A Battery Aging Assessment Method and System for Battery Swapping Cabinets Based on Charging Pattern Clustering

Abstract

This invention discloses a method and system for battery aging assessment of battery swapping cabinets based on charging pattern clustering, belonging to the field of battery testing technology for battery swapping cabinets. The method includes: analyzing battery charging cycles according to the charging cycle of the swapping cabinet to construct multi-dimensional charging features; performing clustering learning based on these features to construct multiple charging pattern clusters; performing health analysis on the battery based on these clusters to obtain battery health status parameters; retrieving real-time charging data and matching the charging pattern clusters with the health status parameters to determine the target charging pattern cluster; and performing battery aging assessment based on the target charging pattern cluster to generate aging assessment results. This invention solves the technical problems of low assessment accuracy and poor real-time performance caused by the diversity of charging patterns in existing battery aging assessment methods. It achieves refined analysis and dynamic matching of battery health status based on charging pattern clustering, improving the accuracy, scenario adaptability, and real-time responsiveness of battery aging assessment results.

Description

Technical Field

[0001] This invention relates to the field of battery testing technology for battery swapping cabinets, and specifically to a battery aging assessment method and system for battery swapping cabinets based on charging mode clustering. Background Technology

[0002] With the increasing popularity of electric vehicles, battery swapping stations, as crucial facilities for battery exchange and charging, are becoming an essential component of electric vehicle charging infrastructure. Battery swapping stations offer a more efficient way to replace batteries for electric vehicles, shortening charging time and improving charging efficiency compared to traditional charging stations. However, the health condition of the batteries in the swapping station directly impacts the performance of the electric vehicle and the user experience. Battery aging is an unavoidable issue during battery use; excessive aging not only affects the battery's range but may also pose safety hazards.

[0003] Existing battery aging assessment methods typically rely on simple capacity degradation measurements, lacking comprehensive analysis of batteries under different charging modes and failing to adequately consider the complex operating environments and charging conditions batteries face in actual use. This approach may not accurately reflect the true aging state of the battery, leading to insufficient accuracy and reliability of the assessment results. Therefore, how to conduct comprehensive and accurate aging assessments based on battery charging modes has become a significant technical challenge in battery management and maintenance. Summary of the Invention

[0004] This application provides a battery aging assessment method and system for battery swapping cabinets based on charging mode clustering, which solves the technical problems of low assessment accuracy and poor real-time performance caused by the diversity of charging modes in existing battery aging assessment methods.

[0005] The first aspect of this application provides a method for assessing battery aging in battery swapping cabinets based on charging pattern clustering, the method comprising:

[0006] The battery swapping cabinet's batteries are analyzed for charging cycles according to their charging periods to construct multi-dimensional charging features. Clustering learning is then performed based on these features to construct multiple charging mode clusters. Battery health analysis is then conducted on the batteries based on these multiple charging mode clusters to obtain battery health status parameters. Real-time charging data of the batteries is retrieved and combined with the battery health status parameters to match and determine the target charging mode cluster. Finally, battery aging assessment is performed on the batteries according to the target charging mode cluster to generate battery aging assessment results.

[0007] A second aspect of this application provides a battery aging assessment system for battery swapping cabinets based on charging pattern clustering, the system comprising:

[0008] The battery swapping station comprises the following modules: a charging cycle analysis module, a battery health analysis module, and a battery health assessment module. The charging cycle analysis module performs charging cycle analysis on the battery swapping station's batteries according to the charging cycle of the swapping station, constructing multi-dimensional charging characteristics. The battery health analysis module performs clustering learning based on the multi-dimensional charging characteristics, constructing multiple charging mode clusters, and performs battery health analysis on the battery swapping station's batteries based on these multiple charging mode clusters, obtaining battery health status parameters. The matching and determination module retrieves real-time charging data of the battery swapping station's batteries and combines it with the battery health status parameters to perform matching and determination on the multiple charging mode clusters, identifying the target charging mode cluster. The aging assessment module performs battery aging assessment on the battery swapping station's batteries according to the target charging mode cluster, generating battery aging assessment results.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] First, by analyzing the charging cycles of the batteries in the battery swapping station, multi-dimensional charging characteristics are extracted to assess the battery's health status. Then, through detailed analysis of the battery charging process, charging mode clusters are constructed, and battery health analysis is performed based on these clusters to obtain battery health status parameters. Next, battery charging data is monitored in real time, and the health status parameters are matched with multiple charging mode clusters to determine the most suitable charging mode cluster for the current battery. Finally, an aging assessment is performed on the battery based on the target charging mode cluster, generating detailed aging assessment results to help predict battery lifespan. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the battery aging assessment method for battery swapping cabinets based on charging mode clustering provided in an embodiment of this application.

