A charging anti-interference method and device based on an electromagnetic environment and a storage medium

Abstract

This application discloses a charging interference immunity method, device, and storage medium based on electromagnetic environment, which is used to improve the anti-interference capability and charging performance of charging modules in complex electromagnetic environments. The method includes: collecting composite interference data of the charging module; extracting the spectral features, temporal features, and statistical features of the composite interference data, and constructing them into a high-dimensional interference feature vector and a Bayesian weight vector; constructing an interference identification model; inputting the high-dimensional interference feature vector and Bayesian weight vector into the interference identification model to obtain inference results; implementing a dynamic interference compensation strategy for the charging module based on a preset RBF mapping model; monitoring the charging module to obtain an energy conversion efficiency dataset and an electromagnetic interference immunity quantification index set; calculating the interference immunity deviation value based on the energy conversion efficiency dataset and a preset energy efficiency threshold, combined with the weighted scoring results of the electromagnetic interference immunity quantification index set; and adjusting the compensation parameters of the dynamic interference compensation strategy according to the interference immunity deviation value.

Description

Technical Field

[0001] This application relates to the field of anti-interference technology, and in particular to a charging anti-interference method, apparatus and storage medium based on electromagnetic environment. Background Technology

[0002] A charging module is a functional unit that provides power conversion and transmission services for various electrical devices in scenarios such as electric vehicle charging, energy storage system charging and discharging, and industrial equipment power supply. Specifically, it includes the power conversion unit of electric vehicle charging piles, the charging and discharging conversion module of energy storage power stations, the core control module of outdoor mobile chargers, and the power conversion unit of industrial customized power supply equipment. It undertakes functions such as accurate conversion of AC power from the grid side to DC power from the load side, dynamic power adjustment, and stable voltage and current output.

[0003] In practical application scenarios such as industrial plants, public charging stations, and outdoor energy storage stations, the electromagnetic environment in which charging modules operate is complex and variable, resulting in various types of electromagnetic interference across a wide frequency band. Existing technologies typically employ measures such as adding EMC filters, optimizing shielding structures, or strengthening grounding to suppress interference.

[0004] However, these measures are difficult to achieve ideal suppression when faced with complex and ever-changing electromagnetic interference, which in turn affects the power conversion accuracy and output stability of the charging module, resulting in a decline in charging performance. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides a charging anti-interference method, apparatus, and storage medium based on electromagnetic environment, which can improve the anti-interference capability and charging performance of charging modules in complex electromagnetic environments.

[0006] The technical solution provided in this application is described below: The first aspect of this application provides a charging interference immunity method based on an electromagnetic environment, the charging interference immunity method comprising: When the charging module starts working, composite interference data of the charging module is collected. The composite interference data is used to characterize electromagnetic interference on the power grid side, inside the charging module, and in the communication link. The spectral features, temporal features, and statistical features of the composite interference data are extracted respectively, and constructed into a high-dimensional interference feature vector and a Bayesian weight vector; An interference recognition model is constructed using the MobileNetV3-Small neural network, and the quantized and compressed interference recognition model is deployed in the charging module. The high-dimensional interference feature vector and the Bayesian weight vector are input into the interference identification model according to a preset period to obtain the inference result, which includes the interference type, interference intensity and inference confidence. When the reasoning confidence reaches a preset confidence threshold, a dynamic interference compensation strategy is executed on the charging module based on a preset RBF mapping model; After implementing the dynamic interference compensation strategy, the charging module is monitored to obtain a power conversion efficiency dataset and an electromagnetic interference quantification index set. Based on the power conversion efficiency dataset and the preset efficiency threshold, and combined with the weighted scoring results of the electromagnetic immunity quantification index set, the immunity deviation value is calculated. The compensation parameters of the dynamic interference compensation strategy are adjusted based on the anti-interference deviation value.

[0007] Optionally, the dynamic interference compensation strategy applied to the charging module based on a preset RBF mapping model includes: Based on the interference type and the interference intensity, a compensation weight matrix and an interference scene identifier are generated by calling the interference rule base of the preset RBF mapping model. Calculate the information entropy of each compensation dimension of the compensation weight matrix; Based on the information entropy of each compensation dimension, an interference suppression contribution is generated; Compensation priority weights are assigned based on the interference suppression contribution. Based on the compensation priority weight and the interference scenario identifier, a dynamic interference compensation strategy is executed through the benchmark parameter library of the RBF mapping model. The dynamic interference compensation strategy includes LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation.

[0008] Optionally, the step of executing a dynamic interference compensation strategy based on the compensation priority weight and the interference scene identifier through the baseline parameter library of the RBF mapping model includes: Based on the interference scene identifier, the corresponding benchmark parameter set is matched through the benchmark parameter library of the RBF mapping model; The compensation priority weights are sorted in descending order, and the execution order of LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy is determined according to the sorting result. According to the execution order and the reference parameter set, the LLC resonant frequency compensation, the PFC circuit coefficient compensation, and the reverse interference signal compensation are performed on the charging module in segments.

[0009] Optionally, adjusting the compensation parameters of the dynamic interference compensation strategy based on the anti-interference deviation value includes: The disturbance rejection deviation value is divided into several adjustment levels, and the initial step size coefficient is calculated according to the adjustment level. Extract the energy efficiency variation characteristics of the power conversion energy efficiency dataset, and calculate the compensation effect attenuation coefficient based on the energy efficiency variation characteristics; The initial step size coefficient is corrected by the compensation effect attenuation coefficient to obtain the final step size coefficient; The adjustment amount of the compensation parameters of the dynamic interference compensation strategy is calculated based on the final step size coefficient and the compensation priority weight. Based on the adjustment amount of the compensation parameters, the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy are iterated respectively.

[0010] Optionally, the step of constructing an interference recognition model using a MobileNetV3-Small neural network and deploying the quantized and compressed interference recognition model in the charging module includes: A simulated environment of hybrid electromagnetic interference was created using virtualization technology, and the initial parameters of the interference identification model were configured based on the MobileNetV3-Small neural network architecture. Input a test set into the interference identification model to optimize the initial parameters; The interference identification model after optimizing the initial parameters is compressed using the INT8 static quantization method and then deployed in the charging module after compilation.

[0011] Optionally, after iterating the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy according to the compensation parameter adjustment amount, the charging anti-interference method further includes: When the anti-disturbance deviation value reaches the preset deviation threshold range, a compensation correlation map is established based on the adjusted compensation parameters and the inference results; When the subsequent inference result matches the feature element in the association graph, the corresponding target compensation parameter in the association graph is directly called, and the dynamic interference compensation strategy is executed according to the target compensation parameter.

[0012] Optionally, the charging interference suppression method further includes: Real-time monitoring of the charging module's operating status parameters, including input voltage fluctuation range, output current stability, and power device temperature; When the operating status parameters exceed the preset safety threshold, the execution of the dynamic interference compensation strategy is suspended, and a preset protection anti-interference mechanism is activated. The preset protection anti-interference mechanism is used to suppress the impact of sudden strong interference on the charging module.

