Intelligent management and control method and system applied to vehicle-mounted wireless charger

Abstract

The invention discloses an intelligent management and control method and system applied to a vehicle-mounted wireless charger, and the method comprises the steps: obtaining a reflected magnetic field signal, carrying out the detection and positioning of a to-be-charged device, and generating the positioning information of the to-be-charged device; judging whether a safe link can be established or not, performing charging protocol matching and data acquisition, and constructing a multi-modal fusion feature of the to-be-charged device; vehicle-mounted end dynamic monitoring information is obtained, real-time driving scene analysis is carried out based on the vehicle-mounted end dynamic monitoring information, and driving scene analysis information is obtained; extracting real-time battery state features and historical charging behavior features of the to-be-charged equipment based on the multi-modal fusion features, and performing battery health analysis on the target to-be-charged equipment to obtain battery health analysis information; the charging strategy knowledge graph is constructed, the optimal charging strategy of the target to-be-charged equipment in the current driving scene is formulated, a regulation and control instruction is generated in combination with the positioning information of the to-be-charged equipment to perform wireless charging intelligent management and control, the charging efficiency is improved, and the battery health is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of in-vehicle wireless charging control, and particularly to an intelligent control method and system applied to in-vehicle wireless charging. Background Art

[0002] Traditional in-vehicle wireless charging systems generally adopt static and passive control logics, and their technical implementation highly depends on preset electromagnetic coupling parameters and fixed power output modes. Existing solutions achieve device positioning through a single magnetic field sensor or simple multi-coil scanning, but do not consider the dynamic position offsets caused by bumps and steering during vehicle driving, resulting in positioning errors and a decrease in magnetic field energy transmission efficiency. In a multi-device charging scenario, traditional methods allocate charging resources based on the physical contact order of devices, lacking quantitative analysis of the device power urgency, battery health status, and user usage habits, causing charging delays for critical devices or inefficient repeated charging problems. Under the traditional control logic, personalized charging control cannot be performed based on the device to be charged, thus affecting the battery of the device to be charged.

[0003] The above technical defects lead to significant bottlenecks in the efficiency and safety of in-vehicle wireless charging systems. Especially in the complex working conditions of high dynamics and multi-scenarios of intelligent electric vehicles, it is difficult to meet the dual requirements of users for efficient power supply and battery life. There is an urgent need for an intelligent control method that integrates multi-modal perception, dynamic strategy optimization, and cross-domain collaborative control to achieve intelligent charging control and take into account charging efficiency and battery health. Summary of the Invention

[0004] The present invention overcomes the defects of the prior art and provides an intelligent control method and system applied to in-vehicle wireless charging. An important objective is to improve charging efficiency and ensure battery health.

[0005] To achieve the above objective, in the first aspect of the present invention, an intelligent control method applied to in-vehicle wireless charging is provided, including: Repeatedly scanning the in-vehicle wireless charging area to obtain a reflected magnetic field signal, detecting and positioning the device to be charged based on the reflected magnetic field signal, and generating positioning information of the device to be charged; After detecting and positioning the device to be charged, determining whether a secure link can be established with the device to be charged, performing charging protocol matching and data collection, and constructing a multi-modal fusion feature of the device to be charged; Monitoring the vehicle dynamics during the driving of the target vehicle to obtain in-vehicle dynamic monitoring information, and performing real-time driving scenario analysis based on the in-vehicle dynamic monitoring information to obtain driving scenario analysis information; Extract the real-time battery state features and historical charging behavior features of the device to be charged based on the multimodal fusion features of the device to be charged, perform battery health analysis on the target device to be charged, and obtain battery health analysis information; Construct a charging strategy knowledge graph, formulate an optimal charging strategy for the target device to be charged in the current driving scenario, and generate a control command in combination with the positioning information of the device to be charged for intelligent wireless charging control.

[0006] In this solution, the repeated scanning of the on-vehicle wireless charging area to obtain the reflected magnetic field signal, and the detection and positioning of the device to be charged based on the reflected magnetic field signal to generate the positioning information of the device to be charged specifically include: The electromagnetic induction coil array built into the wireless charging board uses a time-sharing activation strategy to scan the charging area, and sequentially sends high-frequency signals to the coils at different positions to obtain the reflected magnetic field signal; Import the obtained reflected magnetic field signal into a high-precision analog-to-digital converter for conversion to obtain a reflected magnetic field digital signal, use a digital filtering algorithm to preprocess the reflected magnetic field digital signal, and generate a signal intensity matrix according to the preprocessed reflected magnetic field digital signal; Extract the intensity feature of the reflected magnetic field signal at the current moment according to the signal intensity matrix. If the intensity of the reflected magnetic field signal at the current moment is greater than the preset signal intensity threshold and the number of scans of the corresponding scanned area is greater than the preset scan number threshold, it is determined that there is a device to be charged in the target area and it is marked as a locked area; When it is detected that there is a device to be charged on the target wireless charging board, obtain the reflected magnetic field signal of the locked area, select the three coils with the optimal geometric distribution in the locked area as reference nodes, and calculate the distance estimation value from the device to be charged to each node based on the path loss model; Construct a non-linear geometric equation set according to the reference node coordinates and the distance estimation value, convert it into an overdetermined linear equation set through linearization processing, and use the weighted least squares method to solve the optimal two-dimensional plane coordinates to generate the positioning information of the device to be charged.

[0007] In this solution, after detecting and positioning the device to be charged, it is judged whether a secure link can be established with the device to be charged, and the charging protocol is matched and data is collected to construct the multimodal fusion features of the device to be charged, specifically including: After detecting and positioning the device to be charged, obtain the unique identifier stored in the vehicle-mounted terminal, convert the unique identifier through a hash function and use it as the key derivation seed input of the elliptic curve encryption algorithm to generate a public-private key pair of the vehicle-mounted terminal and send it to the device to be charged terminal for key exchange; After completing the key exchange, the vehicle-mounted end uses the AES-GCM encryption algorithm to generate a challenge code and sends it to the device-end through a secure channel. The device-end accepts the challenge code, signs it, and sends it back to the vehicle-mounted end for identity authentication. If the identity authentication result of the vehicle-mounted end to the device to be charged is successfully verified within the preset number of verifications, a secure link is established with the device to be charged and the charging protocol characteristic parameters of the device to be charged are obtained through the encrypted channel; Input the acquired charging protocol characteristic parameters into the preset charging protocol library and perform encoding matching with the pre-built hash tree structure, traverse the hash tree nodes and compare them with the characteristic hash values ​​in the preset protocol library, output the protocol matching results and corresponding working mode parameters, and obtain the charging protocol matching information; The vehicle-mounted end obtains real-time battery status data and historical charging behavior data of the device-end user from the device-end through the established secure link, and obtains the user's vehicle-mounted charging behavior data from the vehicle-mounted database to generate first data collection information; Calculating the data standard deviation of the first data acquisition information, eliminating the corresponding data that is not within the preset standard deviation range based on the calculated data standard deviation, supplementing it through linear interpolation, and obtaining data preprocessing information; The positioning information of the device to be charged is obtained, and combined with the charging protocol matching information and the data preprocessing information, an attention mechanism is introduced to perform multimodal feature fusion to obtain the multimodal fusion features of the device to be charged.