[0013] Figure 2 This is a schematic diagram of the battery aging assessment system for battery swapping cabinets based on charging mode clustering, provided in an embodiment of this application.

[0014] Explanation of reference numerals in the attached diagram: 11 Charging cycle analysis module, 12 Battery health analysis module, 13 Matching determination module, 14 Aging assessment module. Detailed Implementation

[0015] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0016] Example 1, as Figure 1 As shown, this application provides a battery aging assessment method for battery swapping cabinets based on charging pattern clustering. The method includes:

[0017] Based on the charging cycle of the battery swapping cabinet, a charging cycle analysis of the battery swapping cabinet is performed to construct multi-dimensional charging characteristics.

[0018] In this embodiment, the charging cycle of the battery swapping cabinet refers to the complete process of the battery from the start to the end of charging. The system first collects key parameters such as voltage, current, and temperature of the battery in real time during the charging process, recording the changes of these parameters over time, according to the charging cycle of the battery swapping cabinet. Then, outlier removal and missing value imputation are performed on these collected key parameters to obtain a usable sequence of key parameters. Next, charging cycle analysis is performed on these processed key parameters to construct a complete charging cycle state parameter set. Based on these charging cycle state parameters, multi-stage analysis is conducted to form multi-dimensional charging characteristics. These multi-dimensional charging characteristics can comprehensively reflect the state and performance of the battery in each charging cycle, providing accurate data support for battery health analysis and aiding in subsequent aging assessment and lifespan prediction.

[0019] Furthermore, charging cycle analysis of the batteries in the battery swapping cabinet is performed according to the charging cycle of the cabinet to construct multi-dimensional charging characteristics. The methods include:

[0020] A time-series analysis is performed on the batteries in the battery swapping cabinet according to the charging cycle to extract charging time-series data, which includes an initial voltage sequence, an initial current sequence, and an initial temperature sequence. Interference analysis is then performed on the initial voltage sequence, the initial current sequence, and the initial temperature sequence to identify abnormal abrupt changes. Based on these abnormal abrupt changes, data is removed from the initial voltage sequence, the initial current sequence, and the initial temperature sequence to generate a data missing sequence. A linear interpolation algorithm is then used to complete the data missing sequence, generating a voltage sequence, a current sequence, and a temperature sequence. The voltage sequence, the current sequence, and the temperature sequence are then subjected to charging cycle analysis to construct charging cycle state parameters. Based on these charging cycle state parameters, multi-stage analysis is performed to construct the multi-dimensional charging characteristics.