[0013] A second aspect of this application provides a charging interference immunity system based on an electromagnetic environment, the charging interference immunity system comprising: The acquisition unit is used to acquire composite interference data of the charging module when the charging module starts working. The composite interference data is used to characterize electromagnetic interference on the power grid side, inside the charging module, and in the communication link. The extraction unit is used to extract the spectral features, temporal features, and statistical features of the composite interference data, and construct them into a high-dimensional interference feature vector and a Bayesian weight vector. The building unit is used to build an interference recognition model through the MobileNetV3-Small neural network and deploy the quantized and compressed interference recognition model in the charging module; The inference unit is used to input the high-dimensional interference feature vector and the Bayesian weight vector into the interference identification model according to a preset period to obtain the inference result, which includes the interference type, interference intensity and inference confidence. The compensation unit is used to perform a dynamic interference compensation strategy on the charging module based on a preset RBF mapping model when the reasoning confidence reaches a preset confidence threshold. The monitoring unit is used to monitor the charging module after the dynamic interference compensation strategy is executed, and to obtain the power conversion efficiency dataset and the electromagnetic interference quantification index set. The calculation unit is used to calculate the disturbance rejection deviation value based on the power conversion efficiency dataset and the preset efficiency threshold, and in combination with the weighted scoring results of the electromagnetic disturbance rejection quantification index set. The adjustment unit is used to adjust the compensation parameters of the dynamic interference compensation strategy according to the anti-interference deviation value.

[0014] Optionally, the compensation unit is specifically used for: Based on the interference type and the interference intensity, a compensation weight matrix and an interference scene identifier are generated by calling the interference rule base of the preset RBF mapping model. Calculate the information entropy of each compensation dimension of the compensation weight matrix; Based on the information entropy of each compensation dimension, an interference suppression contribution is generated; Compensation priority weights are assigned based on the interference suppression contribution. Based on the compensation priority weight and the interference scenario identifier, a dynamic interference compensation strategy is executed through the benchmark parameter library of the RBF mapping model. The dynamic interference compensation strategy includes LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation.

[0015] Optionally, the compensation unit is specifically used for: The step of executing a dynamic interference compensation strategy based on the compensation priority weight and the interference scene identifier through the baseline parameter library of the RBF mapping model includes: Based on the interference scene identifier, the corresponding benchmark parameter set is matched through the benchmark parameter library of the RBF mapping model; The compensation priority weights are sorted in descending order, and the execution order of LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy is determined according to the sorting result. According to the execution order and the reference parameter set, the LLC resonant frequency compensation, the PFC circuit coefficient compensation, and the reverse interference signal compensation are performed on the charging module in segments.

[0016] Optionally, the adjustment unit is specifically used for: The disturbance rejection deviation value is divided into several adjustment levels, and the initial step size coefficient is calculated according to the adjustment level. Extract the energy efficiency variation characteristics of the power conversion energy efficiency dataset, and calculate the compensation effect attenuation coefficient based on the energy efficiency variation characteristics; The initial step size coefficient is corrected by the compensation effect attenuation coefficient to obtain the final step size coefficient; The adjustment amount of the compensation parameters of the dynamic interference compensation strategy is calculated based on the final step size coefficient and the compensation priority weight. Based on the adjustment amount of the compensation parameters, the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy are iterated respectively.

[0017] Optionally, the building unit is specifically used for: A simulated environment of hybrid electromagnetic interference was created using virtualization technology, and the initial parameters of the interference identification model were configured based on the MobileNetV3-Small neural network architecture. Input a test set into the interference identification model to optimize the initial parameters; The interference identification model after optimizing the initial parameters is compressed using the INT8 static quantization method and then deployed in the charging module after compilation.

[0018] Optionally, it also includes the calling unit, specifically used for: When the anti-disturbance deviation value reaches the preset deviation threshold range, a compensation correlation map is established based on the adjusted compensation parameters and the inference results; When the subsequent inference result matches the feature element in the association graph, the corresponding target compensation parameter in the association graph is directly called, and the dynamic interference compensation strategy is executed according to the target compensation parameter.

[0019] Optionally, the protection unit is also included, specifically for: Real-time monitoring of the charging module's operating status parameters, including input voltage fluctuation range, output current stability, and power device temperature; When the operating status parameters exceed the preset safety threshold, the execution of the dynamic interference compensation strategy is suspended, and a preset protection anti-interference mechanism is activated. The preset protection anti-interference mechanism is used to suppress the impact of sudden strong interference on the charging module.

[0020] A third aspect of this application provides a charging interference suppression device based on an electromagnetic environment, the charging interference suppression device comprising: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to execute the first aspect and any one of the optional methods in the first aspect.

[0021] A fourth aspect of this application provides a computer-readable storage medium on which a program is stored, which, when executed on a computer, performs the methods of the first aspect and any one of the first aspects.

[0022] As can be seen from the above technical solutions, this application has the following beneficial effects: 1. Based on the complex and ever-changing electromagnetic environment in which the charging module is located, the interference identification model is constructed to infer the type, intensity and reliability of the interference faced by the current charging module. Then, a dynamic interference compensation strategy is executed based on these inference results to realize the perception and quantitative evaluation of multi-source composite interference, thereby enabling the dynamic interference compensation strategy to improve the anti-interference capability of the charging module in a targeted manner. 2. When implementing the dynamic interference compensation strategy, the dynamic interference compensation strategy is executed based on the interference identification results through a preset RBF mapping model. The compensation parameters of the dynamic interference compensation strategy are adjusted according to the interference deviation formed by the power conversion efficiency and electromagnetic interference immunity quantification index. This allows the dynamic interference compensation strategy to match the best processing scheme according to the current interference scenario and to be optimized and adjusted according to the dynamic changes of interference. This forms an anti-interference scheme that includes identification, matching, execution, and optimization, thereby improving the anti-interference capability and charging performance of the charging module in complex electromagnetic environments. 3. Meanwhile, the interference identification model is built using the MobileNetV3-Small network architecture, which is lightweight and can adapt to the limited storage resources of the charging module. While ensuring the accuracy of interference identification and the efficiency of inference, it reduces the resource consumption of the charging module by the complex model. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic flowchart of an embodiment of the charging interference immunity method based on electromagnetic environment provided in this application; Figure 2 A schematic flowchart of another embodiment of the charging interference immunity method based on electromagnetic environment provided in this application; Figure 3 A schematic diagram of an embodiment of the charging interference suppression system based on the electromagnetic environment provided in this application; Figure 4 This is a schematic diagram of an embodiment of the charging interference suppression device based on the electromagnetic environment provided in this application. Detailed Implementation

[0025] This application provides a charging interference immunity method, apparatus, and storage medium based on electromagnetic environment, which improves the interference immunity and charging performance of charging modules in complex electromagnetic environments. It should be noted that the charging interference immunity method based on electromagnetic environment provided in this application is applicable to charging terminals located in complex electromagnetic environments, such as charging terminals in industrial parks and automotive charging terminals.

