Remote monitoring system and method applied to electric vehicle lithium battery state estimation

Abstract

The invention discloses a remote monitoring system and method applied to electric vehicle lithium battery state estimation, and belongs to the technical field of lithium battery state monitoring, and the method comprises the steps: obtaining a current charging scene parameter through the interaction information of a charging facility and a vehicle terminal, and constructing a charging scene feature set; the method comprises the following steps: acquiring multi-dimensional state data of an electric vehicle lithium battery in a charging mode based on a vehicle-mounted sensor, and constructing an association mapping library of a charging scene and lithium battery state change; the charging state change of the lithium battery in different charging scenes is analyzed, a lithium battery state evaluation model is constructed, and the abnormal risk of the lithium battery is evaluated; according to an evaluation result, dynamically adjusting a data sampling frequency, and setting an anti-interference frequency modulation sequence; dynamically adjusting the data transmission frequency of the vehicle end and the remote monitoring terminal according to the charging scene and the network environment; and after the monitoring terminal receives the data, the health state trend of the lithium battery is predicted by combining the association mapping library and the risk assessment result, and the health state of the lithium battery is pre-warned.

Description

Technical Field

[0001] This invention relates to the field of lithium battery state monitoring technology, specifically to a remote monitoring system and method for estimating the state of lithium batteries in electric vehicles. Background Technology

[0002] With the increasing popularity of electric vehicles, charging scenarios are becoming more diverse, including home slow charging, public fast charging, and wireless charging. In these different charging scenarios, key parameters such as the current and temperature changes of electric vehicle lithium batteries vary significantly. For example, in fast charging scenarios, the charging power is relatively high, and the current and temperature of the lithium battery change rapidly, requiring high-frequency data sampling to accurately capture the battery status. In wireless charging scenarios, the complex electromagnetic environment may interfere with the accuracy of the sensors used for monitoring, affecting the reliability of the monitoring data.

[0003] However, existing monitoring technologies mostly employ fixed monitoring strategies, meaning that parameters such as sampling frequency and alarm thresholds are preset regardless of changes in the charging scenario, making it difficult to flexibly adapt to the needs of different charging scenarios. For example, in fast charging scenarios, this fixed strategy may lead to a conflict between high-frequency data requirements and communication bandwidth. Due to limited communication bandwidth, the real-time transmission of high-frequency data is constrained, easily resulting in the omission of critical transient data, which in turn affects the accuracy of fault warnings. In wireless charging scenarios, due to the presence of electromagnetic interference, traditional monitoring systems lack targeted anti-interference mechanisms, leading to a decline in monitoring data quality and making it difficult to accurately reflect the actual state of lithium batteries. Summary of the Invention

[0004] The purpose of this invention is to provide a remote monitoring system and method for estimating the state of lithium batteries in electric vehicles, in order to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a remote monitoring method for estimating the state of lithium batteries in electric vehicles, comprising: Step S100: Obtain relevant parameters of the current charging scenario through the interaction information between the charging facility and the vehicle terminal, and construct a charging scenario feature set; at the same time, based on various vehicle sensors, collect multi-dimensional state data of electric vehicle lithium batteries in real time under the charging mode, and construct a correlation mapping library between the charging scenario and the state changes of lithium batteries. Step S200: Based on the constructed association mapping library, analyze the changes in the charging state of the lithium battery under different charging scenarios, construct a lithium battery state assessment model, and assess whether there is a risk of abnormal state of the lithium battery under the current charging scenario. Step S300: Based on the state risk assessment results of the lithium battery, dynamically adjust the data sampling frequency under each charging scenario; set different anti-interference frequency modulation sequences, and dynamically adjust the data transmission frequency between the vehicle and the remote monitoring terminal according to the current charging scenario and network environment; Step S400: After receiving the transmitted data, the monitoring terminal combines the associated mapping library and the state risk assessment results to predict the health status trend of the lithium battery in the current charging scenario, comprehensively assess the health status of the lithium battery, and issue a health status warning for the lithium battery.

[0006] Furthermore, step S100 includes: Step S101: After the electric vehicle connects to the charging facility, the charging facility and the vehicle's battery management system establish a connection through a communication protocol, collect the current charging power, record the current charging mode P, and convert it into a binary vector using one-hot encoding; simultaneously, the grid voltage V is monitored in real time by the voltage sensor built into the charging facility. grid And the fluctuation range ΔV, stored as the interval [V grid -ΔV, V grid +ΔV]; Combined with the onboard environmental sensor to obtain the ambient temperature T env Furthermore, the load rate Y is obtained by calculating the ratio of the current charging power to the rated power of the charging facility; the collected variables are normalized to map the data to the [0, 1] interval; all features are integrated into a charging scenario feature vector F=[P,V] grid ,T env [,Y]; Step S102: During the charging process, the on-board sensor network collects the state parameters of the lithium battery in real time, sets the initial sampling frequency, monitors the internal temperature T of the battery pack through the temperature sensor, and obtains the voltage V of the individual cells through the voltage sensor. i and the total voltage V of the battery pack total V i The voltage of each low-i individual cell in the battery pack is represented; the current sensor measures the charging current I through the Hall effect; simultaneously, the battery management system calculates the state of charge (SOC) of the battery based on the coulomb counting method and calculates the internal resistance R using the AC impedance spectroscopy method; all state parameters are synchronized and aligned using timestamps, with an allowable time deviation of no more than 1ms; finally, all state parameters are integrated into the lithium battery state vector S. t =[V i V total ,I,SOC,R],S t This represents the complete state of the battery at time t; Step S103: For each charging process, the collected scene feature vector F and the corresponding state vector S are compared. t Pair them according to time sequence to form key-value pairs (F,S) t); Store all data pairs as a dataset, forming an association mapping library D: D={(F k ,S t,k )} N k=1 Where N represents the total number of data pairs, F k S represents the scene feature vector of the k-th charging session. t,k This represents the lithium battery state vector at time t during the k-th charge; after each charge, newly collected data pairs are automatically appended to the associated mapping library D.