[0021] Preferably, during the battery charging process in the battery swapping cabinet, a time-series analysis is first performed on the battery according to the charging cycle. This involves recording the voltage value at each time point from the beginning to the end of charging, forming an initial voltage sequence; recording the current value at each time point during charging, forming an initial current sequence; and recording the temperature value at each time point during charging, forming an initial temperature sequence. These initial sequences are then integrated into a single set to form charging time-series data, providing the foundational time-series data for subsequent charging cycle analysis. Subsequently, because external factors or abnormal equipment fluctuations may occur during charging, causing sudden changes in the data at certain times that fail to reflect the normal charging state, to improve the accuracy and reliability of the data, interference analysis is performed on the initial voltage, initial current, and initial temperature sequences in the charging time-series data based on preset normal ranges for voltage, current, and temperature. Data points outside the normal range are identified and defined as abnormal mutation points, representing external interference or sudden problems during battery charging. Subsequently, for detected anomalous abrupt changes, the corresponding time period data is removed from the initial voltage, current, and temperature sequences to generate a data missing sequence. Then, a linear interpolation algorithm is used to fill in the missing data based on the values ​​of adjacent known data points, generating complete voltage, current, and temperature sequences. These filled-in voltage, current, and temperature sequences are then used for further charging cycle analysis. Specifically, according to the charging cycle, the same time-series analysis, interference analysis, data removal, and data filling processes are performed on each complete charging cycle from full charge to full discharge, resulting in voltage, current, and temperature sequences for multiple charging cycles. These sequences are then integrated and stored to construct charging cycle state parameters. Finally, based on the charging cycle state parameters, the battery charging process is divided into multiple stages, such as constant current charging, constant voltage charging, and constant temperature charging. For each stage, a corresponding time-domain analysis is performed to extract the time-domain features of each stage, thereby constructing multi-dimensional charging features. These charging features not only reflect the health status of the battery, but also reveal the battery's performance under different charging modes, providing rich data support and accurate analysis results for battery aging assessment and health analysis.

[0022] Furthermore, based on the charging cycle state parameters, a multi-stage analysis is performed to construct the multi-dimensional charging characteristics. The method includes:

[0023] Based on the charging cycle state parameters and the voltage sequence, a stage is defined to determine the constant voltage charging stage; time-domain analysis is performed on the battery in the battery swapping cabinet according to the constant voltage charging stage to obtain the first time-domain feature; based on the charging cycle state parameters and the current sequence, a stage is defined to determine the constant current charging stage; time-domain analysis is performed on the battery in the battery swapping cabinet according to the constant current charging stage to obtain the second time-domain feature; based on the charging cycle state parameters and the temperature sequence, a stage is defined to determine the constant temperature charging stage; time-domain analysis is performed on the battery in the battery swapping cabinet according to the constant temperature charging stage to obtain the third time-domain feature; the first time-domain feature, the second time-domain feature, and the third time-domain feature are correlated and integrated to construct the multi-dimensional charging feature.

[0024] Optionally, after obtaining the charging cycle state parameters, the charging cycle state parameters are combined with the corresponding voltage sequence, current sequence, and temperature sequence to divide the process into stages. During this process, when the battery voltage sequence reaches a set constant value, it indicates the battery has entered a constant-voltage charging state; that is, the current gradually decreases, and the battery continues to charge at a constant voltage. At this time, the data corresponding to the constant-voltage charging state in the voltage sequence is extracted from the charging cycle state parameters based on the timestamp, thus determining the constant-voltage charging stage. When the battery current sequence reaches a set constant value, it indicates the battery has entered a constant-current charging state. The same method is used to extract the data corresponding to the time from the charging cycle state parameters to determine the constant-current charging stage. When the battery temperature sequence reaches a set constant value, it indicates the battery has entered a constant-temperature charging state. The same method is also used to extract the data corresponding to the time from the charging cycle state parameters to determine the constant-temperature charging stage. After identifying the different charging stages, for the constant-voltage charging stage, the battery maintains a constant voltage while the current gradually decreases. Therefore, time-domain analysis of the constant-voltage charging stage reveals the changes in current and temperature over time, allowing for the calculation of data such as the current decay rate, charging time, constant-voltage charging rate, and temperature change rate, constituting the first time-domain feature. For the constant-current charging stage, the battery maintains a constant current while the voltage gradually increases over time. Therefore, time-domain analysis of the constant-current charging stage reveals the changes in voltage and temperature over time, allowing for the calculation of data such as the voltage increase rate, charging time, constant-current charging rate, and temperature change rate, constituting the second time-domain feature. For the constant-temperature charging stage, the temperature remains stable and fluctuates within a certain range. In this case, analysis of the temperature sequence calculates data such as the temperature fluctuation amplitude and temperature change rate, constituting the third time-domain feature. Finally, the first time-domain features, the second time-domain features, and the third time-domain features are correlated and integrated to form multi-dimensional charging features. These multi-dimensional charging features can not only reveal the health status of the battery, but also help to evaluate the aging status and charging performance of the battery through comprehensive analysis at different stages.