[0008] In this solution, the vehicle dynamics of the monitoring target vehicle during driving obtains the vehicle-mounted dynamic monitoring information, and real-time driving scene analysis is performed based on the vehicle-mounted dynamic monitoring information to obtain driving scene analysis information, which specifically includes: Based on the on-board multi-source sensor array, the vehicle dynamics of the target vehicle during driving are monitored to obtain on-board dynamic monitoring information, wherein the on-board dynamic monitoring information includes driving environment data of the target vehicle and driving state data of the target vehicle; Extracting the target vehicle driving state data according to the vehicle-mounted dynamic monitoring information and performing time series processing to obtain a driving state time series sequence, importing the driving state time series sequence into a pre-trained time series convolutional network for deep feature mining, and using an extended causal convolutional structure to stack multiple residual blocks to analyze the time series motion state of the target vehicle, and generate a motion state feature vector; Extract the target vehicle driving environment data based on the vehicle-mounted dynamic monitoring information, import it into the environment perception model built with YOLOv5 as the framework for analysis, identify the road type and traffic density, calculate the relative movement speed of the target ahead by combining the optical flow method, generate a traffic condition score, and obtain a visual state feature vector; Input the motion state feature vector and the visual state feature vector into the cross-modal attention fusion module. Use the temporal motion feature as the query vector and the visual feature as the key-value vector. Calculate the cross-modal correlation weight matrix through the multi-head attention mechanism, screen out the strongly correlated feature combinations, and perform temporal dependence modeling through the gated recurrent unit to generate a joint representation vector; Obtain the navigation information of the target vehicle through the in-vehicle system of the target vehicle, extract the navigation distance feature of the current navigation path for driving duration estimation, and construct a driving feature vector by combining the driving duration and the navigation distance; Introduce a big data network to obtain the historical driving scenario features of different driving scenario categories. Based on the joint representation vector and the driving feature vector, construct the current driving scenario portrait of the target vehicle, compare it with the historical driving scenario features, and analyze the current driving scenario category of the target vehicle to obtain the driving scenario analysis information.

[0009] In this solution, the real-time battery state feature and the historical charging behavior feature of the device to be charged are extracted based on the multi-modal fusion feature of the device to be charged, and the battery health of the target device to be charged is analyzed to obtain the battery health analysis information, which specifically includes: Extract the real-time battery state feature and the historical charging behavior feature of the device to be charged through the multi-modal fusion feature of the device to be charged to obtain the first feature extraction information, where the first feature extraction information includes the remaining power, voltage, current, temperature, internal resistance, number of cycles, and historical charge-discharge curve; Based on the first feature extraction information, construct a standardized battery state matrix of the device to be charged, adopt a hierarchical evaluation strategy for battery health analysis, and import the standardized battery state matrix into the extended Kalman filter; Define the state vector as the battery capacity attenuation rate and the internal resistance growth rate, the observation vector as the real-time voltage-current curve, construct the state equation and perform discretization processing to transform it into a state transition matrix, and iteratively update the Jacobian matrix in combination with the real-time charge-discharge data to generate the battery SOH estimation value of the device to be charged; Subsequently, take the standardized battery state matrix as the input and import it into the LSTM network. Use the historical charge-discharge cycle data as the initial feature and convert it into a time-aware feature vector through the time embedding layer to perform long-term capacity attenuation prediction and output a probabilistic SOH degradation curve; Integrate the SOH estimation value and the probabilistic SOH degradation curve using the Bayesian fusion algorithm. Take the real-time SOH estimation value as the prior distribution and the long-term prediction probability as the likelihood function, and calculate the posterior distribution based on Bayes' theorem; Solve the expected value and confidence interval of the posterior distribution, introduce variational inference to accelerate the calculation, use KL divergence to minimize the difference between the approximate posterior and the true distribution, and finally output the comprehensive SOH score of the battery of the device to be charged to obtain battery health analysis information.

[0010] In this solution, the construction of the charging strategy knowledge graph, formulating the optimal charging strategy for the target device to be charged in the current driving scenario, and generating a control command in combination with the positioning information of the device to be charged for intelligent wireless charging management specifically include: Use the big data network to obtain historical wireless charging instances in different driving scenarios, extract features from each historical wireless charging instance to obtain the driving scenario features, charging device status features, and charging strategy features corresponding to each historical wireless charging instance; Using the driving scenario features as the initial classification basis, initially classify each historical wireless charging instance to generate several category sets, and calculate the Euclidean distance values between the corresponding historical wireless charging instances within each category set based on the charging device status features; Based on the calculated Euclidean distance values, use the K-mean clustering algorithm to further classify each category set to generate several category subsets, and associate the driving scenario features corresponding to each category subset to obtain the category division result information; Based on the category division result information, construct entity triples according to the driving scenario features, charging device status features, and charging strategy features, and construct a charging strategy knowledge graph through the entity triples; Obtain battery health analysis information, driving scenario analysis information, and multi-modal fusion features of the device to be charged, and input them into the charging strategy knowledge graph to analyze the appropriate charging strategy for the current driving scenario and charging device status; Set charging protocol adaptation constraint conditions based on the multi-modal fusion features of the device to be charged, use the cosine metric algorithm to analyze the matching degree between each charging strategy stored in the charging strategy knowledge graph and the current charging event, and generate several candidate charging strategies; Screen each candidate charging strategy through the set charging protocol adaptation constraint conditions, eliminate the charging strategies that do not meet the charging protocol supported by the current device to be charged, obtain the remaining candidate charging strategies and sort them according to the matching degree with the current charging event, and select the one with the highest matching degree as the final charging strategy; Obtain the positioning information of the device to be charged, generate a wireless charging control command through the final charging strategy, and generate a coil position activation command for the wireless charging board in combination with the positioning information of the device to be charged to perform intelligent charging management on the target device to be charged.

[0011] The second aspect of the present invention provides an intelligent control system applied to in-vehicle wireless charging, which system includes: a memory and a processor. The memory contains an intelligent control method program for in-vehicle wireless charging. When the intelligent control method program for in-vehicle wireless charging is executed by the processor, the following steps are implemented: Repeatedly scan the in-vehicle wireless charging area to obtain a reflected magnetic field signal, detect and locate the device to be charged based on the reflected magnetic field signal, and generate device-to-be-charged location information; After detecting and locating the device to be charged, determine whether a secure link can be established with the device to be charged, perform charging protocol matching and data collection, and construct a multi-modal fusion feature of the device to be charged; Monitor the vehicle dynamics during the driving of the target vehicle to obtain vehicle-mounted dynamic monitoring information, and perform real-time driving scenario analysis based on the vehicle-mounted dynamic monitoring information to obtain driving scenario analysis information; Extract the real-time battery state feature and historical charging behavior feature of the device to be charged based on the multi-modal fusion feature of the device to be charged, perform battery health analysis on the target device to be charged, and obtain battery health analysis information; Construct a charging strategy knowledge graph, formulate the optimal charging strategy for the target device to be charged in the current driving scenario, and generate a control command in combination with the device-to-be-charged location information for intelligent control of wireless charging.