[0007] Furthermore, step S200 includes: Step S201: Process the state vector sequence {S} in the associated database. t,k} Perform time-series difference calculation, and approximate the calculation by using the difference between adjacent sampling points: ΔS t =S t,k -S t-1,k ;where S t-1,k Let {ΔS} represent the state vector of the lithium battery at time t-1 during the k-th charge, and obtain the characteristic sequence of the rate of change of state. t,k The feature sequences are standardized, and the number of clusters M is determined using the silhouette coefficient method. The M value with the largest silhouette coefficient is selected, and the K-Means algorithm is used to cluster the state change features to divide the charging behavior patterns, resulting in M ​​cluster centers {C}. m} M m=1 C m C represents the center vector of the m-th cluster; each center C m The vector is a continuous numerical vector representing the feature distribution of a normal charging mode, and a unique label L is assigned to each cluster center. m L m Let represent the category label of the m-th cluster. The label is a discrete integer used to identify the cluster category to which the data point belongs; for any state change rate feature ΔS t Labels are obtained by clustering clusters using nearest neighbor matching; the cluster centers and labels are integrated into a charging pattern library C={(C m L m )} M m=1 ; Step S202: Collect historical data over a period of time, and construct a training set by labeling outlier samples. Each sample includes: a scene feature vector F, and a lithium battery state change rate feature ΔS. t and clustering label L m ;Discrete label L mThe data is converted into one-hot encoding and concatenated with numerical features to form the input vector X. An XGBoost gradient boosting tree is used to construct a lithium battery state assessment model. The input is X, and the output y∈{0,1} represents whether the state is normal (0 for normal, 1 for abnormal). During training, the dataset is divided into training, validation, and test sets according to a set ratio. The model classification performance is optimized by minimizing the cross-entropy loss function. A preset number of training rounds is set, and training is terminated when the validation set loss does not decrease for e consecutive rounds. The model performance is evaluated using the test set based on accuracy and F1 score. Step S203: During the charging process, the current scene feature F and the state change rate ΔS are compared. t Input the trained model, calculate the Euclidean distance to the cluster centers, match the nearest cluster label, and output the anomaly probability G at time t. t When the anomaly probability G t Exceeding the preset threshold θ G When the time comes, an alert will be triggered immediately, and a report containing a timestamp, scene parameters, anomaly probability, and anomaly type will be automatically generated.

[0008] Furthermore, step S300 also includes: Step S301: Based on the anomaly probability output by the model, the state risk of the lithium battery is divided into three levels: when G t <=θ G1 When θ is set to low risk, it indicates a normal state, and the initial sampling frequency is maintained; when θ G1 <G t <=θ G2 When the risk is set to medium, indicating a potential anomaly in the state, the sampling frequency is increased to f1 = k1 × f0, where k1 > 1; when G t >θ G2 When the risk is set to high, indicating an abnormal state, the sampling frequency is increased to f2 = k2 × f0, where k2 > k1; θ G1 θ G2 The preset risk threshold is defined, k1 and k2 are frequency adjustment coefficients, and f0 represents the initial sampling frequency; during the charging process, the abnormal probability G is monitored. t The sampling frequency of the vehicle-mounted sensors is dynamically adjusted according to the risk level: when G t When the risk level rises from low to medium, the sampling frequency is switched from f0 to f1; when G t When the risk level rises from medium to high, the sampling frequency is switched from f1 to f2; when G t When the risk level drops back to low, gradually restore the initial sampling frequency f0; Step S302: According to the communication protocol, the vehicle terminal and the remote monitoring terminal set the minimum communication frequency f. min and maximum communication frequency f maxBased on the electromagnetic ambient noise intensity, a pseudo-random frequency modulation sequence is generated using a linear congruent generator: Q={q1,q2,...,q} a}, where q a This represents the a-th communication frequency, where a represents the sequence length, and q... a ∈[f min f max The data transmission quality index Z is calculated based on packet loss and delay: Z = α × (1 - PLR) + β × (1 / U); where PLR ​​represents the packet loss rate, which is the proportion of lost data packets counted through the ACK confirmation mechanism, and U represents the delay parameter, which records the time difference between sending and receiving data packets; α and β represent the adjustment weights of the packet loss rate and the delay parameter, respectively; the frequency that maximizes Z is selected from the sequence Q as the data transmission frequency for the current charging scenario, and the frequencies are sorted according to the Z value, with the top n selected as the candidate frequency group; based on the selected transmission frequency, the data packets for the current charging scenario are constructed using vector F, lithium battery state vector St, anomaly probability, and clustering labels and sent to the remote monitoring terminal.