[0026] Please see Figure 1 This application first provides an embodiment of a charging immunity method in an electromagnetic environment, which includes: S101. When the charging module starts working, collect the composite interference data of the charging module. The composite interference data is used to characterize the electromagnetic interference on the power grid side, inside the charging module, and in the communication link. Charging modules are typically used in industrial parks, parking lots, and other similar environments where numerous industrial devices (such as motors, high-frequency heaters, and radio frequency equipment) and power electronic devices exist. These devices generate electromagnetic radiation during operation, which radiates into space as electromagnetic waves. This radiation penetrates the charging module's casing or couples to its internal circuitry via cables. Simultaneously, electromagnetic signals generated by fluctuations in the charging module's internal current and grid voltage also couple to the module, causing signal distortion in the communication link or control circuitry. Specifically, at the moment the charging module starts operating, initial startup radiated interference from surrounding equipment and harmonic interference caused by inrush current from the grid simultaneously couple to the module. The instantaneous switching noise of internal power devices during startup also adds to the external interference, making the charging module susceptible to interference. Therefore, it is necessary to collect composite interference data from the moment the charging module starts operating to ensure the accuracy and completeness of the data collected in this complex electromagnetic environment, facilitating subsequent improvements to the charging module's immunity to interference. The collected composite interference data includes data from three aspects: the power grid side, the charging module internals, and the communication link. For example, data collected from the power grid side includes voltage sag and swell waveforms, voltage flicker parameters, and coupled interference data resulting from the superposition of power frequency electromagnetic fields and the radiation fields of surrounding equipment; switching noise amplitudes of power devices inside the charging module, resonant frequencies of the LLC resonant cavity, and transformer leakage flux radiation field strength; and signal amplitude jitter, phase shift, inter-symbol interference, and electromagnetic coupling coefficients between communication cables and power cables at the charging module interface on the communication link side. Composite interference data is multi-source and multi-frequency, including both steady-state and transient interference data. Therefore, after collecting the composite interference data, data preprocessing such as data cleaning, noise suppression, and data normalization is required to eliminate noise, redundancy, and format differences, ensuring the accuracy and effectiveness of subsequent feature extraction.

[0027] S102. Extract the spectral features, temporal features, and statistical features of the composite interference data respectively, and construct them into a high-dimensional interference feature vector and a Bayesian weight vector. After preprocessing the composite interference data, it is necessary to extract its spectral features, time-domain features, and statistical features to provide identifiable data for subsequent interference identification model inference. When extracting spectral features, Fourier Transform (FFT) is used to transform the composite interference data in the frequency domain, decomposing the continuous time-domain signal into discrete frequency components. For example, after obtaining the frequency-amplitude distribution spectrum from the grid-side interference data using FFT, the amplitudes corresponding to the 2nd to 50th harmonics and the power spectral density of each harmonic are extracted, and the total harmonic distortion rate is calculated according to the formula. ; in, For the first The effective value of the subharmonics The system extracts the fundamental RMS value and locates the characteristic frequency peaks and corresponding frequency points formed by harmonic superposition (such as the peak frequencies of the 3rd and 5th dominant harmonics). For interference data inside the charging module, segmented extraction is used, with FFT analysis performed on the high-frequency band of 1MHz-100MHz to extract the frequency and amplitude corresponding to the spectral peak of the switching noise of the power switching device. The resonant frequency offset and resonant peak amplitude fluctuation of the LLC resonant cavity are calculated using a resonant frequency fitting algorithm. For communication link interference data, FFT transformation is mainly performed on the communication signal operating frequency band and surrounding interference frequency bands (the 250kbps band corresponding to the CAN bus and the 2.4GHz / 5.8GHz RF band) to extract the characteristic frequency of RF coupling interference, the amplitude of spectral spurious components, and the spectral broadening coefficient. Time-domain feature extraction is based on the preprocessed time-domain raw waveform data, obtaining key time-domain information through feature point detection and parameter calculation. Statistical feature extraction involves statistical analysis of spectral and time-domain features, calculating the mean, variance, standard deviation, skewness, kurtosis, and coefficient of variation of each feature.

[0028] When constructing the high-dimensional interference feature vector, spectral features, temporal features, and statistical features are extracted and constructed in the order of "spectral features - temporal features - statistical features." Each element in the vector corresponds to a quantized value of a specific feature, covering the multi-dimensional attributes of the interference. When constructing the Bayesian weight vector, the weight coefficients are determined through statistical analysis based on the actual contribution of each feature to the differentiation of interference types under complex electromagnetic environments. For example, high-discrimination key features such as the amplitude of the 3rd / 5th harmonics, the center frequency of switching noise, and the range of communication amplitude jitter are assigned high weight coefficients of 0.8-1.0; while redundant features such as small variance fluctuations in steady-state interference and low-amplitude spectral spurious components are assigned low weight coefficients of 0.1-0.3. It should be noted that during the construction of the Bayesian weight vector, the sum of the weight coefficients of all features must be normalized to 1 to form a Bayesian weight vector with the same dimension as the high-dimensional interference feature vector.

[0029] S103. An interference recognition model is constructed using the MobileNetV3-Small neural network, and the quantized and compressed interference recognition model is deployed in the charging module. The interference identification model employs a three-layer structure: a feature extraction layer, a bottleneck layer, and a classification layer. The feature extraction layer utilizes depthwise separable convolution, splitting standard convolution into depthwise convolution and pointwise convolution. Depthwise convolution groups the input high-dimensional interference feature vectors according to their dimensions, enabling independent feature extraction for each dimension. Pointwise convolution, on the other hand, performs channel fusion on the output of the depthwise convolution, mapping multi-dimensional features to preset high-dimensional feature channels, reducing computational cost while preserving the correlation of interference features. The bottleneck layer uses the inverted residual block structure unique to MobileNetV3-Small and embeds a squeeze-and-excitation attention mechanism, which enhances the expression of key features while preventing model overfitting. The classification layer uses a global average pooling layer to convert the feature map output by the bottleneck layer into a one-dimensional feature vector. This vector is then mapped to a predefined interference type classification space (e.g., power grid harmonic interference, switching noise interference, and radio frequency coupling interference) through a fully connected layer. Finally, a softmax activation function is used to output the recognition probability of each type of interference. To adapt to the hardware resource limitations of the charging module embedded system, the interference recognition model undergoes lightweighting after construction to reduce its size and storage and computing resource consumption. The model is then compiled into an executable file supported by the charging module embedded system and burned into the charging module's memory. After deployment, the interference recognition model can directly read the high-dimensional interference feature vector and Bayesian weight vector for inference.

[0030] S104. Input the high-dimensional interference feature vector and Bayesian weight vector into the interference identification model according to the preset period to obtain the inference result, which includes the interference type, interference intensity and inference credibility. Within each preset cycle, the charging module's control unit synchronously inputs the constructed high-dimensional interference feature vector and Bayesian weight vector into the quantized and compressed interference identification model deployed in the charging module's memory for inference, providing a quantized basis for subsequent interference compensation. The interference identification model enhances the high-dimensional interference feature vector through a feature weighting layer combined with Bayesian weight vectors. For example, it increases the response weight of key features with high weight coefficients, such as the amplitude of the 3rd / 5th harmonics, the center frequency of switching noise, and the range of communication amplitude jitter, while weakening the influence of low-weight redundant features, ensuring the model focuses on core interference information for inference. After sequential processing by the feature extraction layer, bottleneck layer, and classification layer of the interference identification model, the inference result is output based on the mapping relationship between interference features and interference types. The output inference results include interference type, interference intensity, and inference reliability. The interference type is specifically classified into predefined categories such as grid harmonic interference, switch noise interference, radio frequency coupling interference, and cable coupling interference, used to pinpoint the specific source of the current dominant interference. Interference intensity is determined by comparing the model's output feature quantization values ​​with predefined intensity grading thresholds (feature amplitude ranges corresponding to mild, moderate, and severe interference). For example, when the total harmonic distortion (THD) on the grid side is ≥10% or the communication link bit error rate is ≥5%, it is judged as severe interference. Inference reliability is represented by the probability value of the corresponding interference type output by the Softmax activation function of the model's classification layer, with a value range of [0,1]. The closer the probability value is to 1, the higher the reliability of the inference result. After each inference, the interference identification model transmits the inference results to the charging module's control unit in real time for compensation judgment.

