the calculation fitting accuracy, the K-GRU method and the MC method have good consistency. Under the condition of the same number of samples in the uncertainty surrogate model for human electromagnetic exposure safety analysis, for the same training samples, the calculation fitting accuracy of GRU is relatively poor, and the consistency is weaker than that of the K-GRU method, indicating that the ability of the K-GRU method in the data fitting task has been significantly improved. In addition, in terms of the calculation time cost, it takes 7 minutes and 25 seconds to calculate the human electromagnetic exposure using the multi-physics finite element simulation software to evaluate a set of uncertainty factors in a single scenario. The calculation time costs of the three methods are shown in Table 3:

[0162] Table 3 Calculation time costs of different methods

[0163] MC GRU K-GRU Computation time cost / min 74167 3726 3710

[0164] The data in Table 3 show that in the same scenario, the time cost of 10,000 times of the MC method is very high, while the calculation time cost of the K-GRU method is only 5% of that of the MC method. Therefore, the K-GRU method can ensure to improve the calculation efficiency while ensuring stable calculation fitting accuracy.

[0165] In addition, when considering the uncertainty of the human body position and related parameters in the EV-WPT system, the induced electric field dose inside the human body increases significantly. For the human body without an intracranial artery stent implanted, it is within the safe range in all exposure scenarios. However, for patients with an intracranial artery stent implanted, the electromagnetic exposure dose inside their bodies may exceed the ICNIRP guideline standard limit in some scenarios, and the probability of exceeding the limit is very high. The E max average value and the probability P of exceeding the standard under each scenario condition are shown in Table 4 and Table 5.

[0166] Table 4 E under different exposure scenarios max mean value

[0167]

[0168] The data in Table 4 show that whether an intracranial artery stent is implanted or not has a significant impact on human electromagnetic exposure. The human body without an intracranial artery stent implanted is safe in all exposure scenarios under the influence of uncertainty factors. However, for the human body with an intracranial artery stent implanted, there is a possibility of exceeding the limit under the influence of uncertainty factors, especially when the human body is in two scenarios outside the vehicle (the side and the rear of the vehicle). Specifically, in the in-vehicle sitting position scenario, the E max mean value of the brain and the E max mean value of the human body increase very little, while in the standing position scenarios on the side and the rear of the vehicle, the E max mean value of the brain and the E maxThe mean values all increased significantly. The E in the brain under each scenario max The mean values increased by 42.5 times, 48.8 times, and 75.9 times respectively, and the E inside the human body max The mean values increased by 3.0 times, 7.0 times, and 8.6 times respectively. In all three exposure scenarios, the human body with an intracranial artery stent exceeded the limit. The above phenomenon is because the human body in the car is protected by the shielding effect of the car chassis, and the increment of the electromagnetic exposure dose it receives is less. For the human body in the two cases of the side and the rear of the car, although there is also partial shielding by the car chassis, due to the influence of uncertain factors, the misalignment between the transmitting coil and the receiving coil and the change of the human body's own position result in a reduction in the spatial distance between the human body and the WPT system, thereby greatly increasing the possibility of being exposed to the leaked electromagnetic field of the WPT system. And due to the metal material characteristics of the intracranial artery stent, the influence on the leaked electromagnetic field is very obvious. Therefore, the electromagnetic exposure dose of the human body in the two scenarios of the side and the rear of the car is very high, and the E in the brain max The mean values reached 7.5061 V / m and 8.7391 V / m respectively, and the E of the human body max The mean values reached 30.2804 V / m and 23.2442 V / m respectively.

[0169] Table 5 E under different exposure scenarios max Probability P of exceeding the standard limit value

[0170]

[0171] The data in Table 5 show that after the intracranial artery stent is implanted in the human body, the E in the brain max is actually higher than the E of the whole human body max in terms of the probability of exceeding the limit. This is because although the human body in the car is shielded by the car chassis, the distance between the intracranial artery stent and the head from the WPT system is also weakened, and the distribution of the entire electromagnetic field is not uniform and is non-linear, making the influence of the intracranial artery stent on the brain relatively larger, resulting in a higher probability of exceeding the limit for the brain of the human body in the car compared to the probability of exceeding the limit for the whole human body. Although the E in the brain max The mean value is lower than the ICNIRP standard limit value, but from the perspective of the entire numerical probability distribution, there is still a possibility of exceeding the limit. In the above three types of exposure scenarios, the probability of exceeding the limit of the E in the brain of the human body with an intracranial artery stent max is 2.5%, 11.5%, and 12.5% respectively, while the probability of exceeding the limit of the E of the whole human body max reached 1.5%, 96%, and 71.5% respectively.

[0172] The above results further illustrate that medical implants made of metal materials will significantly change the electromagnetic field distribution inside the human body, significantly enhance the induced electric field intensity around the implant, and there is a possibility of exceeding the standard limit values of the ICNIRP guidelines. Therefore, in order to ensure human safety, patients with intracranial artery stents should pay attention to maintaining a sufficient safety distance when using the EV-WPT system. Moreover, since there is already a possibility of exceeding the standard limit values in the ICNIRP guidelines when the power of the WPT system of the present invention is 11 kW, it is also necessary to limit the power of the WPT system to reduce the risks and hazards of human electromagnetic exposure.

[0173] In summary, the present invention evaluates and analyzes the electromagnetic exposure safety issues of the human body with intracranial artery stent implants around the EV-WPT system under the influence of uncertainty factors through the K-GRU method. The results show that the K-GRU method can effectively calculate the uncertainty quantification problem of the electromagnetic exposure of the human body with intracranial artery stent implants in the EV-WPT system, providing a scientific basis for human electromagnetic exposure protection and having high engineering application value.