[0009] Furthermore, step S400 includes: Step S401: The monitoring terminal receives real-time data packets transmitted by the vehicle terminal, performs timestamp alignment, missing value imputation, and normalization on the received data; an LSTM network is used to construct a lithium battery health status prediction model; the real-time scene features F and the state vector S are then compared. t The model is fused to form a comprehensive feature vector as input. The hidden layer consists of two LSTM units and one fully connected layer, and the output layer has two neurons, each corresponding to the output of the lithium battery's capacity decay rate ΔC. t and internal resistance growth rate ΔR t Historical charging data is extracted from the association mapping library, and samples are divided according to time windows to form input sequences and label sequences. The Adam optimizer and mean squared error loss function are used to update the model parameters through backpropagation. Training is terminated when the validation set loss does not decrease for e consecutive rounds. The real-time feature vector is input into the trained model to obtain the predicted values ​​of capacity decay rate and internal resistance growth rate at the current moment. Based on the current predicted values ​​and charging scenario parameters, the rolling prediction method is used to predict the trend of lithium battery health status within the future Tp time period and generate a trend curve. Step S402: Combining the prediction results and real-time status data, calculate the health status assessment index of the lithium battery: Use an exponential decay model to estimate the remaining lifespan of the lithium battery based on the predicted capacity decay rate; calculate the health index of the lithium battery based on the internal resistance growth rate; compare the obtained results with a preset threshold; if the calculated result is less than the preset threshold, the lithium battery is determined to be in an unhealthy state; compare the health assessment results with the anomaly probability G. tThe system integrates and calculates a comprehensive risk index; sets an early warning threshold; when the comprehensive risk index exceeds the preset threshold, an early warning is triggered, generating early warning information, including a timestamp, scenario parameter F, health status indicators, predicted trends, anomaly probabilities, and early warning prompts, which is then sent back to the monitoring terminals of relevant personnel; relevant information from this charging process is appended to the association mapping library D for updating and storage. This remote monitoring system for estimating the state of lithium batteries in electric vehicles includes an association mapping module, a lithium battery state assessment module, a dynamic sampling and data transmission module, and a health status prediction module. The association mapping module obtains relevant parameters of the current charging scenario through the interaction information between the charging facility and the vehicle terminal, and constructs a charging scenario feature set; at the same time, based on various vehicle sensors, it collects multi-dimensional state data of electric vehicle lithium batteries in real time under the charging mode, and constructs an association mapping library between the charging scenario and the state changes of lithium batteries. The lithium battery state assessment module analyzes the changes in the charging state of the lithium battery under different charging scenarios based on the constructed association mapping library, constructs a lithium battery state assessment model, and assesses whether there is a risk of abnormal state of the lithium battery under the current charging scenario. The dynamic sampling and data transmission module dynamically adjusts the data sampling frequency under various charging scenarios based on the state risk assessment results of the lithium battery; it sets different anti-interference frequency modulation sequences and dynamically adjusts the data transmission frequency between the vehicle and the remote monitoring terminal according to the current charging scenario and network environment. After receiving the transmitted data at the monitoring terminal, the health status prediction module combines the associated mapping library and the status risk assessment results to predict the health status trend of the lithium battery in the current charging scenario, comprehensively assess the health status of the lithium battery, and issue a health status warning for the lithium battery.

[0010] Compared with the prior art, the beneficial effects achieved by the present invention are: This invention obtains charging scenario parameters through the interaction between charging facilities and vehicle terminals, and combines on-board sensors to collect multi-dimensional state data of lithium batteries in real time. It constructs a correlation mapping library between charging scenarios and changes in lithium battery state, realizing comprehensive capture and recording of changes in lithium battery state. This not only provides a data foundation for subsequent state assessment, but also shows the intrinsic relationship between charging scenarios and lithium battery state through the correlation mapping library, providing strong support for subsequent assessment of lithium battery state, thereby improving the accuracy and reliability of state monitoring. Based on the constructed association mapping library, this invention analyzes the state changes of lithium batteries under different charging scenarios and constructs a state assessment model to assess the risk of anomalies. According to the assessment results, the data sampling frequency and data transmission frequency are dynamically adjusted, which not only ensures the real-time and accuracy of key data, but also effectively reduces the redundancy and energy consumption of data transmission. This dynamic adjustment mechanism not only improves the monitoring efficiency, but also enhances the adaptability of the monitoring system to different charging scenarios and network environments, making remote monitoring more flexible and efficient. This invention constructs a health status prediction model by combining a monitoring terminal with an associated mapping library and state risk assessment results, enabling the prediction and timely early warning of the future health status of lithium batteries. At the same time, the early warning information is sent back to the monitoring terminal in real time, ensuring that problems can be detected and dealt with in a timely manner, thereby effectively extending the service life of lithium batteries, improving the safety and reliability of electric vehicles, and realizing remote and efficient monitoring and early warning of the status of electric vehicle lithium batteries. Attached Figure Description