[0031] S105. When the reasoning credibility reaches the preset credibility threshold, a dynamic interference compensation strategy is executed on the charging module based on the preset RBF mapping model. The inference results of the interference identification model have varying confidence levels. If compensation is directly performed based on low-confidence inference results, errors in judging the interference type and intensity may lead to a mismatch between the dynamic interference compensation strategy and the actual interference. For example, misjudging grid harmonic interference as radio frequency interference and blindly initiating reverse signal compensation could not only fail to suppress the interference but also introduce new circuit parameter conflicts. Therefore, the dynamic interference compensation strategy should only be activated when the interference identification model has sufficient reliability in judging the interference type and intensity. For instance, when the inference confidence level is 0.8 and the value falls between 0.8 and 1 (a preset confidence threshold), the dynamic interference compensation strategy is executed. The dynamic interference compensation strategy, based on real-time identified interference type, intensity, and charging module operating conditions, uses adaptive adjustments to circuit parameters, generation of reverse cancellation signals, and optimization of electromagnetic isolation configuration to specifically suppress or cancel electromagnetic interference affecting the charging module's operation, achieving on-demand compensation and thus improving the charging module's anti-interference capability.

[0032] S106. After implementing the dynamic interference compensation strategy, monitor the charging module to obtain the power conversion efficiency dataset and the electromagnetic immunity quantification index set. After the dynamic interference compensation strategy is implemented, the charging module's monitoring system is immediately activated. Within a preset monitoring period (e.g., 20ms), it continuously collects the charging module's operating status parameters, such as input-side power parameters, output-side power parameters, grid-side harmonic suppression indicators, and internal electromagnetic interference indicators of the charging module. The power conversion efficiency is calculated using the input and output power parameters. The proportions of copper loss, iron loss, and switching loss are then analyzed based on the charging module's operating conditions. Finally, the original input and output parameters, conversion efficiency, and energy consumption data are integrated by timestamp to form a complete power conversion efficiency dataset. The grid-side harmonic suppression indicators include total harmonic distortion (THD) and the proportion of major harmonic components, which can reflect the effectiveness of grid-side interference control. The charging module's internal electromagnetic interference indicators include switching noise intensity and internal electromagnetic radiation level, which are used to represent the charging module's internal interference suppression capability. At the same time, output terminal operation indicators including output voltage ripple and output current fluctuation amplitude, and communication link immunity indicators including signal amplitude stability, phase offset degree, and bit error rate are collected. The grid-side harmonic suppression indicators, charging module internal electromagnetic interference indicators, output terminal operation indicators, and communication link immunity indicators are classified and integrated according to monitoring timestamps to form a structured set of electromagnetic immunity quantitative indicators.

[0033] S107. Based on the power conversion efficiency dataset and the preset efficiency threshold, and combined with the weighted scoring results of the electromagnetic immunity quantification index set, calculate the immunity deviation value. The disturbance rejection deviation value is used to quantify the degree of deviation between the actual effect of the dynamic interference compensation strategy and the expected target, providing a quantitative basis for subsequent adjustment of compensation parameters. The calculation of the disturbance rejection deviation value first retrieves the preset energy efficiency threshold stored in the charging module control unit and calculates the energy efficiency deviation component according to the formula: ; in, This indicates the preset power conversion efficiency threshold. This represents the actual conversion efficiency, and the deviation is calculated only when the actual efficiency is below a threshold, divided by... Eliminate dimensions to achieve normalization; This indicates the preset total energy consumption threshold. This represents the actual total energy consumption and loss. This indicates the preset threshold for the proportion of switching losses. This indicates the percentage of actual switching losses. , , This represents the weighting coefficient, with a total weight of 1. This represents the energy efficiency deviation component, with a value ranging from 0 to 1. The larger the value of the energy efficiency deviation component, the more serious the deviation from the expected energy efficiency.

[0034] Then, a weighted score is calculated on the quantitative index set of electromagnetic immunity, and the score results are converted into electromagnetic immunity deviation components. The formula is: ; in, This represents the total harmonic distortion (THD) score on the power grid side. This indicates the score for internal switch noise intensity. Indicates the communication bit error rate score. This indicates the output voltage ripple coefficient score; The indicator weights are allocated according to their impact on module stability. , , , This represents the electromagnetic immunity deviation component, with a value ranging from 0 to 1. The larger the value, the worse the electromagnetic immunity effect.

[0035] Finally, the immunity deviation value is calculated based on the energy efficiency deviation component and the electromagnetic immunity deviation component, using the following formula: ; in, This indicates the overall weight of the energy efficiency dimension (e.g., 0.4). This indicates the overall weight of the disturbance resistance dimension (e.g., 0.6), which can be adjusted according to the application scenario. This represents the disturbance rejection deviation value.

[0036] S108. Adjust the compensation parameters of the dynamic interference compensation strategy according to the anti-interference deviation value.

[0037] The initial parameters of the dynamic interference compensation strategy are generated based on the benchmark parameter library of the preset RBF mapping model. They can only adapt to the initial state of a specific interference scenario. However, during the actual operation of the charging module, both the electromagnetic environment and the operating conditions of the charging module itself are dynamically changing. The initial parameters are difficult to continuously match the dynamic changes. Long-term operation is prone to problems such as the decay of interference suppression effect and the decrease of energy conversion efficiency. Therefore, it is necessary to adjust the parameters in a targeted manner so that the dynamic interference compensation strategy always adapts to the actual operating requirements. Since the dynamic interference compensation strategy covers multi-dimensional compensation for the electromagnetic interference of the charging module, and the anti-interference deviation value can clearly define the degree of deviation in each dimension, adjusting the compensation parameters based on the anti-interference deviation value can specifically adjust the specific compensation parameters of each dimension in the dynamic interference compensation strategy.

[0038] In this embodiment, given that the charging module is in a complex and variable electromagnetic environment, an interference identification model is constructed to infer the type, intensity, and reliability of interference currently faced by the charging module. Based on these inference results, a dynamic interference compensation strategy is executed to achieve the perception and quantitative assessment of multi-source composite interference. This allows the dynamic interference compensation strategy to specifically improve the charging module's anti-interference capability. When executing the dynamic interference compensation strategy, the interference identification results are used as a basis, and a preset RBF mapping model is used to execute the dynamic interference compensation strategy. An anti-interference deviation is formed based on the energy conversion efficiency and electromagnetic interference immunity quantification indicators, and the compensation parameters of the dynamic interference compensation strategy are adjusted. This allows the dynamic interference compensation strategy to match the best processing scheme according to the current interference scenario and to be optimized and adjusted according to dynamic changes in interference, thus forming an anti-interference scheme that includes identification, matching, execution, and optimization. This improves the charging module's anti-interference capability and charging performance in complex electromagnetic environments. Simultaneously, the interference identification model is constructed using the MobileNetV3-Small network architecture, which is lightweight and can adapt to the limited storage resources of the charging module. This reduces the resource consumption of the complex model on the charging module while ensuring interference identification accuracy and inference efficiency.