[0012] The present invention discloses an intelligent control method and system applied to in-vehicle wireless charging, including: obtaining a reflected magnetic field signal, detecting and locating the device to be charged, and generating device-to-be-charged location information; determining whether a secure link can be established, performing charging protocol matching and data collection, and constructing a multi-modal fusion feature of the device to be charged; obtaining vehicle-mounted dynamic monitoring information, performing real-time driving scenario analysis based on the vehicle-mounted dynamic monitoring information, and obtaining driving scenario analysis information; extracting the real-time battery state feature and historical charging behavior feature of the device to be charged based on the multi-modal fusion feature, performing battery health analysis on the target device to be charged, and obtaining battery health analysis information; constructing a charging strategy knowledge graph, formulating the optimal charging strategy for the target device to be charged in the current driving scenario, combining the device-to-be-charged location information to generate a control command for intelligent control of wireless charging, improving the charging efficiency and ensuring battery health. Description of the Drawings

[0013] In order to more clearly illustrate the technical solutions in the embodiments or exemplary examples of the present invention, the following will briefly introduce the drawings required for use in the embodiments or exemplary descriptions. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the drawings shown.

[0014] Figure 1 Flow chart of an intelligent control method applied to in-vehicle wireless charging provided by an embodiment of the present invention; Figure 2 Flow chart of an intelligent charging control method for in-vehicle wireless charging provided by an embodiment of the present invention; Figure 3 Block diagram of an intelligent control system applied to in-vehicle wireless charging provided by an embodiment of the present invention; The realization, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0015] In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments may be combined with each other.

[0016] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0017] Figure 1 Flow chart of an intelligent control method applied to in-vehicle wireless charging provided by an embodiment of the present invention; As Figure 1 shown, the present invention provides a flow chart of an intelligent control method applied to in-vehicle wireless charging, including: S102, repeatedly scan the in-vehicle wireless charging area to obtain a reflected magnetic field signal, detect and locate the device to be charged based on the reflected magnetic field signal, and generate the location information of the device to be charged; S104, when the device to be charged is detected and located, determine whether a secure link can be established with the device to be charged, perform charging protocol matching and data collection, and construct a multi-modal fusion feature of the device to be charged; S106, monitor the vehicle dynamics during the driving of the target vehicle to obtain vehicle-side dynamic monitoring information, and perform real-time driving scenario analysis based on the vehicle-side dynamic monitoring information to obtain driving scenario analysis information; S108, extract the real-time battery state feature and historical charging behavior feature of the device to be charged based on the multi-modal fusion feature of the device to be charged, perform battery health analysis on the target device to be charged, and obtain battery health analysis information; S110. Construct a charging strategy knowledge graph, formulate the optimal charging strategy for the target device to be charged in the current driving scenario, and generate a control instruction in combination with the positioning information of the device to be charged for intelligent wireless charging control.

[0018] Further, in a preferred embodiment of the present invention, the method for repeatedly scanning the in-vehicle wireless charging area to obtain a reflected magnetic field signal, detecting and positioning the device to be charged based on the reflected magnetic field signal, and generating positioning information of the device to be charged specifically includes: The electromagnetic induction coil array built in the wireless charging board adopts a time-sharing activation strategy to scan the charging area, and sequentially sends high-frequency signals to coils at different positions to obtain the reflected magnetic field signal; Import the obtained reflected magnetic field signal into a high-precision analog-to-digital converter for conversion to obtain a digital reflected magnetic field signal, preprocess the digital reflected magnetic field signal using a digital filtering algorithm, and generate a signal strength matrix based on the preprocessed digital reflected magnetic field signal; Extract the intensity characteristics of the reflected magnetic field signal at the current moment according to the signal strength matrix. If the intensity of the reflected magnetic field signal at the current moment is greater than the preset signal intensity threshold, and the number of scans of the corresponding scanned area is greater than the preset scan number threshold, it is determined that there is a device to be charged in the target area, and it is marked as a locked area; When it is detected that there is a device to be charged on the target wireless charging board, obtain the reflected magnetic field signal of the locked area, select the three coils with the best geometric distribution in the locked area as reference nodes, and calculate the distance estimation values from the device to be charged to each node based on the path loss model; Construct a non-linear geometric equation set according to the reference node coordinates and the distance estimation values, convert it into an overdetermined linear equation set through linearization processing, and use the weighted least squares method to solve the optimal two-dimensional plane coordinates to generate the positioning information of the device to be charged.

[0019] It should be noted that after the system starts, the electromagnetic induction coil array embedded in the wireless charging board sequentially activates each independent coil unit in a time-division multiplexing mode. Each coil emits a high-frequency electromagnetic signal with a specific frequency (such as 6.78 MHz) to the charging area during the activation period and simultaneously receives the magnetic field signal reflected by the receiving coil of the device to be charged. To avoid multi-coil signal interference, a time-division multiple access (TDMA) mechanism is adopted to ensure that each coil emits a signal only within the allocated time slot, and the emission interval between adjacent coils is controlled within 10 μs. The reflected magnetic field signal is digitized by a high-precision analog-to-digital converter (ADC) at a sampling rate of 1 MHz to generate a time-domain signal sequence containing amplitude and phase information. Subsequently, environmental noise is filtered out through a digital filtering algorithm (such as a FIR low-pass filter based on a window function), and the signal is normalized. Finally, a reflection signal intensity matrix indexed by the coil position is output, and the matrix dimension is consistent with the physical layout of the charging board. Subsequently, feature extraction is performed on the signal intensity matrix, and the mean and variance of the signal intensity in each coil area are statistically calculated through a sliding window. When the mean value of the signal intensity in a certain area exceeds the preset threshold for three consecutive scans, and the variance is lower than the fluctuation tolerance, it is determined that there is a metal foreign object or a device to be charged in this area. To exclude occasional interference, historical scan data is verified: if the target area is marked more than 3 times within 5 consecutive scan cycles, it is confirmed as a valid target and marked as a locked area, and the precise positioning process is started.