[0011] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a remote monitoring method applied to the state estimation of lithium batteries in electric vehicles. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Please see Figure 1 This invention provides a technical solution: a remote monitoring method for estimating the state of lithium batteries in electric vehicles, comprising: Step S100: Obtain relevant parameters of the current charging scenario through the interaction information between the charging facility and the vehicle terminal, and construct a charging scenario feature set; at the same time, based on various vehicle sensors, collect multi-dimensional state data of electric vehicle lithium batteries in real time under the charging mode, and construct a correlation mapping library between the charging scenario and the state changes of lithium batteries. Step S200: Based on the constructed association mapping library, analyze the changes in the charging state of the lithium battery under different charging scenarios, construct a lithium battery state assessment model, and assess whether there is a risk of abnormal state of the lithium battery under the current charging scenario. Step S300: Based on the state risk assessment results of the lithium battery, dynamically adjust the data sampling frequency under each charging scenario; set different anti-interference frequency modulation sequences, and dynamically adjust the data transmission frequency between the vehicle and the remote monitoring terminal according to the current charging scenario and network environment; Step S400: After receiving the transmitted data, the monitoring terminal combines the associated mapping library and the state risk assessment results to predict the health status trend of the lithium battery in the current charging scenario, comprehensively assess the health status of the lithium battery, and issue a health status warning for the lithium battery.

[0014] Furthermore, step S100 includes: Step S101: After the electric vehicle connects to the charging facility, the charging facility and the vehicle's battery management system establish a connection through a communication protocol, collect the current charging power, record the current charging mode P, and convert it into a binary vector using one-hot encoding; simultaneously, the grid voltage V is monitored in real time by the voltage sensor built into the charging facility. grid And the fluctuation range ΔV, stored as the interval [V grid -ΔV, V grid +ΔV]; Combined with the onboard environmental sensor to obtain the ambient temperature T env Furthermore, the load rate Y is obtained by calculating the ratio of the current charging power to the rated power of the charging facility; the collected variables are normalized to map the data to the [0, 1] interval; all features are integrated into a charging scenario feature vector F=[P,V] grid ,T env [,Y]; Step S102: During the charging process, the on-board sensor network collects the state parameters of the lithium battery in real time, sets the initial sampling frequency, monitors the internal temperature T of the battery pack through the temperature sensor, and obtains the voltage V of the individual cells through the voltage sensor. i and the total voltage V of the battery pack total V i The voltage of each low-i individual cell in the battery pack is represented; the current sensor measures the charging current I through the Hall effect; simultaneously, the battery management system calculates the state of charge (SOC) of the battery based on the coulomb counting method and calculates the internal resistance R using the AC impedance spectroscopy method; all state parameters are synchronized and aligned using timestamps, with an allowable time deviation of no more than 1ms; finally, all state parameters are integrated into the lithium battery state vector S. t =[V i V total ,I,SOC,R],S t This represents the complete state of the battery at time t; Step S103: For each charging process, the collected scene feature vector F and the corresponding state vector S are compared. t Pair them according to time sequence to form key-value pairs (F,S) t); Store all data pairs as a dataset, forming an association mapping library D: D={(F k ,S t,k )} N k=1 Where N represents the total number of data pairs, F k S represents the scene feature vector of the k-th charging session. t,k This represents the lithium battery state vector at time t during the k-th charge; after each charge, newly collected data pairs are automatically appended to the associated mapping library D.

[0015] Furthermore, step S200 includes: Step S201: Process the state vector sequence {S} in the associated database. t,k} Perform time-series difference calculation, and approximate the calculation by using the difference between adjacent sampling points: ΔS t =S t,k -S t-1,k ;where S t-1,k Let {ΔS} represent the state vector of the lithium battery at time t-1 during the k-th charge, and obtain the characteristic sequence of the rate of change of state. t,k The feature sequences are standardized, and the number of clusters M is determined using the silhouette coefficient method. The M value with the largest silhouette coefficient is selected, and the K-Means algorithm is used to cluster the state change features to divide the charging behavior patterns, resulting in M ​​cluster centers {C}. m} M m=1 C m C represents the center vector of the m-th cluster; each center C m The vector is a continuous numerical vector representing the feature distribution of a normal charging mode, and a unique label L is assigned to each cluster center. m L m Let represent the category label of the m-th cluster. The label is a discrete integer used to identify the cluster category to which the data point belongs; for any state change rate feature ΔS t Labels are obtained by clustering clusters using nearest neighbor matching: L m =argmin m ||ΔS t -C m ||;Integrate cluster centers and tags into a charging pattern library C={(C m L m )} M m=1 ; Step S202: Collect historical data over a period of time, and construct a training set by labeling outlier samples. Each sample includes: a scene feature vector F, and a lithium battery state change rate feature ΔS. t and clustering label L m ;Discrete label Lm Converted to one-hot encoding, it is represented as OneHot(L m The input vector X=[F,ΔS] is formed by concatenating it with numerical features. t OneHot(L m The XGBoost gradient boosting tree is used to construct a lithium battery state assessment model. The input is X, and the output y∈{0,1} represents whether the state is normal, 0 represents normal, and 1 represents abnormal. During training, the dataset is divided into training set, validation set and test set in a ratio of 7:1.5:1.5. The model classification performance is optimized by minimizing the cross-entropy loss function. The preset number of training rounds is set. Training is terminated when the validation set loss does not decrease for e consecutive rounds. The model performance is evaluated by accuracy and F1 score using the test set. Step S203: During the charging process, the current scene feature F and the state change rate ΔS are compared. t Input the trained model, calculate the Euclidean distance to the cluster centers, match the nearest cluster label, and output the anomaly probability G at time t. t Selecting the anomaly probability threshold θ based on the ROC curve G When the anomaly probability G t Exceeding the preset threshold θ G When the time comes, an alert will be triggered immediately, and a report containing a timestamp, scene parameters, anomaly probability, and anomaly type will be automatically generated.