[0039] Please see Figure 2 This application provides another embodiment of a charging immunity method for electromagnetic environments, which includes: S201. When the charging module starts working, collect the composite interference data of the charging module. The composite interference data is used to characterize the electromagnetic interference on the power grid side, inside the charging module, and in the communication link. S202. Extract the spectral features, temporal features, and statistical features of the composite interference data respectively, and construct them into a high-dimensional interference feature vector and a Bayesian weight vector. Steps S201 and S202 are similar to steps S101 and S102 described above, and will not be repeated here.

[0040] S203. Use virtualization technology to create a simulated environment of hybrid electromagnetic interference, and configure the initial parameters of the interference identification model based on the MobileNetV3-Small neural network architecture. When constructing the interference identification model, virtualization simulation technology is used to build a mixed electromagnetic interference simulation environment that includes multiple types of interference such as power grid harmonic interference, radio frequency coupling interference, and switching noise interference. The intensity, frequency, and superposition method of each interference type can be flexibly adjusted to simulate the complex electromagnetic scenarios that the charging module may encounter in actual operation. The MobileNetV3-Small neural network architecture, with its lightweight and low computational consumption characteristics, is used to construct the interference identification model, better adapting to the operational requirements of the embedded hardware of the charging module. After completing the interference identification model construction, initial parameters need to be configured according to the interference identification model structure, including convolutional layer weights, batch normalization parameters, and activation function coefficients. The convolutional layer weights are initialized using a Xavier uniform distribution, and the batch normalization parameters are initialized with a mean of 0 and a variance of 1, ensuring that the model's initial state has basic feature extraction and classification capabilities.

[0041] S204. Input the test set into the interference identification model to optimize the initial parameters; The test set is used to optimize the initial parameters of the interference recognition model, ensuring its performance is at its best and improving its inference efficiency in practical applications. The test set data originates from various interference signal characteristics (such as time-domain waveforms, frequency-domain spectra, and harmonic component proportions) collected in a mixed electromagnetic interference simulation environment, with each interference type clearly labeled. The test set is divided into a training set and a validation set at a preset ratio (e.g., 7:3), and input into the interference recognition model for iterative training until the model's recognition performance on the validation set stabilizes and meets the target, yielding the optimized model parameters.

[0042] S205 uses the INT8 static quantization method to compress the interference identification model after optimizing the initial parameters, and then deploys it in the charging module after compilation.

[0043] To address the storage and computing power limitations of the embedded hardware in the charging module, this application employs the INT8 static quantization method to compress the optimized interference identification model. By statistically analyzing the value range and distribution characteristics of the parameters at each layer of the interference identification model, the quantization scaling factor and offset are determined. Simultaneously, the parameter format of the interference identification model is converted; for example, 32-bit floating-point parameters are converted to 8-bit integer parameters. This significantly reduces the storage capacity and computational complexity of the interference identification model while maintaining its accuracy. Finally, the quantized interference identification model is compiled for compatibility, generating an executable file adapted to the embedded processor of the charging module. The compiled model is then deployed to the control unit of the charging module via a communication interface, providing algorithmic support for subsequent real-time interference identification.

[0044] S206. Input the high-dimensional interference feature vector and Bayesian weight vector into the interference identification model according to the preset period to obtain the inference result, which includes the interference type, interference intensity and inference credibility. Step S206 is similar to step S104 above, and will not be described again here.

[0045] S207. Based on the interference type and interference intensity, call the interference rule base of the preset RBF mapping model to generate a compensation weight matrix and interference scene identifier; When the reasoning confidence level reaches a preset confidence threshold, a dynamic interference compensation strategy needs to be executed to improve the charging module's anti-interference capability. When executing the dynamic interference compensation strategy, it is necessary to adapt it specifically to the current interference scenario. After obtaining the interference type and corresponding interference intensity quantification value obtained from the interference identification model, the interference rule library in the preset RBF mapping model is called. This interference rule library is built based on a large amount of measured data and simulation verification, storing association rules for different interference types, intensities, and compensation dimensions. Based on the current interference type and intensity, the corresponding association logic in the rule library is matched to generate a compensation weight matrix, and a unique interference scenario identifier is generated simultaneously.

[0046] S208. Calculate the information entropy of each compensation dimension of the compensation weight matrix; For the generated compensation weight matrix, extract the initial weight values ​​corresponding to each compensation dimension, and calculate the information entropy of each compensation dimension according to the information entropy formula. : ; in, Statement No. The initial weighting of each compensation dimension. The total number of compensation dimensions. The information entropy value reflects the uncertainty of the weight distribution of the compensation dimension. The smaller the entropy value, the stronger the targeting of the dimension to the current disturbance; the larger the entropy value, the more dispersed the weight distribution and the weaker the targeting.

[0047] S209. Generate the interference suppression contribution based on the information entropy of each compensation dimension; After obtaining the information entropy of each compensation dimension, it is necessary to convert the information entropy of each compensation dimension into an interference suppression contribution value through normalization. First, the information entropy of all compensation dimensions is summed to obtain the total entropy value. Then, based on the ratio of the reciprocal of the information entropy of each dimension to the reciprocal of the total entropy value, normalization calibration is performed to ensure that the sum of the interference suppression contributions of all dimensions is 1. The interference suppression contribution value is negatively correlated with the information entropy. The smaller the information entropy, the larger the contribution value, indicating that the compensation dimension has a more significant effect on suppressing the current interference, and vice versa.

[0048] S210. Assign compensation priority weights based on the contribution of interference suppression; The interference suppression contribution of each compensation dimension is directly mapped to the compensation priority weight. The contribution value is the priority weight ratio of the corresponding dimension, ensuring that the compensation dimension with the higher contribution value gets the higher priority weight and is adjusted first.

[0049] In one specific embodiment, since the charging module is in a complex electromagnetic environment, it may be subject to interference from three aspects: grid-side interference, internal interference within the charging module, and communication link interference. The dynamic interference compensation strategy aims to improve the charging module's anti-interference capability in a coordinated manner based on these three aspects of interference. Specifically, based on the compensation priority weight and interference scenario identifier, the dynamic interference compensation strategy is executed through the reference parameter library of the RBF mapping model. The dynamic interference compensation strategy includes LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation. The following steps S211 to S213 will be explained in detail: S211. Based on the interference scene identifier, match the corresponding reference parameter set through the reference parameter library of the RBF mapping model; The charging module control unit retrieves a reference parameter library pre-set in the RBF mapping model based on the generated interference scenario identifier. This reference parameter library contains structured storage of empirically optimal parameter ranges for three compensation methods: LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation. By matching the interference scenario identifier with the reference parameter library, the corresponding reference parameter set for the interference scenario is extracted, including parameters such as the adjustment range of the LLC resonant frequency, the reference value of the PFC circuit power factor correction coefficient, and the amplitude, phase, and frequency tracking rules of the reverse interference signal.

[0050] S212. Sort the compensation priority weights in descending order, and determine the execution order of LLC resonant frequency compensation, PFC circuit coefficient compensation and reverse interference signal compensation in the dynamic interference compensation strategy according to the sorting results. The compensation priority weights are sorted in descending order. The sorting result corresponds to the execution order of the dynamic interference compensation strategy. The compensation dimension with the highest priority weight is executed first, followed by the next highest weight, and the lowest weight is executed last. For example, if the sorting result shows that PFC circuit coefficient compensation has the highest weight, LLC resonant frequency compensation has the second highest weight, and reverse interference signal compensation has the lowest weight, then PFC circuit coefficient compensation is executed first, followed by LLC resonant frequency compensation, and finally reverse interference signal compensation.