[0020] Furthermore, after the locked area is determined, three reference coil nodes with the optimal geometric distribution are selected from this area. The selection criterion is based on the principle of plane geometry optimization: three coils with the largest spacing and non-collinear distribution are preferentially selected (for example, the vertices of an equilateral triangle) to minimize the positioning error. After the reference nodes are selected, the distance estimation values from the device to each node are calculated based on the path loss model. Specifically, the reflected signal intensity is substituted into the attenuation model where is the reference distance, is the signal intensity at and the path loss exponent

[0021] is calibrated through experiments (typical values in the vehicle environment are 2.5 - 3.5), and the distances between the device and each node are inversely deduced. Based on the physical coordinates of the reference nodes and the distance estimation values, a non-linear equation system is constructed and converted into an overdetermined linear equation system through linearization. The weighted least squares method (WLS) is introduced for solution: a weight matrix is constructed according to the confidence levels of the distance estimation values (inversely weighted by the signal intensity variance), and the two-dimensional coordinates of the device are calculated through matrix operations. Finally, the plane coordinate information is output, providing a spatial reference for subsequent dynamic alignment and energy transmission.Further, in a preferred embodiment of the present invention, after detecting and locating the device to be charged, it is determined whether a secure link can be established with the device to be charged, and a charging protocol matching and data collection are performed to construct a multi-modal fusion feature of the device to be charged, which specifically includes: After detecting and locating the device to be charged, obtain the unique identifier stored in the vehicle-mounted terminal, convert the unique identifier through a hash function and input it as the key derivation seed of the elliptic curve encryption algorithm, generate a public-private key pair of the vehicle-mounted terminal and send it to the device to be charged for key exchange; After completing the key exchange, the vehicle-mounted terminal uses the AES-GCM encryption algorithm to generate a challenge code and send it to the device terminal through a secure channel. The device terminal accepts the challenge code, signs it and sends it back to the vehicle-mounted terminal for identity verification; If within the preset number of verification times, the vehicle-mounted terminal's identity verification result for the device to be charged terminal is successful verification, establish a secure link with the device to be charged terminal and obtain the charging protocol characteristic parameters of the device to be charged through the ciphertext channel; Input the obtained charging protocol characteristic parameters into a preset charging protocol library for encoding matching with a pre-constructed hash tree structure, traverse the hash tree nodes and compare them with the characteristic hash values in the preset protocol library, output the protocol matching result and the corresponding working mode parameters, and obtain the charging protocol matching information; The vehicle-mounted terminal obtains the real-time battery status data and the device terminal user's historical charging behavior data from the device to be charged through the established secure link, and obtains the user's vehicle-mounted charging behavior data from the vehicle-mounted database to generate the first data collection information; Calculate the data standard deviation of the first data collection information, eliminate the corresponding data that is not within the preset standard deviation range based on the calculated data standard deviation, and supplement it through linear interpolation to obtain the data preprocessing information; Obtain the positioning information of the device to be charged, combine the charging protocol matching information and the data preprocessing information, introduce an attention mechanism for multi-modal feature fusion, and obtain the multi-modal fusion feature of the device to be charged.

[0022] It should be noted that after detecting and locating the device to be charged, the unique identifier of the device (such as the device serial number or hardware encryption ID) is extracted from the encrypted database stored on the vehicle side. The identifier is unidirectionally transformed through the HMAC-SHA-256 hash function to generate a fixed-length digest value, which is used as the key derivation seed and input into the deterministic key generation algorithm based on the elliptic curve to generate the temporary public-private key pair on the vehicle side. Subsequently, the public key is sent to the device to be charged through the secure communication protocol, triggering the device side to generate a public-private key pair based on the same algorithm and return the public key. The two parties calculate the shared key based on the elliptic curve protocol and derive the final session key using the hash algorithm to complete the end-to-end key negotiation. After the key exchange is completed, the vehicle side generates a 128-bit random number as the challenge code, encrypts it using the AES-GCM algorithm in combination with the session key, and sends it to the device side. After the device side decrypts to obtain the challenge code, it generates signature data based on the elliptic curve digital signature algorithm using the local private key and encrypts it and sends it back to the vehicle side. The vehicle side verifies the validity of the signature using the public key of the device side. If the verification fails, a challenge is re-initiated (up to 3 times). If it fails continuously, the device is determined to be illegal and the process is terminated. After the verification passes, a two-way authentication secure link based on the session key is established between the two parties to ensure the confidentiality and integrity of subsequent communications. Through the secure link, the vehicle side sends a protocol query instruction to the device side to obtain the charging protocol characteristic parameters supported by the device (such as protocol version, power level, frequency range, etc., which are encoded fields). Subsequently, the protocol parameters are input into the pre-constructed hash tree structure for fast matching: each leaf node of the hash tree corresponds to the characteristic hash value of a protocol in the preset protocol library. By comparing the hash values layer by layer, the target protocol node is located, and the protocol identifier that matches successfully and its corresponding working parameters are output to complete the protocol adaptation.

[0023] Furthermore, the vehicle side obtains the real-time battery status data of the device (such as the current battery level, temperature, number of charge cycles, etc.) and the user's historical charging behavior data stored locally on the device (such as charging time period distribution, average charging rate) through the secure link. At the same time, the vehicle charging records of the user (such as historical charging location preferences, charging interruption frequency) are extracted from the vehicle database. For the above multi-source data, the standard deviation of each dimension is calculated, and the threshold range is set using the dynamic 3σ principle. After removing the outlier data points, the missing values are filled based on the time series context using the linear interpolation method to generate a data preprocessing result with continuous time series and consistent distribution. Finally, the device location information, protocol matching parameters, and preprocessed multi-dimensional data are input into the cross-modal attention fusion model: the location information is converted into a 128-dimensional vector through the spatial encoding layer, the protocol parameters are mapped into a 64-dimensional vector through the feature embedding layer, and the preprocessed data extracts the time series features through the LSTM network and compresses them into a 256-dimensional vector. The three are input into the multi-head attention layer to calculate the correlation weights between different modalities, and the key features are dynamically aggregated and then the multi-modal fusion feature vector is output.

[0024] Further, in a preferred embodiment of the present invention, the vehicle dynamics of the monitored target vehicle during driving are used to obtain on-vehicle dynamic monitoring information, and real-time driving scenario analysis is performed based on the on-vehicle dynamic monitoring information to obtain driving scenario analysis information, which specifically includes: Based on the on-vehicle multi-source sensor array, the vehicle dynamics of the monitored target vehicle during driving are used to obtain on-vehicle dynamic monitoring information, and the on-vehicle dynamic monitoring information includes target vehicle driving environment data and target vehicle driving state data; According to the on-vehicle dynamic monitoring information, the target vehicle driving state data is extracted and subjected to time series processing to obtain a driving state time series sequence. The driving state time series sequence is imported into a pre-trained temporal convolutional network for deep feature mining. An extended causal convolutional structure is used to analyze the temporal motion state of the target vehicle by stacking multiple residual blocks, and a motion state feature vector is generated; According to the on-vehicle dynamic monitoring information, the target vehicle driving environment data is extracted and imported into an environment perception model constructed with YOLOv5 as the framework for analysis to identify the road type and traffic density. The relative motion speed of the front target is calculated by combining the optical flow method, and a traffic condition score is generated to obtain a visual state feature vector; The motion state feature vector and the visual state feature vector are input into a cross-modal attention fusion module. The temporal motion feature is used as the query vector, and the visual feature is used as the key-value vector. The cross-modal correlation weight matrix is calculated through a multi-head attention mechanism, and a strongly correlated feature combination is selected. Temporal dependence modeling is performed through a gated recurrent unit to generate a joint representation vector; The navigation information of the target vehicle is obtained through the in-vehicle system of the target vehicle, the navigation distance feature of the current navigation path is extracted for driving duration estimation, and a driving feature vector is constructed by combining the driving duration and the navigation distance; A big data network is introduced to obtain historical driving scenario features of different driving scenario categories. Based on the joint representation vector and the driving feature vector, the current driving scenario portrait of the target vehicle is constructed and compared with the historical driving scenario features to analyze the current driving scenario category of the target vehicle and obtain driving scenario analysis information.