[0016] Furthermore, step S300 includes: Step S301: Based on the anomaly probability output by the model, the state risk of the lithium battery is divided into three levels: when G t <=θ G1 When θ is set to low risk, it indicates a normal state, and the initial sampling frequency is maintained; when θ G1 <G t <=θ G2 When the risk is set to medium, indicating a potential anomaly in the state, the sampling frequency is increased to f1 = k1 × f0, where k1 > 1; when G t >θ G2 When the risk is set to high, indicating an abnormal state, the sampling frequency is increased to f2 = k2 × f0, where k2 > k1; θ G1 θ G2 The preset risk threshold is defined, k1 and k2 are frequency adjustment coefficients, and f0 represents the initial sampling frequency; during the charging process, the abnormal probability G is monitored. t The sampling frequency of the vehicle-mounted sensors is dynamically adjusted according to the risk level: when G t When the risk level rises from low to medium, the sampling frequency is switched from f0 to f1; when G t When the risk level rises from medium to high, the sampling frequency is switched from f1 to f2; when Gt When the risk level drops back to low, gradually restore the initial sampling frequency f0; Step S302: According to the communication protocol, the vehicle terminal and the remote monitoring terminal set the minimum communication frequency f. min and maximum communication frequency f max Based on the electromagnetic ambient noise intensity, a pseudo-random frequency modulation sequence is generated using a linear congruent generator: Q={q1,q2,...,q} a}, where q a This represents the a-th communication frequency, where a represents the sequence length, and a = T. charge / T switch T charge Indicates the estimated charging time, T switch Indicates the frequency switching period, q a ∈[f min f max The data transmission quality index Z is calculated based on packet loss and delay: Z = α × (1 - PLR) + β × (1 / U); where PLR ​​represents the packet loss rate, which is the proportion of lost data packets counted through the ACK confirmation mechanism, and U represents the delay parameter, which records the time difference between sending and receiving data packets; α and β represent the adjustment weights of the packet loss rate and the delay parameter, respectively; the frequency that maximizes Z is selected from the sequence Q as the data transmission frequency for the current charging scenario, and the frequencies are sorted according to the Z value, with the top n selected as the candidate frequency group; based on the selected transmission frequency, the data packets for the current charging scenario are constructed using vector F, lithium battery state vector St, anomaly probability, and clustering labels and sent to the remote monitoring terminal.

[0017] Furthermore, step S400 includes: Step S401: The monitoring terminal receives real-time data packets transmitted by the vehicle terminal, performs timestamp alignment, missing value imputation, and normalization on the received data; an LSTM network is used to construct a lithium battery health status prediction model; the real-time scene features F and the state vector S are then compared. t The model is fused to form a comprehensive feature vector as input. The hidden layer consists of two LSTM units and one fully connected layer, and the output layer has two neurons, each corresponding to the output of the lithium battery's capacity decay rate ΔC. t and internal resistance growth rate ΔR t Historical charging data is extracted from the association mapping library, and samples are divided according to time windows to form input sequences and label sequences. The Adam optimizer and mean squared error loss function are used to update the model parameters through backpropagation. Training is terminated when the validation set loss does not decrease for e consecutive rounds. The real-time feature vector is input into the trained model to obtain the predicted values ​​of capacity decay rate and internal resistance growth rate at the current moment. Based on the current predicted values ​​and charging scenario parameters, the rolling prediction method is used to predict the trend of lithium battery health status within the future Tp time period and generate a trend curve. Step S402: Combining the prediction results and real-time status data, calculate the health status assessment index of the lithium battery: using the exponential decay model, estimate the remaining lifespan of the lithium battery based on the predicted capacity decay rate: RUL t =In(C / C t ) / In(1-ΔC t ); where RUL t C represents the remaining lifespan of the lithium battery at time t, and C represents the capacity failure threshold. t This represents the current capacity of the lithium battery at time t; based on the internal resistance growth rate, the health index of the lithium battery is calculated: SOH. t =1-[(Σ Tp i=1 ΔR t+i ) / R], where SOH t Represents the lithium battery health index at time t, R represents the initial internal resistance of the lithium battery, and ΔR t+i This represents the predicted growth rate of the internal resistance at the i-th time step in the future; Step S403: Compare the obtained result with a preset threshold. If the calculated result is less than the preset threshold, the lithium battery is determined to be in an unhealthy state; compare the health assessment result with the anomaly probability G. t Integration, calculation of the comprehensive risk index: R t =w1×[(θ RUL -RUL t ) / θ RUL ]+w2×[(θ SOH -SOH t ) / θ SOH ]+w3×G t , where R t θ represents the overall risk index of the lithium battery at time t. RUL The preset threshold representing the remaining service life, θ SOH This represents the preset threshold for the health index; w1, w2, and w3 represent RUL respectively. t SOH t G t The weights in the comprehensive risk index; and w1+w2+w3=1; set the early warning threshold θ. R When the comprehensive risk index R t When the preset threshold is exceeded, an early warning is triggered, and early warning information is generated, including a timestamp, scene parameter F, health status indicators, predicted trend, abnormal probability, and early warning prompt, which is then sent back to the monitoring terminal of relevant personnel; relevant information of this charging process is appended to the associated mapping library D for update and storage.