[0025] Clustering learning is performed based on the multidimensional charging features to construct multiple charging mode clusters. Battery health analysis is then performed on the batteries in the battery swapping cabinet based on these multiple charging mode clusters to obtain battery health status parameters.

[0026] In one embodiment, multidimensional charging features extracted during battery charging are used as input data for clustering learning, thereby dividing the charging features into multiple charging mode clusters. Each cluster represents a specific charging mode of a battery during charging. The similarity in charging features within each cluster indicates that the batteries are operating under similar charging environments. Subsequently, each charging mode cluster is input into a battery health assessment model for analysis. This model is based on a deep neural network and trained using historical charging features and historical state parameters. The training steps typically include forward propagation, loss calculation (e.g., mean squared error), backpropagation, and parameter optimization (e.g., Adam optimizer). After receiving the charging mode clusters, the battery health assessment model performs health analysis on the charging features within the clusters based on the learned knowledge, generating battery health state parameters for each charging mode. These parameters typically include remaining capacity, charging efficiency, health state index, battery internal resistance, maximum charging voltage, and depth of discharge, reflecting the actual health status of the battery and providing data support for subsequent battery aging assessments.

[0027] Furthermore, based on the aforementioned multidimensional charging features, clustering learning is performed to construct multiple charging pattern clusters. The method includes:

[0028] The multidimensional charging features are dimensionality reduced to obtain low-dimensional feature vectors; clustering learning is performed based on the low-dimensional feature vectors to determine K clusters, where K is a positive integer greater than 0; clustering is performed based on the K clusters to construct multiple initial data clusters; the multiple initial data clusters are traversed for optimization and verification to construct the multiple charging mode clusters.

[0029] Preferably, for the obtained multidimensional charging features, common dimensionality reduction techniques such as principal component analysis (PCA) or t-SNE are used to convert the multidimensional charging features into low-dimensional feature vectors. Taking PCA as an example, the low-dimensional feature vectors are obtained by eigenvalue decomposition of the covariance matrix, typically the top k principal component components with the largest variance contribution. Dimensionality reduction reduces computational complexity and improves the efficiency and accuracy of subsequent analysis. Then, based on these low-dimensional feature vectors, the silhouette coefficient method is used to determine the number of clusters K to be specified during clustering. This K value is a positive integer greater than 0. Afterwards, according to the determined number of clusters K, K cluster centers are randomly initialized from the low-dimensional feature vectors, and the Euclidean distance is used to calculate the distance between each low-dimensional feature vector and the K cluster centers. Each low-dimensional feature vector is then assigned to its nearest cluster. At this point, the battery charging features within each cluster are similar, representing a state of charging mode. Next, the mean of all low-dimensional eigenvectors within a cluster is calculated and used as the new cluster center. Low-dimensional eigenvectors are then reassigned to the nearest cluster. This process is repeated until the cluster centers no longer change significantly or the maximum set number of iterations is reached, forming multiple initial data clusters. Then, because these initial data clusters may contain noise or unreasonable groupings, they are optimized and validated through data center analysis and intra-cluster consistency checks to gradually improve the clustering quality and ensure that each cluster represents a charging mode. The optimized results constitute multiple charging mode clusters, which clearly describe the impact of different charging modes on battery health and provide a valid basis for subsequent battery health analysis and aging assessment.

[0030] Furthermore, clustering learning is performed based on the low-dimensional feature vectors to determine the number of K clusters. The method includes:

[0031] Based on the low-dimensional feature vector, fuzzy boundary division is performed, and a candidate threshold for the number of clusters is set; the low-dimensional feature vector is contour calculated according to the candidate threshold for the number of clusters to generate multiple contour coefficients, and the contour coefficients correspond to the number of clusters; the multiple contour coefficients are arranged in descending order, and the first contour coefficient is extracted as the K number of clusters.