[0025] It should be noted that vehicle dynamic data is collected in real time through an in-vehicle sensor array (including an inertial measurement unit, wheel speed sensors, a steering angle sensor, and a camera): the inertial measurement unit provides three-axis acceleration and angular velocity, the wheel speed sensors obtain the rotational speeds of each wheel, the steering angle sensor records the steering wheel angle, and the camera captures the image of the road ahead and the positions of surrounding vehicles. After the above data is fused by Kalman filtering, vehicle driving state data (such as longitudinal speed, lateral acceleration, yaw angular velocity) and driving environment data (such as lane line curvature, distance to the vehicle ahead, traffic sign type) are output to form complete in-vehicle dynamic monitoring information. For the driving state data, the system divides parameters such as speed, acceleration, and steering angle by a sliding window according to a preset time window to construct a time series data set; then it is input into a pre-trained temporal convolutional network (TCN), which adopts a dilated causal convolutional structure (the dilation factor grows exponentially, such as 1, 2, 4, 8), and extracts multi-scale temporal features layer by layer through stacking 6 residual blocks (each block contains a convolutional layer, weight normalization, ReLU activation, and skip connection). The network compresses the temporal dimension through global average pooling at the output end to generate a motion state feature vector, which characterizes the dynamic behavior patterns of the vehicle under conditions such as hard acceleration, cornering, and frequent starts and stops.

[0026] Furthermore, for driving environment data, the front camera video stream is input into the environment perception module with the YOLOv5s lightweight model as the core: The model extracts road image features through the CSPDarknet backbone network and uses PANet for multi-scale fusion to output object detection results (such as the bounding boxes and categories of vehicles, pedestrians, and traffic signs); at the same time, the Lucas-Kanade optical flow method is applied to consecutive frame images to calculate the pixel displacement vector of the front target, and the relative motion speed is estimated by combining camera calibration parameters and vehicle speed. The road type recognition module counts the number of lane lines, traffic sign density, and the distribution of surrounding vehicles based on the detection results and outputs a road type score; the traffic density is calculated by normalizing the number of vehicles detected per unit time. Finally, the traffic condition score ranging from 0 to 100 is generated by fusing the rate of speed change and road complexity and encoded into a multi-dimensional visual state feature vector through a fully connected layer. Subsequently, the motion state and visual features are deeply fused through a cross-modal attention mechanism: The motion feature vector is used as the query vector Q, and the visual feature vector is used as the key-value vectors K and V and input into an 8-head attention layer. Each attention head projects Q and K into a multi-dimensional subspace, calculates the attention weight matrix through scaled dot product, and filters out strongly correlated feature combinations; after the multi-head outputs are concatenated and normalized by LayerNorm, they are input into a gated recurrent unit for temporal dependence modeling, and finally a multi-dimensional joint representation vector is generated to capture the spatio-temporal coupling relationship of "vehicle dynamics - environment state". Finally, navigation information is obtained through the in-vehicle computer, and features such as the remaining driving distance, path curvature, and speed limit data are extracted, and the estimated driving duration is calculated by combining the real-time vehicle speed. The driving duration, remaining distance, and path complexity (curvature integral) are input into a multi-layer perceptron to generate a driving feature vector. Based on the joint representation vector and the driving feature vector, a portrait of the current driving scenario of the target vehicle is constructed, compared with the historical driving scenario features, and the current scenario category is analyzed through similarity calculation, and the driving scenario analysis information is output to provide the scenario context for the dynamic optimization of the charging strategy.

[0027] Furthermore, in a preferred embodiment of the present invention, the real-time battery state feature and historical charging behavior feature of the device to be charged are extracted based on the multi-modal fusion feature of the device to be charged, and battery health analysis is performed on the target device to be charged to obtain battery health analysis information, specifically including: The real-time battery state feature and historical charging behavior feature of the device to be charged are extracted through the multi-modal fusion feature of the device to be charged to obtain the first feature extraction information, and the first feature extraction information includes remaining power, voltage, current, temperature, internal resistance, number of cycles, and historical charge and discharge curves; A standardized battery state matrix of the device to be charged is constructed based on the first feature extraction information, and a hierarchical evaluation strategy is adopted for battery health analysis, and the standardized battery state matrix is imported into an extended Kalman filter; Define the state vector as the battery capacity attenuation rate and the internal resistance growth rate, and the observation vector as the real-time voltage-current curve. Construct the state equation and perform discretization processing to transform it into a state transition matrix. Combine the real-time charge and discharge data to iteratively update the Jacobian matrix, and generate the estimated value of the battery SOH of the device to be charged; Subsequently, import the standardized battery state matrix as input into the LSTM network. Use the time embedding layer to convert the historical charge and discharge cycle data into a time series-aware feature vector as the initial feature, perform long-term capacity attenuation prediction, and output a probabilistic SOH degradation curve; Use the Bayesian fusion algorithm to perform weighted integration on the SOH estimated value and the probabilistic SOH degradation curve. Take the real-time SOH estimated value as the prior distribution and the long-term prediction probability as the likelihood function, and calculate the posterior distribution based on Bayes' theorem; Solve the expected value and confidence interval of the posterior distribution, introduce variational inference to accelerate the calculation, and use the KL divergence to minimize the difference between the approximate posterior and the true distribution. Finally, output the comprehensive SOH score of the battery of the device to be charged to obtain the battery health analysis information.

[0028] It should be noted that in the battery health analysis stage, based on multi-modal fusion features, real-time battery state parameters of the device to be charged (including remaining power, voltage, current, temperature, internal resistance, etc.) and historical charging behavior data (such as cycle times, charge and discharge rate distributions, historical capacity attenuation records, etc.) are used to construct the first feature extraction information including the time series dimension. First, parameters such as voltage, current, and temperature are normalized. After eliminating the dimension differences, the time series data is sliced by a fixed window to form a standardized battery state matrix. A hierarchical evaluation strategy is adopted: at the bottom layer, an extended Kalman filter (EKF) is used for real-time state estimation, and at the top layer, an LSTM network is used to perform long-term capacity attenuation prediction. In the real-time layer, the state vector is defined as the battery capacity attenuation rate and the internal resistance growth rate, and the observation vector is the voltage-current dynamic response in the real-time charge and discharge curve. A non-linear state equation is established to describe the evolution relationship between the capacity and the internal resistance, and the state equation is discretized by a first-order Taylor expansion to be transformed into a state transition matrix. The observation noise covariance matrix is updated through the real-time collected voltage-current data, and the Jacobian matrix partial derivative is iteratively calculated to correct the prediction error covariance, and finally the estimated values of the battery capacity attenuation rate and the internal resistance growth rate at the current moment are output. In the long-term prediction layer, the standardized battery state matrix is input into a bidirectional LSTM network: the network maps the historical charge and discharge cycle data into time series feature vectors through a time embedding layer, uses the forget gate and input gate mechanisms to screen key historical nodes, and captures the long-term dependence relationship of capacity attenuation to generate the SOH probability distribution curve within future cycles. The Bayesian fusion algorithm is introduced: the SOH estimated value output by the EKF is used as the prior distribution, and the probability curve predicted by the LSTM is used as the likelihood function, and the posterior distribution is calculated based on Bayes' theorem. The variational inference approximation is adopted: it is assumed that the posterior follows a Gaussian distribution, the true distribution is approximated by minimizing the KL divergence, and the variational parameters are optimized using stochastic gradient descent, and finally the comprehensive SOH score and the deterioration trend warning are output. This fusion mechanism combines the sensitivity of real-time data with the stability of long-term prediction, providing a high-confidence health state input for the dynamic charging strategy.