[0018] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A remote monitoring method for estimating the state of lithium batteries in electric vehicles, characterized in that: The method includes: Step S100: Obtain relevant parameters of the current charging scenario through the interaction information between the charging facility and the vehicle terminal, and construct a charging scenario feature set; at the same time, based on various vehicle sensors, collect multi-dimensional state data of electric vehicle lithium batteries in real time under the charging mode, and construct a correlation mapping library between the charging scenario and the state changes of lithium batteries. Step S200: Based on the constructed association mapping library, analyze the changes in the charging state of the lithium battery under different charging scenarios, construct a lithium battery state assessment model, and assess whether there is a risk of abnormal state of the lithium battery under the current charging scenario. Step S300: Based on the state risk assessment results of the lithium battery, dynamically adjust the data sampling frequency under each charging scenario; set different anti-interference frequency modulation sequences, and dynamically adjust the data transmission frequency between the vehicle and the remote monitoring terminal according to the current charging scenario and network environment; Step S400: After receiving the transmitted data, the monitoring terminal combines the associated mapping library and the state risk assessment results to predict the health status trend of the lithium battery in the current charging scenario, comprehensively assess the health status of the lithium battery, and issue a health status warning for the lithium battery.

2. The remote monitoring method for estimating the state of lithium batteries in electric vehicles according to claim 1, characterized in that: Step S100 includes: Step S101: After the electric vehicle connects to the charging facility, the charging facility and the vehicle's battery management system establish a connection through a communication protocol, collect the current charging power, record the current charging mode P, and convert it into a binary vector using one-hot encoding; simultaneously, the grid voltage V is monitored in real time by the voltage sensor built into the charging facility. grid And the fluctuation range ΔV, stored as the interval [V grid -ΔV, V grid +ΔV]; Combined with the onboard environmental sensor to obtain the ambient temperature T env Furthermore, the load rate Y is obtained by calculating the ratio of the current charging power to the rated power of the charging facility; the collected variables are normalized to map the data to the [0, 1] interval; all features are integrated into a charging scenario feature vector F=[P,V] grid ,T env [,Y]; Step S102: During the charging process, the on-board sensor network collects the state parameters of the lithium battery in real time, sets the initial sampling frequency, monitors the internal temperature T of the battery pack through the temperature sensor, and obtains the voltage V of the individual cells through the voltage sensor. i and the total voltage V of the battery pack total V i This represents the voltage of each low-i individual cell in the battery pack; the current sensor measures the charging current I through the Hall effect; simultaneously, the battery management system calculates the battery's state of charge (SOC) based on the coulomb counting method and calculates the internal resistance R using the AC impedance spectroscopy method; all state parameters are synchronized and aligned using timestamps, and all state parameters are integrated into the lithium battery state vector S. t =[V i V total ,I,SOC,R],S t This represents the complete state of the battery at time t; Step S103: For each charging process, the collected scene feature vector F and the corresponding state vector S are compared. t Pair them according to time sequence to form key-value pairs (F,S) t ); Store all data pairs as a dataset, forming an association mapping library D: D={(F k ,S t,k )} N k=1 Where N represents the total number of data pairs, F k S represents the scene feature vector of the k-th charging session. t,k This represents the lithium battery state vector at time t during the k-th charge; after each charge, newly collected data pairs are automatically appended to the associated mapping library D.