[0032] Optionally, before cluster analysis, to determine the optimal K value, a Gaussian membership function is first used to calculate the membership degree between each low-dimensional feature vector and all candidate cluster centers, constructing a membership degree matrix to reflect the relationship between the low-dimensional feature vector and the cluster. Then, based on the membership degree matrix and combined with domain knowledge or the natural distribution of data, a candidate threshold for the number of clusters is set to minimize the overlapping areas of different clusters while ensuring close clustering of data points within each cluster. Generally, if a candidate K value leads to excessive overlap between clusters or a relatively uniform distribution of membership degrees, this K value may be unsuitable; conversely, when membership degrees are concentrated and data points within clusters are close together, this K value is more appropriate. Afterward, for each low-dimensional feature vector, its corresponding silhouette coefficient is calculated according to the determined candidate threshold for the number of clusters. These silhouette coefficients range from -1 to 1. A value closer to 1 indicates better clustering quality, closer clustering, and greater separation between clusters; a value close to 0 indicates that the data point may be located on the cluster boundary, while a value close to -1 indicates that the data point may have been incorrectly assigned to the wrong cluster. After obtaining the silhouette coefficients of all low-dimensional feature vectors, for each possible K value, the average silhouette coefficient of all low-dimensional feature vectors is calculated to obtain the average silhouette coefficient of each cluster number K. By sorting the average silhouette coefficients under different K values ​​in descending order, the K value corresponding to the largest average silhouette coefficient is selected as the optimal number of clusters. This provides a reasonable and reliable foundation for subsequent cluster learning and data analysis, thereby ensuring the accuracy of cluster analysis and the quality of clustering results.

[0033] Furthermore, the method for constructing the multiple charging mode clusters by traversing the multiple initial data clusters for optimization and verification includes:

[0034] Data center analysis is performed based on the multiple initial data clusters to determine multiple data cluster centers; data points are identified by traversing the multiple initial data clusters, and multiple cluster sample points are extracted based on the identification results; according to the multiple initial data clusters, the multiple cluster sample points are matched with the multiple data cluster centers to generate multiple data point combinations, each of the multiple data point combinations having one and only one data cluster center and multiple cluster sample points; verification is performed based on the multiple data cluster centers and the multiple cluster sample points, and the multiple initial data clusters are optimized based on the verification results to construct the multiple charging mode clusters.

[0035] Optionally, after obtaining multiple initial data clusters, a data center analysis is performed on each initial data cluster to calculate the mean of each initial data cluster and determine multiple data cluster centers. Then, the multiple initial data clusters are traversed, and the distance between each traversed data point and the data cluster center is calculated. These distances are used to label these data points, and points closer to the cluster center are selected as cluster sample points based on these labels. Next, these cluster sample points are matched with the data cluster centers to generate multiple data point combinations. Each data point combination contains one cluster center and multiple cluster sample points for further validation and optimization. Then, the data clusters are validated and optimized based on the data cluster centers and cluster sample points in each data point combination. The validity of the initial clustering is checked by verifying whether the similarity between the cluster sample points and the cluster center in each data point combination is sufficiently high. Finally, based on the validation results, the cluster boundaries are readjusted, and data points that do not meet the requirements are reassigned to other clusters to avoid cluster overlap or incorrect sample point division, thus obtaining multiple final charging mode clusters, providing an accurate basis for subsequent battery aging assessment, lifespan prediction, and other analyses.

[0036] Furthermore, based on the verification of the multiple data cluster centers and the multiple cluster sample points, and the optimization of the multiple initial data clusters according to the verification results, the multiple charging mode clusters are constructed. The method includes:

[0037] Based on the multiple data point combinations, the average distance between the centers of the multiple data clusters and the multiple cluster sample points is calculated, generating multiple average distance values; the total average distance is calculated, and if the average distance value within any data point combination exceeds Q times the total average distance, the data point combination is identified as a loose cluster, where Q is a positive integer greater than 1; when the loose cluster exists, clustering optimization is performed on the multiple data point combinations, the multiple initial data clusters are updated, and the multiple charging mode clusters are constructed.