[0029] Furthermore, in a preferred embodiment of the present invention, the construction of the charging strategy knowledge graph, formulating the optimal charging strategy for the target device to be charged in the current driving scenario, and generating a control command in combination with the positioning information of the device to be charged for intelligent wireless charging management specifically include: Using the big data network to obtain historical wireless charging instances in different driving scenarios, and extracting features from each historical wireless charging instance to obtain the driving scenario features, charging device condition features, and charging strategy features corresponding to each historical wireless charging instance; Taking the driving scenario features as the initial classification basis, initially classifying each historical wireless charging instance to generate several category sets, and calculating the Euclidean distance values between the corresponding several historical wireless charging instances within each category set based on the charging device condition features; Based on the calculated Euclidean distance values, the K-mean clustering algorithm is used to further classify each category set, generating several category subsets, and associating each category subset with the corresponding driving scenario features to obtain the category division result information; Based on the category division result information, entity triples are constructed according to the driving scenario features, charging device status features, and charging strategy features, and a charging strategy knowledge graph is constructed through the entity triples; The battery health analysis information, driving scenario parsing information, and multi-modal fusion features of the device to be charged are obtained and input into the charging strategy knowledge graph to analyze the appropriate charging strategy for the current driving scenario and charging device status; Based on the multi-modal fusion features of the device to be charged, charging protocol adaptation constraint conditions are set, and the cosine metric algorithm is used to analyze the matching degree of each charging strategy stored in the charging strategy knowledge graph with the current charging event, generating several candidate charging strategies; Each candidate charging strategy is screened through the set charging protocol adaptation constraint conditions, the charging strategies that do not conform to the charging protocol supported by the current device to be charged are excluded, the remaining candidate charging strategies are obtained and sorted according to the matching degree with the current charging event, and the one with the highest matching degree is selected as the final charging strategy; The positioning information of the device to be charged is obtained, a wireless charging control instruction is generated through the final charging strategy, and a coil position activation instruction for the wireless charging board is generated in combination with the positioning information of the device to be charged to perform intelligent charging management on the target device to be charged.

[0030] It should be noted that, first, a large amount of historical wireless charging instance data is retrieved from the cloud big data platform, and the driving scenario features (such as road type, vehicle speed distribution, environmental temperature, etc.), charging device status features (such as battery SOH, internal resistance, maximum charging current, etc.) and charging strategy features (such as charging power, frequency, heat dissipation mode, etc.) of each record are extracted to form a multi-dimensional feature vector. Based on the driving scenario features, the historical data is classified at a coarse-grained level to generate an initial category set; subsequently, within each category, the Euclidean distance matrix of the device status features is calculated, and the K-means clustering algorithm is used for fine-grained division. For example, the "highway" scenario is further divided into sub-categories such as "high-temperature long-distance cruise" and "low-temperature short-distance high-speed", and finally a multi-level mapping relationship of scenario-device status-charging strategy is established. A charging strategy knowledge graph is constructed based on the classification results: the driving scenario, device status, and charging strategy are defined as entity nodes in the graph, connected by relationship edges such as "scenario-applicable strategy" and "device status-strategy constraint", and attribute edges are introduced to annotate specific parameters. When a real-time charging task is triggered, the current battery health score, driving scenario analysis result, and device multi-modal features are input into the knowledge graph, and the associated strategy nodes are retrieved through the graph traversal algorithm to generate an initial candidate set containing 20-30 historical strategies. For the candidate strategy set, by setting charging protocol adaptation constraint conditions: based on the protocol support list in the device multi-modal features, a binary protocol compatibility matrix is constructed, and the cosine similarity algorithm is used to calculate the matching degree of each candidate strategy with the current device protocol parameters. After constraint screening, the remaining strategies are sorted by matching degree, and finally the optimal strategy is output. During the strategy execution stage, two-level control instructions are generated according to the device location information and the final charging strategy parameters. The intelligent charging control of in-vehicle wireless charging is realized, that is, while ensuring the charging efficiency, the health status of the charging device is also taken into account.

[0031] Figure 2 Flowchart of an in-vehicle wireless charging intelligent charging control method provided by an embodiment of the present invention; As Figure 2 shown, the present invention provides a flowchart of an in-vehicle wireless charging intelligent charging control method, including: S202, repeatedly scan the in-vehicle wireless charging area to detect and locate the device to be charged. After detecting and locating the device to be charged, determine whether a secure link can be established with the device to be charged, perform charging protocol matching and data collection, and construct a multi-modal fusion feature; S204, obtain the in-vehicle dynamic monitoring information, perform real-time driving scenario analysis based on the in-vehicle dynamic monitoring information, and perform battery health analysis on the target device to be charged through the constructed multi-modal fusion feature; S206, input the battery health analysis information, driving scenario analysis information, and multi-modal fusion features of the device to be charged into the charging strategy knowledge graph to analyze the adaptive charging strategy for the current driving scenario and charging device status; S208, use the cosine metric algorithm to analyze the matching degree between each charging strategy stored in the charging strategy knowledge graph and the current charging event, generate several candidate charging strategies, and select the final charging strategy for each candidate charging strategy through the set charging protocol adaptation constraints; S210, generate a wireless charging control instruction through the final charging strategy, combine the positioning information of the device to be charged to generate a coil position activation instruction for the wireless charging board, and perform intelligent charging management on the target device to be charged.

[0032] Further, in an intelligent management and control method for in-vehicle wireless charging provided by the present invention, the following steps are further included: Obtain the historical charging temperature data of the wireless charging module, perform data preprocessing on the obtained historical charging temperature data and use time series interpolation to fill in the missing values, and construct a first data set according to the preprocessed historical charging temperature data; Extract features from the first data set to obtain a thermal state feature vector, including the battery surface temperature gradient, radiator efficiency coefficient, power loss rate, and environmental thermal resistance factor, to obtain a thermal state feature set; Divide the thermal state of the wireless charging module into a safe state, a warning state, and a critical state, generate a thermal state space based on the thermal state feature set, calculate the state transition probability of different thermal states based on the thermal state space, and construct a thermal state transition matrix; Iteratively optimize the state transition probability matrix and the observation probability matrix through the Baum-Welch algorithm, and use the forward-backward algorithm to calculate the likelihood of the state sequence during the training process until convergence to output the final thermal state transition probability, and construct a thermal state analysis model; Obtain the temperature data of the wireless charging module through a real-time monitoring array, input it into the thermal state analysis model to analyze the temperature change state of the wireless charging module at a future moment, and obtain thermal state analysis result information; Judge the thermal state type of the wireless charging module within a future moment based on the thermal state analysis result information. If there is a critical state within a preset time range, perform heat dissipation regulation based on a preset heat dissipation strategy.