[0051] S213. According to the execution order and reference parameter set, perform LLC resonant frequency compensation, PFC circuit coefficient compensation and reverse interference signal compensation on the charging module in segments.

[0052] Taking the sorting result in step S212 as an example, the dynamic interference compensation strategy is executed in three stages according to the sorting result. First, PFC circuit coefficient compensation is performed. Based on the power factor correction coefficient range in the reference parameter set, the duty cycle of the switching transistors and the PI adjustment parameters of the PFCBoost converter are adjusted to optimize the power factor correction coefficient, enhance the suppression capability of dominant harmonics on the grid side, and improve the input power utilization rate. During the execution of PFC circuit coefficient compensation, the total harmonic distortion (THD) and power factor of the grid input current are monitored in real time. When the THD drops below the preset value, the compensation for this stage is confirmed to be complete. Subsequently, based on the stability of PFC circuit compensation, LLC resonant frequency compensation is performed. According to the resonant frequency adjustment range in the reference parameters, the inductor and capacitor matching parameters of the LLC resonant converter are finely adjusted to optimize the default value of the resonant frequency, thereby improving the impedance characteristics and filtering effect of the resonant cavity, while reducing the additional noise generated by the high-frequency switching of the switching transistors. Finally, for any residual harmonic interference not completely eliminated in the first two stages, reverse interference signal compensation is performed. Based on reference parameters, the built-in signal generator is activated to generate a reverse compensation signal with an amplitude 1.2 times that of the interference signal, an opposite phase, and a frequency that tracks the dominant harmonic frequency. This signal is then injected into the power circuit of the charging module through a coupling coil. The three compensation stages are performed sequentially to reduce circuit parameter disturbances caused by simultaneous adjustments in multiple dimensions.

[0053] S214. After implementing the dynamic interference compensation strategy, monitor the charging module to obtain the power conversion efficiency dataset and the electromagnetic immunity quantification index set. S215. Based on the power conversion efficiency dataset and the preset efficiency threshold, and combined with the weighted scoring results of the electromagnetic immunity quantification index set, calculate the immunity deviation value. Steps S214 and S215 are similar to steps S106 and S107 above, and will not be repeated here.

[0054] S216. Divide the disturbance rejection deviation value into several adjustment levels, and calculate the initial step size coefficient according to the adjustment level; To improve the efficiency of adjusting compensation parameters, several adjustment levels (e.g., three levels) are defined based on the disturbance rejection deviation value. These levels represent different degrees of deviation, and the initial step size coefficient is positively correlated with the adjustment level of the disturbance rejection deviation value. The more severe the deviation, the larger the base magnitude of the parameter adjustment, requiring a rapid reduction in the gap between the compensation effect and the expected target. Dividing the disturbance rejection deviation value into several adjustment levels allows for a quick determination of the required adjustment magnitude and the corresponding adjustment operation.

[0055] S217. Extract the energy efficiency change characteristics of the power conversion energy efficiency dataset, and calculate the compensation effect attenuation coefficient based on the energy efficiency change characteristics; The energy efficiency variation features extracted from the power conversion efficiency dataset include the fluctuation trend of power conversion efficiency, the rate of change of total energy consumption, the fluctuation amplitude of the switching loss ratio, and the stability parameters of energy efficiency indicators. The compensated power conversion efficiency is calculated using these energy efficiency variation features. Then, based on three types of features—the iterative change rate of power conversion efficiency, the relative deviation of total energy consumption, and the stability coefficient of the switching loss ratio—different weights are assigned to each (the sum of the weights is 1, which can be dynamically adjusted according to the interference scenario), and the results are summed to obtain a normalized result. Finally, this result is mapped to a specific compensation effect attenuation coefficient value. A lower compensation effect attenuation coefficient indicates a more stable compensation effect, while a higher coefficient indicates a decreased compensation effect. When the compensation effect decreases, the initial step size coefficient needs to be adjusted.

[0056] Normalization of various energy efficiency variation characteristics The formula is: ; in, for Extracted values ​​of energy efficiency change characteristics. for The theoretical minimum value of a class feature (e.g., fluctuation range of 0%). for The theoretical maximum value of a characteristic (such as a fluctuation range of 5%) needs to be set according to the actual operating conditions of the charging module. Compensation effect attenuation coefficient The formula is: ; in, The assigned weights can be dynamically adjusted.

[0057] S218. Correct the initial step size coefficient using the compensation effect attenuation coefficient to obtain the final step size coefficient; The initial step size coefficient is adjusted according to the step size formula: ; in, This represents the final step size coefficient. This represents the initial step size coefficient. This represents the attenuation coefficient of the compensation effect.

[0058] S219. Calculate the adjustment amount of the compensation parameters for the dynamic interference compensation strategy based on the final step size coefficient and the compensation priority weight; The adjustment amount of the compensation parameter is calculated according to the formula: ; in, Indicates the first Adjustment amount for each compensation dimension Indicates the first Compensation priority weights in the compensation dimension.

[0059] S220. Based on the adjustment amount of the compensation parameters, iterate the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy respectively.

[0060] Based on the parameter adjustment amounts for each compensation dimension, the specific parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation are iteratively optimized. For example, if the current value of the PFC circuit power factor correction coefficient is 0.9 and the adjustment amount is 0.04, the parameter will be updated to 0.94 after iteration, thereby improving harmonic suppression capability and power utilization. During the iteration process, parameters are updated sequentially according to the compensation priority weight from high to low. After each iteration of a compensation dimension is completed, the module's operating parameters (such as harmonic distortion rate, conversion efficiency, and ripple coefficient) are collected in real time to ensure that parameter adjustments do not cause operational abnormalities, thus optimizing the dynamic interference compensation strategy and adapting to the dynamic changes in the electromagnetic environment and charging module operating conditions.

[0061] S221. When the anti-disturbance deviation value reaches the preset deviation threshold range, a compensation correlation map is established based on the adjusted compensation parameters and inference results. After iterative optimization of the compensation parameters, the anti-interference deviation value is recalculated. When the anti-interference deviation value reaches the preset deviation threshold range, it is determined that the current compensation parameters have reached the optimal state for adapting to the current interference scenario. At this point, based on the adjusted optimal compensation parameters (including the specific parameter values ​​for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation) and the corresponding interference inference results, a compensation correlation map is constructed. The interference feature combinations in the compensation correlation map are associated with the optimal compensation parameters. Among them, the interference feature combinations include key feature elements such as interference type, intensity level, and electromagnetic environment scenario label.

[0062] S222. When the subsequent inference result matches the feature element in the association graph, the corresponding target compensation parameter in the association graph is directly called, and the dynamic interference compensation strategy is executed according to the target compensation parameter.

[0063] During subsequent operation of the charging module, after completing new inference through the interference identification model, the new inference result is matched with interference feature elements in the compensation correlation graph in multiple dimensions. During the matching process, a feature similarity algorithm is used to accurately compare core features such as interference type and intensity level. If the similarity reaches a preset matching threshold (e.g., above 95%), the match is considered successful. At this point, the corresponding target compensation parameter set can be directly retrieved from the correlation graph without repeatedly executing the parameter iteration optimization process, quickly initiating the dynamic interference compensation strategy. If no corresponding feature element is matched, the process returns to the parameter iteration optimization process, initiating parameter adjustment and graph updates to ensure the continuous expansion of the coverage and adaptability of the compensation correlation graph. The correlation graph shortens the compensation response time under similar interference scenarios, while reducing computational power consumption and improving the execution efficiency of the dynamic anti-interference strategy.