[0038] Optionally, for each data point combination, the Euclidean distance between the cluster center and all cluster sample points is calculated, and the average distance value is determined. These average distance values ​​are used to measure the cluster compactness; the smaller the value, the more concentrated the sample points within the cluster, and the higher the clustering quality. Subsequently, a total average distance is calculated for all data point combinations. This total average distance is the average of the average distances between all cluster centers and sample points, representing the overall compactness of all clusters. Then, a threshold Q is defined to identify loose clusters. When any average distance value exceeds Q times the total average distance, the cluster is considered a loose cluster and is identified. For the initial data clusters identified as loose clusters, some data points within the loose cluster are reassigned to clusters more similar to them, or the cluster boundaries are reassessed, and the cluster center position is updated by recalculating the mean of the data points within the cluster to ensure that the cluster center represents the central position of the data points within the cluster. After the above optimization and update steps, the resulting multiple charging mode clusters will be more closely spaced and have better differentiation. These charging mode clusters can more accurately reflect the health status of the battery under different charging modes, providing more reliable data support for subsequent battery aging assessments, performance analyses, etc.

[0039] The real-time charging data of the battery in the battery swapping cabinet is retrieved and combined with the battery health status parameters to match and determine the target charging mode cluster.

[0040] In one embodiment, the system monitors and acquires various real-time charging data of the battery in the battery swapping cabinet during the current charging process. This real-time charging data typically includes information such as battery voltage, current, temperature, remaining capacity, charging efficiency, health status index, and battery internal resistance, reflecting the battery's state during a specific charging process. Subsequently, the real-time charging data is matched with each battery health status parameter. The similarity between the real-time charging data and each battery health status parameter is calculated using Euclidean distance, and the closest set of battery health status parameters is selected. Then, the charging mode cluster corresponding to this set of battery health status parameters is obtained, and this charging mode cluster is used as the target charging mode cluster. This target charging mode cluster best reflects the current battery health status and charging characteristics during the charging process, ensuring the accuracy of subsequent battery aging assessments.

[0041] The battery aging assessment of the battery swapping cabinet is performed according to the target charging mode cluster, and the battery aging assessment results are generated.

[0042] In one embodiment, the battery in the battery swapping cabinet is first evaluated using charging characteristics and health status data from the target charging mode cluster over multiple charging cycles. The evaluation values ​​are then organized chronologically into a historical health status sequence, reflecting the battery's changing trends across different charging cycles. Subsequently, a regression algorithm is used to train the battery health status parameters of the target charging mode cluster, generating a baseline aging curve. This baseline aging curve represents the battery's aging trend under normal conditions. Next, the historical health status sequence is compared with the baseline aging curve, and a dynamic time warping algorithm is used to adjust the time axis, thereby accurately aligning the battery's health data with the baseline aging curve. Then, based on the comparison results, the remaining usable lifespan of the battery is predicted—that is, the time the battery can continue to be used effectively in its current state—generating a complete battery aging assessment result. This assessment result includes information such as the battery's health status, aging rate, and remaining usable lifespan, providing a basis for decision-making regarding subsequent battery replacement, maintenance, and management.

[0043] Furthermore, the battery aging assessment of the battery swapping cabinet is performed according to the target charging mode cluster to generate battery aging assessment results. The method includes:

[0044] Based on the target charging mode cluster, the battery in the battery swapping cabinet is evaluated for charging over multiple charging cycles to obtain health status estimates for multiple charging cycles. These health status estimates are then serialized according to the charging time sequence to construct a historical health status sequence. Data regression training is performed according to the target charging mode cluster to set a baseline aging curve. The historical health status sequence is compared with the baseline aging curve, and dynamic time warping is performed based on the comparison results to calculate battery aging rate data. Attenuation is evaluated based on the battery aging rate data to construct a battery aging trajectory. Based on the battery aging trajectory, the battery life of the battery in the battery swapping cabinet is predicted to obtain a predicted remaining useful life value, which is then added to the battery aging evaluation result.