[0033] Figure 3 An intelligent management and control system 3 for in-vehicle wireless charging provided by an embodiment of the present invention, the system includes: a memory 31, a processor 32, the memory 31 contains an intelligent management and control method program for in-vehicle wireless charging, and when the intelligent management and control method program for in-vehicle wireless charging is executed by the processor 32, the following steps are implemented: Repeatedly scan the in-vehicle wireless charging area to obtain the reflected magnetic field signal, detect and locate the device to be charged based on the reflected magnetic field signal, and generate the positioning information of the device to be charged; After detecting and locating the device to be charged, determine whether a secure link can be established with the device to be charged, perform charging protocol matching and data collection, and construct the multi-modal fusion features of the device to be charged; Monitor the vehicle dynamics during the driving of the target vehicle to obtain the in-vehicle dynamic monitoring information, and perform real-time driving scenario analysis based on the in-vehicle dynamic monitoring information to obtain the driving scenario analysis information; Extract the real-time battery state features and historical charging behavior features of the device to be charged based on the multi-modal fusion features of the device to be charged, perform battery health analysis on the target device to be charged, and obtain the battery health analysis information; Construct a charging strategy knowledge graph, formulate the optimal charging strategy for the target device to be charged in the current driving scenario, and generate a control command in combination with the positioning information of the device to be charged for intelligent control of wireless charging.

[0034] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed with each other can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical, or other forms.

[0035] The units described as separate components above may or may not be physically separated, and the components shown as units may or may not be physical units; they can be located in one place or distributed to multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0036] In addition, each functional unit in the embodiments of the present invention can be all integrated in one processing unit, or each unit can be separately used as a unit, or two or more units can be integrated in one unit; the above integrated units can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0037] Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes: removable storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs and other various media that can store program codes.

[0038] Alternatively, if the above integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. And the foregoing storage medium includes: removable storage devices, ROM, RAM, magnetic disks, or optical discs and other various media that can store program codes.

[0039] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An intelligent control method applied to in-vehicle wireless charging, characterized in that, Including: Repeatedly scan the in-vehicle wireless charging area to obtain a reflected magnetic field signal, detect and locate the device to be charged based on the reflected magnetic field signal, and generate the positioning information of the device to be charged; After detecting and locating the device to be charged, determine whether a secure link can be established with the device to be charged, perform charging protocol matching and data acquisition, and construct the multi-modal fusion features of the device to be charged; Monitor the vehicle dynamics during the driving of the target vehicle to obtain the vehicle-side dynamic monitoring information, and perform real-time driving scenario analysis based on the vehicle-side dynamic monitoring information to obtain the driving scenario analysis information; Extract the real-time battery state characteristics and historical charging behavior characteristics of the device to be charged based on the multi-modal fusion features of the device to be charged, and perform battery health analysis on the target device to be charged to obtain the battery health analysis information; Construct a charging strategy knowledge graph, formulate the optimal charging strategy for the target device to be charged in the current driving scenario, and generate a control command in combination with the positioning information of the device to be charged for intelligent management and control of wireless charging.

2. The intelligent control method for vehicle-mounted wireless charging according to claim 1, characterized in that, The step of repeatedly scanning the in-vehicle wireless charging area to obtain a reflected magnetic field signal, detecting and locating the device to be charged based on the reflected magnetic field signal, and generating the positioning information of the device to be charged specifically includes: The electromagnetic induction coil array built in the wireless charging board is used to scan the charging area by using a time-sharing activation strategy, and high-frequency signals are sequentially sent to the coils at different positions to obtain the reflected magnetic field signal; Import the obtained reflected magnetic field signal into a high-precision analog-to-digital converter for conversion to obtain a reflected magnetic field digital signal, preprocess the reflected magnetic field digital signal by using a digital filtering algorithm, and generate a signal intensity matrix according to the preprocessed reflected magnetic field digital signal; Extract the intensity characteristics of the reflected magnetic field signal at the current moment according to the signal intensity matrix. If the intensity of the reflected magnetic field signal at the current moment is greater than the preset signal intensity threshold and the number of scans of the corresponding scanned area is greater than the preset scan number threshold, it is determined that there is a device to be charged in the target area and it is marked as a locked area; When a device to be charged is detected on the target wireless charging board, obtain the reflected magnetic field signal of the locked area, select the three coils with the optimal geometric distribution in the locked area as reference nodes, and calculate the distance estimation values from the device to be charged to each node based on the path loss model; Construct a non-linear geometric equation set according to the reference node coordinates and the distance estimation values, convert it into an over-determined linear equation set through linearization processing, and use the weighted least squares method to solve the optimal two-dimensional plane coordinates to generate the positioning information of the device to be charged.

3. The intelligent control method applied to in-vehicle wireless charging according to claim 1, wherein, The step of, after detecting and locating the device to be charged, determining whether a secure link can be established with the device to be charged, performing charging protocol matching and data acquisition, and constructing the multi-modal fusion features of the device to be charged specifically includes: After detecting and locating the device to be charged, obtain the unique identifier stored on the vehicle side, convert the unique identifier through a hash function and use it as the key derivation seed input of the elliptic curve encryption algorithm to generate a public-private key pair on the vehicle side and send it to the device to be charged side for key exchange; After completing the key exchange, the vehicle-mounted end uses the AES-GCM encryption algorithm to generate a challenge code and sends it to the device-end through a secure channel. The device-end accepts the challenge code, signs it, and sends it back to the vehicle-mounted end for identity authentication. If the identity authentication result of the vehicle-mounted end to the device to be charged is successfully verified within the preset number of verifications, a secure link is established with the device to be charged and the charging protocol characteristic parameters of the device to be charged are obtained through the encrypted channel; Input the acquired charging protocol characteristic parameters into the preset charging protocol library and perform encoding matching with the pre-built hash tree structure, traverse the hash tree nodes and compare them with the characteristic hash values ​​in the preset protocol library, output the protocol matching results and corresponding working mode parameters, and obtain the charging protocol matching information; The vehicle-mounted end obtains real-time battery status data and historical charging behavior data of the device-end user from the device-end through the established secure link, and obtains the user's vehicle-mounted charging behavior data from the vehicle-mounted database to generate first data collection information; Calculating the data standard deviation of the first data acquisition information, eliminating the corresponding data that is not within the preset standard deviation range based on the calculated data standard deviation, supplementing it through linear interpolation, and obtaining data preprocessing information; The positioning information of the device to be charged is obtained, and combined with the charging protocol matching information and the data preprocessing information, an attention mechanism is introduced to perform multimodal feature fusion to obtain the multimodal fusion features of the device to be charged.