3. The remote monitoring method for estimating the state of lithium batteries in electric vehicles according to claim 1, characterized in that: Step S200 includes: Step S201: Process the state vector sequence {S} in the associated database. t,k } Perform time-series difference calculation, and approximate the calculation by using the difference between adjacent sampling points: ΔS t =S t,k -S t-1,k ;where S t-1,k Let {ΔS} represent the state vector of the lithium battery at time t-1 during the k-th charge, and obtain the characteristic sequence of the rate of change of state. t,k The feature sequences are standardized, and the number of clusters M is determined using the silhouette coefficient method. The M value with the largest silhouette coefficient is selected, and the K-Means algorithm is used to cluster the state change features to divide the charging behavior patterns, resulting in M ​​cluster centers {C}. m } M m=1 C m C represents the center vector of the m-th cluster; each center C m The vector is a continuous numerical vector representing the feature distribution of a normal charging mode, and a unique label L is assigned to each cluster center. m L m Let represent the category label of the m-th cluster. The label is a discrete integer used to identify the cluster category to which the data point belongs; for any state change rate feature ΔS t Labels are obtained by clustering clusters using nearest neighbor matching; the cluster centers and labels are integrated into a charging pattern library C={(C m L m )} M m=1 ; Step S202: Collect historical data over a period of time, and construct a training set by labeling outlier samples. Each sample includes: a scene feature vector F, and a lithium battery state change rate feature ΔS. t and clustering label L m ;Discrete label L m The data is converted into one-hot encoding and concatenated with numerical features to form the input vector X. An XGBoost gradient boosting tree is used to construct a lithium battery state assessment model. The input is X, and the output y∈{0,1} represents whether the state is normal (0 for normal, 1 for abnormal). During training, the dataset is divided into training, validation, and test sets according to a set ratio. The model classification performance is optimized by minimizing the cross-entropy loss function. A preset number of training rounds is set, and training is terminated when the validation set loss does not decrease for e consecutive rounds. The model performance is evaluated using the test set based on accuracy and F1 score. Step S203: During the charging process, the current scene feature F and the state change rate ΔS are compared. t Input the trained model, calculate the Euclidean distance to the cluster centers, match the nearest cluster label, and output the anomaly probability G at time t. t When the anomaly probability G t Exceeding the preset threshold θ G When the time comes, an alert will be triggered immediately, and a report containing a timestamp, scene parameters, anomaly probability, and anomaly type will be automatically generated.

4. The remote monitoring method for estimating the state of lithium batteries in electric vehicles according to claim 1, characterized in that: Step S300 further includes: Step S301: Based on the anomaly probability output by the model, the state risk of the lithium battery is divided into three levels: when G t <=θ G1 When θ is set to low risk, it indicates a normal state, and the initial sampling frequency is maintained; when θ G1 <G t <=θ G2 When the risk is set to medium, indicating a potential anomaly in the state, the sampling frequency is increased to f1 = k1 × f0, where k1 > 1; when G t >θ G2 When the risk is set to high, indicating an abnormal state, the sampling frequency is increased to f2 = k2 × f0, where k2 > k1; θ G1 θ G2 The preset risk threshold is defined, k1 and k2 are frequency adjustment coefficients, and f0 represents the initial sampling frequency; during the charging process, the abnormal probability G is monitored. t The sampling frequency of the vehicle-mounted sensors is dynamically adjusted according to the risk level: when G t When the risk level rises from low to medium, the sampling frequency is switched from f0 to f1; when G t When the risk level rises from medium to high, the sampling frequency is switched from f1 to f2; when G t When the risk level drops back to low, gradually restore the initial sampling frequency f0; Step S302: According to the communication protocol, the vehicle terminal and the remote monitoring terminal set the minimum communication frequency f. min and maximum communication frequency f max Based on the electromagnetic ambient noise intensity, a pseudo-random frequency modulation sequence is generated using a linear congruent generator: Q={q1,q2,...,q} a }, where q a This represents the a-th communication frequency, where a represents the sequence length, and q... a ∈[f min f max The data transmission quality index Z is calculated based on packet loss and delay: Z = α × (1 - PLR) + β × (1 / U); where PLR ​​represents the packet loss rate, which is the proportion of lost data packets counted through the ACK confirmation mechanism, and U represents the delay parameter, which records the time difference between sending and receiving data packets; α and β represent the adjustment weights of the packet loss rate and the delay parameter, respectively; the frequency that maximizes Z is selected from the sequence Q as the data transmission frequency for the current charging scenario, and the frequencies are sorted according to the Z value, with the top n selected as the candidate frequency group; based on the selected transmission frequency, the data packets for the current charging scenario are constructed using vector F, lithium battery state vector St, anomaly probability, and clustering labels and sent to the remote monitoring terminal.