[0045] Preferably, based on a defined cluster of target charging modes, the system retrieves the corresponding health status assessment model. This model is trained using historical data for that charging mode and can employ a deep neural network as its framework, with the training steps identical to those described above. After retrieving the corresponding health status assessment model, key parameters of the battery in the swapping cabinet across multiple charging cycles, such as voltage, current, and temperature, are input into the model to calculate a comprehensive health status estimate, reflecting the battery's performance over multiple charging cycles. Subsequently, the health status estimates for all charging cycles are arranged according to the charging time sequence, forming a historical health status sequence to reflect the battery's health changes at different time points and during charging cycles. After constructing the historical health status sequence, the system uses the historical data corresponding to the target charging mode cluster for data regression training. Taking an exponential regression model as an example, the time and health status values ​​from the historical data are input into this model. The least squares method is used to minimize the loss function, and the model parameters are updated to fit the data until the loss function reaches its minimum. After completing regression training, a baseline aging curve for the battery is plotted based on the parameters of the exponential regression model. This baseline aging curve is a theoretical standard curve representing the aging process of the battery under typical usage conditions. Next, a dynamic time warping algorithm is used to align the historical health state sequence with the generated baseline aging curve, determining the differences between them and calculating the battery's aging rate—the speed of battery degradation—to help predict the battery's future performance and aging process under the same charging mode. Then, based on the calculated battery aging rate data, the degradation assessment is performed by traversing the historical health state sequence from the starting point, and the battery's aging trajectory is plotted. This trajectory depicts how the battery's health state and performance gradually decline over time, and can be used to predict the battery's future degradation process. Finally, based on this aging trajectory, the position corresponding to the current health state estimate is located, indicating which stage of degradation the battery is currently in, such as early degradation, mid-term degradation, or near-failure. Then, a degradation model corresponding to the current stage is used to estimate the battery's remaining usable life based on the current health state. These degradation models are all constructed based on Long Short-Term Memory (LSTM) networks. After obtaining the predicted remaining useful life of the battery, this predicted remaining useful life will be added to the battery aging assessment results to provide a comprehensive assessment for the future use of the battery and ensure the efficient and reliable operation of the battery in the battery swapping cabinet.

[0046] Furthermore, based on the battery aging rate data, a degradation assessment is performed to construct a battery aging trajectory. The method includes:

[0047] Starting from the last data point in the health status history sequence, and using the battery aging rate data as the attenuation coefficient, the battery attenuation decrease data is obtained by traversing the health status history sequence from the starting point and performing charging cycle increment calculations based on the attenuation coefficient. The battery attenuation decrease data is then visualized to construct the battery aging trajectory.

[0048] Optionally, during battery aging assessment, the last data point in the historical health state sequence is selected as the starting point. This data point represents the battery's health status at the current moment. The battery aging rate data is then used as a degradation coefficient to simulate the battery's health status changes over time and predict future degradation trends. Subsequently, the historical health state sequence is traversed in reverse, processing each data point one by one. At each data point, the number of charging cycles is increased, and the battery's health status decline is calculated based on the battery's degradation coefficient. Specifically, with each additional charging cycle, the battery's health status will decrease by a certain proportion according to the degradation coefficient. The degradation calculation formula is the current health status estimate minus the product of the degradation coefficient and the health status estimate. During the reverse traversal, the battery's health status gradually deteriorates, generating degradation decline data after each charging cycle. This process continues until the starting point of the historical health state sequence is traversed, resulting in a complete battery degradation decline dataset that records the battery's health status change trend over multiple charging cycles, providing detailed evidence for subsequent battery aging trajectory construction and lifespan prediction. Next, this battery degradation dataset will be visualized and plotted as a degradation curve as the final battery aging trajectory. This trajectory shows the degradation process of the battery from its current state to the point of final use, helping to predict the remaining useful life of the battery.