[0064] In some specific embodiments, the charging module may become unsuitable for compensation operations due to sudden strong interference, excessively high temperature, or other reasons. To reduce hardware damage caused by improper compensation, a protection and anti-interference mechanism is added to the charging module. The details are as follows: Real-time monitoring of the charging module's operating status parameters, including input voltage fluctuation range, output current stability, and power device temperature; When the operating status parameters exceed the preset safety threshold, the execution of the dynamic interference compensation strategy is suspended, and the preset protection anti-interference mechanism is activated. The preset protection anti-interference mechanism is used to suppress the impact of sudden strong interference on the charging module.

[0065] During the execution of the dynamic interference compensation strategy, a real-time monitoring mechanism for operating status parameters needs to be activated simultaneously. This mechanism continuously collects core operating status parameters during module operation, including input voltage fluctuation range (e.g., the difference between the instantaneous peak and valley values ​​of the grid input voltage), output current stability (e.g., the ripple coefficient and fluctuation amplitude of the output current), and power device temperature. During real-time monitoring, the collected operating status parameters are compared in real-time with corresponding preset safety thresholds. If any operating status parameter exceeds the preset safety threshold, it is determined that the charging module has encountered a sudden strong interference (e.g., grid lightning surge, external strong electromagnetic pulse impact) or that abnormal parameter adjustment has caused operational risks. In this case, the execution of the dynamic interference compensation strategy needs to be immediately suspended to reduce the exacerbation of module malfunctions due to continuous compensation operations. Simultaneously, a preset protection anti-interference mechanism is automatically activated, which includes multi-level protection measures. Finally, the recovery status of the operating status parameters is continuously monitored. Once all operating status parameters have fallen back to within the preset safety threshold range and have been running stably for a preset time (e.g., 3 seconds), the execution process of the dynamic interference compensation strategy is restarted.

[0066] In this embodiment, virtualization technology is used to build a hybrid electromagnetic interference simulation environment when constructing the interference identification model, and the model parameters are optimized. This provides comprehensive training data support covering multiple types and intensities of interference for the interference identification model, improving the identification accuracy and inference efficiency of multi-source composite interference in complex electromagnetic environments. Furthermore, when calling the RBF mapping model to execute the dynamic interference compensation strategy, the interference scenario is matched with the compensation scheme, making the dynamic interference compensation strategy targeted. At the same time, the dynamic interference compensation strategy, which includes LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation, is executed in segments according to the compensation priority, ensuring that the dynamic interference compensation strategy is executed in an orderly manner. Subsequently, based on the anti-interference deviation value, the compensation parameters of LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation are adjusted in stages to reduce circuit parameter disorder caused by simultaneous multi-dimensional operation, and also to make the dynamic interference compensation strategy adapt to actual interference changes, thereby performing effective compensation and improving the anti-interference capability and charging performance of the charging module. In addition, by constructing a compensation correlation spectrum to reuse the optimal parameters and adding a protection mechanism to protect the charging module, the execution efficiency and operational reliability of the dynamic interference compensation strategy are improved.

[0067] Please see Figure 3 This application also provides a charging interference suppression system based on electromagnetic environment, comprising: The acquisition unit 301 is used to acquire composite interference data of the charging module when the charging module starts working. The composite interference data is used to characterize the electromagnetic interference on the power grid side, inside the charging module, and in the communication link. Extraction unit 302 is used to extract the spectral features, temporal features and statistical features of composite interference data respectively, and construct them into a high-dimensional interference feature vector and a Bayesian weight vector; The building unit 303 is used to build an interference recognition model through the MobileNetV3-Small neural network and deploy the quantized and compressed interference recognition model in the charging module; The inference unit 304 is used to input the high-dimensional interference feature vector and the Bayesian weight vector into the interference identification model according to a preset period to obtain the inference result, which includes the interference type, interference intensity and inference confidence. The compensation unit 305 is used to perform a dynamic interference compensation strategy on the charging module based on a preset RBF mapping model when the reasoning confidence reaches a preset confidence threshold. The monitoring unit 306 is used to monitor the charging module after the dynamic interference compensation strategy is executed, and to obtain the power conversion efficiency dataset and the electromagnetic immunity quantification index set. The calculation unit 307 is used to calculate the disturbance rejection deviation value based on the power conversion efficiency dataset and the preset efficiency threshold, and combined with the weighted scoring results of the electromagnetic disturbance rejection quantification index set. The adjustment unit 308 is used to adjust the compensation parameters of the dynamic interference compensation strategy according to the anti-interference deviation value.

[0068] Optionally, the compensation unit 305 is specifically used for: Based on the interference type and intensity, the interference rule base of the preset RBF mapping model is called to generate a compensation weight matrix and interference scene identifier; Calculate the information entropy of each compensation dimension of the compensation weight matrix; Based on the information entropy of each compensation dimension, the interference suppression contribution is generated; Compensation priority weights are assigned based on the contribution of interference suppression. Based on the compensation priority weight and interference scenario identification, a dynamic interference compensation strategy is executed through the benchmark parameter library of the RBF mapping model. The dynamic interference compensation strategy includes LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation.

[0069] Optionally, the compensation unit 305 is specifically used for: Based on compensation priority weights and interference scenario identifiers, a dynamic interference compensation strategy is executed through the baseline parameter library of the RBF mapping model, including: Based on the interference scene identifier, the corresponding benchmark parameter set is matched through the benchmark parameter library of the RBF mapping model; The compensation priority weights are sorted in descending order, and the execution order of LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy is determined according to the sorting results. According to the execution order and the reference parameter set, LLC resonant frequency compensation, PFC circuit coefficient compensation and reverse interference signal compensation are performed on the charging module in segments.

[0070] Optionally, adjustment unit 308 is specifically used for: The disturbance rejection deviation value is divided into several adjustment levels, and the initial step size coefficient is calculated according to the adjustment level. Extract the energy efficiency variation characteristics of the power conversion energy efficiency dataset, and calculate the compensation effect attenuation coefficient based on the energy efficiency variation characteristics; The initial step size coefficient is corrected by the compensation effect attenuation coefficient to obtain the final step size coefficient; The adjustment amount of the compensation parameters for the dynamic interference compensation strategy is calculated based on the final step size coefficient and the compensation priority weight. Based on the adjustment amount of the compensation parameters, the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy are iterated separately.

[0071] Optionally, building unit 303 is specifically used for: A simulated environment of hybrid electromagnetic interference was created using virtualization technology, and the initial parameters of the interference identification model were configured based on the MobileNetV3-Small neural network architecture. Input the test set into the interference identification model to optimize the initial parameters; The interference identification model with optimized initial parameters is compressed using the INT8 static quantization method and then deployed in the charging module after compilation.

[0072] Optionally, it also includes calling unit 309, specifically used for: When the anti-disturbance deviation value reaches the preset deviation threshold range, a compensation correlation map is established based on the adjusted compensation parameters and inference results; When the subsequent inference results match the feature elements in the association graph, the corresponding target compensation parameters in the association graph are directly called, and a dynamic interference compensation strategy is executed according to the target compensation parameters.