5. The remote monitoring method for estimating the state of lithium batteries in electric vehicles according to claim 1, characterized in that: Step S400 includes: Step S401: The monitoring terminal receives real-time data packets transmitted by the vehicle terminal, performs timestamp alignment, missing value imputation, and normalization on the received data; an LSTM network is used to construct a lithium battery health status prediction model; the real-time scene features F and the state vector S are then compared. t The model is fused to form a comprehensive feature vector as input. The hidden layer consists of two LSTM units and one fully connected layer, and the output layer has two neurons, each corresponding to the output of the lithium battery's capacity decay rate ΔC. t and internal resistance growth rate ΔR t Historical charging data is extracted from the association mapping library, and samples are divided according to time windows to form input sequences and label sequences. The Adam optimizer and mean squared error loss function are used to update the model parameters through backpropagation. Training is terminated when the validation set loss does not decrease for e consecutive rounds. The real-time feature vector is input into the trained model to obtain the predicted values ​​of capacity decay rate and internal resistance growth rate at the current moment. Based on the current predicted values ​​and charging scenario parameters, the rolling prediction method is used to predict the trend of lithium battery health status within the future Tp time period and generate a trend curve. Step S402: Combining the prediction results and real-time status data, calculate the health status assessment index of the lithium battery: Use an exponential decay model to estimate the remaining lifespan of the lithium battery based on the predicted capacity decay rate; calculate the health index of the lithium battery based on the internal resistance growth rate; compare the obtained results with a preset threshold; if the calculated result is less than the preset threshold, the lithium battery is determined to be in an unhealthy state; compare the health assessment results with the anomaly probability G. t The system integrates and calculates a comprehensive risk index; sets an early warning threshold; when the comprehensive risk index exceeds the preset threshold, an early warning is triggered, generating early warning information, including a timestamp, scenario parameter F, health status indicators, predicted trends, abnormal probability, and early warning prompts, which is then sent back to the monitoring terminals of relevant personnel; and appends relevant information from this charging process to the associated mapping library D for updating and storage.

6. A remote monitoring system for estimating the state of lithium batteries in electric vehicles, characterized in that: The system includes an association mapping module, a lithium battery status assessment module, a dynamic sampling and data transmission module, and a health status prediction module. The association mapping module obtains relevant parameters of the current charging scenario through the interaction information between the charging facility and the vehicle terminal, and constructs a charging scenario feature set; at the same time, based on various vehicle sensors, it collects multi-dimensional state data of electric vehicle lithium batteries in real time under the charging mode, and constructs an association mapping library between the charging scenario and the state changes of lithium batteries. The lithium battery state assessment module analyzes the changes in the charging state of the lithium battery under different charging scenarios based on the constructed association mapping library, constructs a lithium battery state assessment model, and assesses whether there is a risk of abnormal state of the lithium battery under the current charging scenario. The dynamic sampling and data transmission module dynamically adjusts the data sampling frequency under various charging scenarios based on the state risk assessment results of the lithium battery; it sets different anti-interference frequency modulation sequences and dynamically adjusts the data transmission frequency between the vehicle and the remote monitoring terminal according to the current charging scenario and network environment. After receiving the transmitted data at the monitoring terminal, the health status prediction module combines the associated mapping library and the status risk assessment results to predict the health status trend of the lithium battery in the current charging scenario, comprehensively assess the health status of the lithium battery, and issue a health status warning for the lithium battery.

7. The remote monitoring system for estimating the state of lithium batteries in electric vehicles according to claim 6, characterized in that: The association mapping module includes a charging scenario parameter acquisition unit, a lithium battery status data acquisition unit, and an association mapping construction unit. The charging scenario parameter acquisition unit acquires charging scenario parameters and performs preprocessing through the interaction between the charging facility and the vehicle terminal to construct a scenario feature vector. The lithium battery status data acquisition unit utilizes the vehicle-mounted sensor network to collect multi-dimensional status data of the lithium battery in real time, and performs synchronous alignment and preprocessing to construct a lithium battery status vector. The association mapping construction unit pairs the collected scene feature vectors with the lithium battery state vectors in a time sequence to form key-value pairs and stores them as an association mapping library.

8. The remote monitoring system for estimating the state of lithium batteries in electric vehicles according to claim 6, characterized in that: The lithium battery status assessment module includes a clustering analysis unit, a status assessment unit, and an anomaly warning unit. The clustering analysis unit performs time-series difference calculation on the state vector sequence in the association mapping library and identifies charging behavior patterns through clustering analysis; The state assessment unit constructs a training set using labeled abnormal samples to train the lithium battery state assessment model. The anomaly warning unit inputs the current scene characteristics and the lithium battery state change rate into the model, calculates the anomaly probability, and triggers a warning based on the threshold.

9. The remote monitoring system for estimating the state of lithium batteries in electric vehicles according to claim 6, characterized in that: The dynamic sampling and data transmission module includes a sampling frequency adjustment unit and a data transmission frequency adjustment unit; The sampling frequency adjustment unit dynamically adjusts the data sampling frequency of the vehicle-mounted sensor according to the anomaly probability output by the model. The data transmission frequency adjustment unit constructs an anti-interference frequency modulation sequence based on the current charging scenario and network environment, and dynamically adjusts the data transmission frequency between the vehicle and the remote monitoring terminal.

10. The remote monitoring system for estimating the state of lithium batteries in electric vehicles according to claim 6, characterized in that: The health status prediction module includes a health status prediction model construction unit and a health status assessment and early warning unit. The health status prediction model building unit preprocesses the received data and integrates real-time scene features and lithium battery state vectors as model inputs, and builds a lithium battery health status prediction model through an LSTM network. The health status assessment and early warning unit outputs the health status trend of the lithium battery through a prediction model and assesses the health status of the lithium battery in combination with real-time status data. Once an abnormal risk is detected, an early warning is immediately triggered and detailed warning information is generated.

Images