4. An intelligent control method applied to in-vehicle wireless charging according to claim 1, characterized in that, The monitoring target vehicle acquires vehicle-mounted dynamic monitoring information during driving, and performs real-time driving scene analysis based on the vehicle-mounted dynamic monitoring information to obtain driving scene analysis information, specifically including: Based on the on-board multi-source sensor array, the vehicle dynamics of the target vehicle during driving are monitored to obtain on-board dynamic monitoring information, wherein the on-board dynamic monitoring information includes driving environment data of the target vehicle and driving state data of the target vehicle; Extracting the target vehicle driving state data according to the vehicle-mounted dynamic monitoring information and performing time series processing to obtain a driving state time series sequence, importing the driving state time series sequence into a pre-trained time series convolutional network for deep feature mining, and using an extended causal convolutional structure to stack multiple residual blocks to analyze the time series motion state of the target vehicle, and generate a motion state feature vector; Extract the target vehicle driving environment data based on the vehicle-mounted dynamic monitoring information, import it into the environment perception model built with YOLOv5 as the framework for analysis, identify the road type and traffic density, calculate the relative movement speed of the target ahead by combining the optical flow method, generate a traffic condition score, and obtain a visual state feature vector; The motion state feature vector and the visual state feature vector are input into a cross-modal attention fusion module, the temporal motion feature is used as a query vector, the visual feature is used as a key value vector, a cross-modal association weight matrix is ​​calculated through a multi-head attention mechanism, a strong correlation feature combination is screened out, and temporal dependency modeling is performed through a gated recurrent unit to generate a joint representation vector; Obtain the navigation information of the target vehicle through the in-vehicle system of the target vehicle, extract the navigation distance feature of the current navigation path for driving duration estimation, and construct a driving feature vector by combining the driving duration and the navigation distance; Introduce a big data network to obtain historical driving scenario features of different driving scenario categories, construct a portrait of the current driving scenario of the target vehicle based on the joint representation vector and the driving feature vector, compare it with the historical driving scenario features, analyze the current driving scenario category of the target vehicle, and obtain driving scenario analysis information.

5. The intelligent control method applied to in-vehicle wireless charging according to claim 1, wherein Extract the real-time battery state feature and historical charging behavior feature of the target charging device based on the multi-modal fusion feature of the charging device, and perform battery health analysis on the target charging device to obtain battery health analysis information, specifically including: Extract the real-time battery state feature and historical charging behavior feature of the charging device through the multi-modal fusion feature of the charging device to obtain the first feature extraction information, where the first feature extraction information includes remaining power, voltage, current, temperature, internal resistance, cycle times, and historical charge and discharge curves; Construct a standardized battery state matrix of the charging device based on the first feature extraction information, adopt a hierarchical evaluation strategy for battery health analysis, and import the standardized battery state matrix into an extended Kalman filter; Define the state vector as the battery capacity attenuation rate and the internal resistance growth rate, the observation vector as the real-time voltage-current curve, construct a state equation and perform discretization processing to transform it into a state transition matrix, and iteratively update the Jacobian matrix in combination with real-time charge and discharge data to generate an estimated value of the battery SOH of the charging device; Subsequently, import the standardized battery state matrix as an input into an LSTM network, convert the historical charge and discharge cycle data into a time-series perception feature vector using a time embedding layer as the initial feature, perform long-cycle capacity attenuation prediction and output a probabilistic SOH degradation curve; Weightedly integrate the SOH estimated value and the probabilistic SOH degradation curve using the Bayesian fusion algorithm, use the real-time SOH estimated value as the prior distribution, the long-cycle prediction probability as the likelihood function, and calculate the posterior distribution based on Bayes' theorem; Solve the expected value and confidence interval of the posterior distribution, introduce variational inference to accelerate the calculation, use the KL divergence to minimize the difference between the approximate posterior and the true distribution, and finally output the comprehensive SOH score of the battery of the charging device to obtain battery health analysis information.

6. The intelligent control method applied to in-vehicle wireless charging according to claim 1, wherein, Construct a charging strategy knowledge graph, formulate the optimal charging strategy for the target charging device in the current driving scenario, and generate a control command in combination with the positioning information of the charging device for intelligent management and control of wireless charging, specifically including: Use a big data network to obtain historical wireless charging instances in different driving scenarios, extract features from each historical wireless charging instance to obtain the driving scenario features, charging device condition features, and charging strategy features corresponding to each historical wireless charging instance; Use the driving scenario features as the initial classification basis, perform initial classification on each historical wireless charging instance to generate several category sets, and calculate the Euclidean distance values between the corresponding historical wireless charging instances within each category set based on the charging device condition features; Based on the calculated Euclidean distance values, the K-mean clustering algorithm is used to further classify each category set, generating several category subsets. The corresponding driving scenario features are associated with each category subset to obtain the category division result information; Based on the category division result information, entity triples are constructed according to the driving scenario features, charging device status features, and charging strategy features. The charging strategy knowledge graph is constructed through the entity triples; Obtain the battery health analysis information, driving scenario analysis information, and multi-modal fusion features of the device to be charged, and input them into the charging strategy knowledge graph to analyze the charging strategy suitable for the current driving scenario and charging device status; Based on the multi-modal fusion features of the device to be charged, set the charging protocol adaptation constraint conditions, and use the cosine metric algorithm to analyze the matching degree of each charging strategy stored in the charging strategy knowledge graph with the current charging event, generating several candidate charging strategies; Screen each candidate charging strategy through the set charging protocol adaptation constraint conditions, eliminate the charging strategies that do not conform to the charging protocol supported by the current device to be charged, obtain the remaining candidate charging strategies, sort them according to the matching degree with the current charging event, and select the one with the highest matching degree as the final charging strategy; Obtain the positioning information of the device to be charged, generate a wireless charging control instruction through the final charging strategy, and combine the positioning information of the device to be charged to generate a coil position activation instruction for the wireless charging board to perform intelligent charging management on the target device to be charged.

7. An intelligent control system applied to in-vehicle wireless charging, characterized in that, The system includes: a memory and a processor. The memory contains an intelligent management method program for in-vehicle wireless charging. When the intelligent management method program for in-vehicle wireless charging is executed by the processor, the following steps are implemented: Repeatedly scan the in-vehicle wireless charging area to obtain the reflected magnetic field signal. Based on the reflected magnetic field signal, detect and locate the device to be charged, generating the positioning information of the device to be charged; After detecting and locating the device to be charged, determine whether a secure link can be established with the device to be charged, perform charging protocol matching and data collection, and construct the multi-modal fusion features of the device to be charged; Monitor the vehicle dynamics of the target vehicle during driving to obtain the in-vehicle dynamic monitoring information. Based on the in-vehicle dynamic monitoring information, perform real-time driving scenario analysis to obtain the driving scenario analysis information; Extract the real-time battery state features and historical charging behavior features of the device to be charged based on the multi-modal fusion features of the device to be charged, and perform battery health analysis on the target device to be charged to obtain the battery health analysis information; Construct a charging strategy knowledge graph, formulate the optimal charging strategy for the target device to be charged in the current driving scenario, and generate a control instruction in combination with the positioning information of the device to be charged to perform intelligent management of wireless charging.

Images