[0073] Optionally, a protection unit 310 is also included, specifically for: Real-time monitoring of the charging module's operating status parameters, including input voltage fluctuation range, output current stability, and power device temperature; When the operating status parameters exceed the preset safety threshold, the execution of the dynamic interference compensation strategy is suspended, and the preset protection anti-interference mechanism is activated. The preset protection anti-interference mechanism is used to suppress the impact of sudden strong interference on the charging module.

[0074] For specific implementation methods, please refer to [the example]. Figure 1 and Figure 2 Examples are not detailed here.

[0075] Please see Figure 4 This application also provides a charging interference suppression device based on electromagnetic environment, comprising: Processor 401, memory 402, input / output unit 403, bus 404; The processor 401 is connected to the memory 402, the input / output unit 403, and the bus 404; The memory 402 stores a program, and the processor 401 calls the program to execute any of the methods described above.

[0076] This application also relates to a computer-readable storage medium on which a program is stored, which, when run on a computer, causes the computer to perform any of the methods described above.

[0077] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0079] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0080] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0081] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A charging interference immunity method based on electromagnetic environment, characterized in that, The charging interference suppression method includes: When the charging module starts working, composite interference data of the charging module is collected. The composite interference data is used to characterize electromagnetic interference on the power grid side, inside the charging module, and in the communication link. The spectral features, temporal features, and statistical features of the composite interference data are extracted respectively, and constructed into a high-dimensional interference feature vector and a Bayesian weight vector; An interference recognition model is constructed using the MobileNetV3-Small neural network, and the quantized and compressed interference recognition model is deployed in the charging module. The high-dimensional interference feature vector and the Bayesian weight vector are input into the interference identification model according to a preset period to obtain the inference result, which includes the interference type, interference intensity and inference confidence. When the reasoning confidence reaches a preset confidence threshold, a dynamic interference compensation strategy is executed on the charging module based on a preset RBF mapping model; After implementing the dynamic interference compensation strategy, the charging module is monitored to obtain a power conversion efficiency dataset and an electromagnetic interference quantification index set. Based on the power conversion efficiency dataset and the preset efficiency threshold, and combined with the weighted scoring results of the electromagnetic immunity quantification index set, the immunity deviation value is calculated. The compensation parameters of the dynamic interference compensation strategy are adjusted based on the anti-interference deviation value.

2. The charging interference suppression method according to claim 1, characterized in that, The dynamic interference compensation strategy for the charging module based on the preset RBF mapping model includes: Based on the interference type and the interference intensity, a compensation weight matrix and an interference scene identifier are generated by calling the interference rule base of the preset RBF mapping model. Calculate the information entropy of each compensation dimension of the compensation weight matrix; Based on the information entropy of each compensation dimension, an interference suppression contribution is generated; Compensation priority weights are assigned based on the interference suppression contribution. Based on the compensation priority weight and the interference scenario identifier, a dynamic interference compensation strategy is executed through the benchmark parameter library of the RBF mapping model. The dynamic interference compensation strategy includes LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation.

3. The charging interference suppression method according to claim 2, characterized in that, The step of executing a dynamic interference compensation strategy based on the compensation priority weight and the interference scene identifier through the baseline parameter library of the RBF mapping model includes: Based on the interference scene identifier, the corresponding benchmark parameter set is matched through the benchmark parameter library of the RBF mapping model; The compensation priority weights are sorted in descending order, and the execution order of LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy is determined according to the sorting result. According to the execution order and the reference parameter set, the LLC resonant frequency compensation, the PFC circuit coefficient compensation, and the reverse interference signal compensation are performed on the charging module in segments.

4. The charging interference suppression method according to claim 2, characterized in that, The step of adjusting the compensation parameters of the dynamic interference compensation strategy based on the anti-interference deviation value includes: The disturbance rejection deviation value is divided into several adjustment levels, and the initial step size coefficient is calculated according to the adjustment level. Extract the energy efficiency variation characteristics of the power conversion energy efficiency dataset, and calculate the compensation effect attenuation coefficient based on the energy efficiency variation characteristics; The initial step size coefficient is corrected by the compensation effect attenuation coefficient to obtain the final step size coefficient; The adjustment amount of the compensation parameters of the dynamic interference compensation strategy is calculated based on the final step size coefficient and the compensation priority weight. Based on the adjustment amount of the compensation parameters, the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy are iterated respectively.

5. The charging interference suppression method according to claim 1, characterized in that, The step of constructing an interference recognition model using a MobileNetV3-Small neural network and deploying the quantized and compressed interference recognition model in the charging module includes: A simulated environment of hybrid electromagnetic interference was created using virtualization technology, and the initial parameters of the interference identification model were configured based on the MobileNetV3-Small neural network architecture. Input a test set into the interference identification model to optimize the initial parameters; The interference identification model after optimizing the initial parameters is compressed using the INT8 static quantization method and then deployed in the charging module after compilation.

6. The charging interference suppression method according to claim 4, characterized in that, After iterating the compensation parameters for LLC resonant frequency compensation, PFC circuit coefficient compensation, and reverse interference signal compensation in the dynamic interference compensation strategy according to the compensation parameter adjustment amount, the charging anti-interference method further includes: When the anti-disturbance deviation value reaches the preset deviation threshold range, a compensation correlation map is established based on the adjusted compensation parameters and the inference results; When the subsequent inference result matches the feature element in the association graph, the corresponding target compensation parameter in the association graph is directly called, and the dynamic interference compensation strategy is executed according to the target compensation parameter.

7. The charging interference suppression method according to any one of claims 1 to 6, characterized in that, The charging interference suppression method further includes: Real-time monitoring of the charging module's operating status parameters, including input voltage fluctuation range, output current stability, and power device temperature; When the operating status parameters exceed the preset safety threshold, the execution of the dynamic interference compensation strategy is suspended, and a preset protection anti-interference mechanism is activated. The preset protection anti-interference mechanism is used to suppress the impact of sudden strong interference on the charging module.

8. A charging interference suppression system based on electromagnetic environment, characterized in that, The charging interference immunity system includes: The acquisition unit is used to acquire composite interference data of the charging module when the charging module starts working. The composite interference data is used to characterize electromagnetic interference on the power grid side, inside the charging module, and in the communication link. The extraction unit is used to extract the spectral features, temporal features, and statistical features of the composite interference data, and construct them into a high-dimensional interference feature vector and a Bayesian weight vector. The building unit is used to build an interference recognition model through the MobileNetV3-Small neural network and deploy the quantized and compressed interference recognition model in the charging module; The inference unit is used to input the high-dimensional interference feature vector and the Bayesian weight vector into the interference identification model according to a preset period to obtain the inference result, which includes the interference type, interference intensity and inference confidence. The compensation unit is used to perform a dynamic interference compensation strategy on the charging module based on a preset RBF mapping model when the reasoning confidence reaches a preset confidence threshold. The monitoring unit is used to monitor the charging module after the dynamic interference compensation strategy is executed, and to obtain the power conversion efficiency dataset and the electromagnetic interference quantification index set. The calculation unit is used to calculate the disturbance rejection deviation value based on the power conversion efficiency dataset and the preset efficiency threshold, and in combination with the weighted scoring results of the electromagnetic disturbance rejection quantification index set. The adjustment unit is used to adjust the compensation parameters of the dynamic interference compensation strategy according to the anti-interference deviation value.

9. A charging interference suppression device based on electromagnetic environment, characterized in that, The charging interference suppression device includes: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains a program that, when executed on a computer, performs the method as described in any one of claims 1 to 7.

Images