[0174] The above is only the preferred implementation mode of the present invention. It should be pointed out that for those skilled in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, which should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicability of the patent.

Claims

1. A method for assessing human electromagnetic exposure in a wireless charging system based on K-GRU, characterized in that: The following steps are involved: Step 1: Establish the equivalent circuit model of the MCR-WPT system; Based on the MCR-WPT system, the SS type compensation circuit structure is adopted to reduce reactive power; Step 2: Build the K-GRU proxy model; A GRU network is used as the basic architecture and a gating mechanism is introduced; the GRU network structure includes an update gate and a reset gate; KAN is introduced to improve the activation function from a fixed form to a learnable B-spline function; KAN is used to decompose high-dimensional functions into a combination of single-variable functions; Step 3: Train the K-GRU proxy model; The K-GRU proxy model is trained to obtain the uncertainty quantification proxy model of human electromagnetic exposure safety in the EV-WPT system; Step 4: Uncertainty quantification analysis; Using the Monte Carlo method to call the proxy model, we get N MC The statistical characteristic parameters of human body internal exposure dose are calculated by combining samples and corresponding output responses.

2. According to claim 1, the method for assessing human electromagnetic exposure in a wireless charging system based on K-GRU is characterized in that: In the SS type compensation circuit, U s is the input high frequency voltage source, I1 and I2 are the currents at the transmitting and receiving ends respectively, R S1 is the transmitting coil T x The equivalent resistance, C S1 is the transmitting coil T x The series compensation capacitors L1 and L2 are the transmitting coil T x and receiving coil R x The coil self-inductance, M is the transmitting coil T x With receiving coil R x The mutual inductance between the coils, C S2 The receiving coil R x The series compensation capacitor, R S2 The receiving coil R x The equivalent resistance, R L is the load resistance; According to Kirchhoff's voltage law of the circuit, the equivalent formula of the loop voltage is obtained: Formula 1: Where Z is the system impedance matrix, w is the resonant angular frequency, and j is the imaginary part of the complex number. When the WPT system undergoes series resonance, the following equation exists: Formula 2: Therefore, there exists formula 3: Formula 3: The SS compensation circuit is in a resonant state to ensure the transmission efficiency of the WPT system. The circuit parameters of the transmitter and receiver are shown in Equation 4 and Equation 5, respectively; Transmitter resonance condition: Formula 4: Resonance conditions at the receiving end: Formula 5: Among them, w is the resonant angular frequency, the value is 2πf, and f is the resonant frequency.

3. According to claim 1, the method for assessing human electromagnetic exposure in a wireless charging system based on K-GRU is characterized in that: The calculation process of the GRU is as follows: Input initialization: Assume that the input sequence is x = [x1, x2, ..., x t ], where x t is the output of the tth step; initialize the hidden state h0; Calculate the reset gate: reset gate r t The calculation formula is as follows: Formula 6: r t =σ(W r ·[h t-1 ,x t ]+b r ); Where σ is the sigmoid activation function, W r is the weight matrix of the reset gate, b r is the bias term, [h t-1 ,x t ] represents the vector concatenated from the previous hidden state and the current input; Calculate candidate hidden states: Use reset gate r t To control the hidden state h of the previous time step t-1 The influence of is as follows: Formula 7: in, is a candidate hidden state, tanh is the hyperbolic tangent function, is the weight matrix of the candidate hidden states, is the bias term; Calculate the update gate: Update gate z t The calculation formula is as follows: Equation 8: z t = σ(W z ·[h t-1 , x t + b z ); Among them, W z is the weight matrix of the update gate, b z is the bias term; Calculate the current hidden state: Use the update gate z t To combine the hidden state h of the previous time step t-1 and the current hidden state h t , the formula is as follows: Formula 9: Among them, 1-z t and z t They respectively represent the proportion of retaining the previous state and adopting the candidate hidden state decided by the update gate.

4. According to claim 1, the method for assessing human electromagnetic exposure in a wireless charging system based on K-GRU is characterized in that: The KAN achieves high-dimensional decomposition by parameterizing each one-dimensional function as a B-spline function. The p-order B-spline is recursively defined as the de Boor-cox formula as follows: Formula 10: Formula 11: Among them, N i,1 (x) is the recursive definition of the first-order B-spline, representing the i-th first-order B-spline; u is called the node vector, u i is called a node, u i+1 is the i+1th node; the half-open interval [u i ,u i+1 ) is the i-th node interval; p is the degree of the basis function, and the i-th p-th B-spline is written as N i,p (u);u i+p is the i+pth node, N i,p-1 (u) is the i-th p-1th degree B-spline, u i+p+1 is the i+p+1th node, N i+1,p-1 (u) is the i+1th p-1th degree B-spline.

5. The method for assessing human electromagnetic exposure in a wireless charging system based on K-GRU according to claim 1, characterized in that: The specific process of step 4 is as follows: Start iterative calculation to reset gate r t , update gate z t 、Current hidden state h t and candidate hidden states For each time step t from 1 to N, do the following: a. Get the input data x of the current time step t ; b. Calculate the reset gate r t ; c. Calculate candidate hidden states d. Calculate the update gate z t ; e. Save the hidden state h of the current time step t For the next iteration, h t-1 =h t ,h t is the final hidden state, containing the sequence information; Calculate using the trained K-GRU proxy model The corresponding output response Calculate Y MC The probability distribution function f y (y), and obtain the statistical characteristic parameters.

Images