[0049] In summary, the embodiments of this application have at least the following technical effects:

[0050] First, charging cycle analysis is performed on the batteries in the battery swapping cabinet according to the charging cycle of the cabinet to construct multi-dimensional charging features. Next, clustering learning is performed based on these multi-dimensional charging features to construct multiple charging mode clusters. Battery health analysis is then performed on the batteries in the swapping cabinet based on these multiple charging mode clusters to obtain battery health status parameters. Then, real-time charging data of the batteries in the swapping cabinet is retrieved and combined with the battery health status parameters to match and determine the multiple charging mode clusters, identifying the target charging mode cluster. Finally, battery aging assessment is performed on the batteries in the swapping cabinet according to the target charging mode cluster, generating battery aging assessment results. This solves the technical problems of low assessment accuracy and poor real-time performance caused by the diversity of charging modes in existing battery aging assessment methods. It achieves the technical effect of refined analysis and dynamic matching of battery health status based on charging mode clustering, improving the accuracy, scenario adaptability, and real-time responsiveness of battery aging assessment results.

[0051] Example 2, based on the same inventive concept as the battery aging assessment method for battery swapping cabinets based on charging mode clustering in the previous examples, such as... Figure 2 As shown, this application provides a battery aging assessment system for battery swapping cabinets based on charging mode clustering. The system includes:

[0052] Charging Cycle Analysis Module 11: Performs charging cycle analysis on the battery in the battery swapping cabinet according to the charging cycle of the battery swapping cabinet, and constructs multi-dimensional charging features; Battery Health Analysis Module 12: Performs cluster learning based on the multi-dimensional charging features to construct multiple charging mode clusters, and performs battery health analysis on the battery in the battery swapping cabinet based on the multiple charging mode clusters to obtain battery health status parameters; Matching Judgment Module 13: Retrieves real-time charging data of the battery in the battery swapping cabinet and combines it with the battery health status parameters to perform matching judgment on the multiple charging mode clusters, and determines the target charging mode cluster; Aging Assessment Module 14: Performs battery aging assessment on the battery in the battery swapping cabinet according to the target charging mode cluster, and generates battery aging assessment results.

[0053] Furthermore, the charging cycle analysis module 11 is used to perform the following method:

[0054] A time-series analysis is performed on the batteries in the battery swapping cabinet according to the charging cycle to extract charging time-series data, which includes an initial voltage sequence, an initial current sequence, and an initial temperature sequence. Interference analysis is then performed on the initial voltage sequence, the initial current sequence, and the initial temperature sequence to identify abnormal abrupt changes. Based on these abnormal abrupt changes, data is removed from the initial voltage sequence, the initial current sequence, and the initial temperature sequence to generate a data missing sequence. A linear interpolation algorithm is then used to complete the data missing sequence, generating a voltage sequence, a current sequence, and a temperature sequence. The voltage sequence, the current sequence, and the temperature sequence are then subjected to charging cycle analysis to construct charging cycle state parameters. Based on these charging cycle state parameters, multi-stage analysis is performed to construct the multi-dimensional charging characteristics.

[0055] Furthermore, the charging cycle analysis module 11 is used to perform the following method:

[0056] Based on the charging cycle state parameters and the voltage sequence, a stage is defined to determine the constant voltage charging stage; time-domain analysis is performed on the battery in the battery swapping cabinet according to the constant voltage charging stage to obtain the first time-domain feature; based on the charging cycle state parameters and the current sequence, a stage is defined to determine the constant current charging stage; time-domain analysis is performed on the battery in the battery swapping cabinet according to the constant current charging stage to obtain the second time-domain feature; based on the charging cycle state parameters and the temperature sequence, a stage is defined to determine the constant temperature charging stage; time-domain analysis is performed on the battery in the battery swapping cabinet according to the constant temperature charging stage to obtain the third time-domain feature; the first time-domain feature, the second time-domain feature, and the third time-domain feature are correlated and integrated to construct the multi-dimensional charging feature.

[0057] Furthermore, the battery health analysis module 12 is used to perform the following methods:

[0058] The multidimensional charging features are dimensionality reduced to obtain low-dimensional feature vectors; clustering learning is performed based on the low-dimensional feature vectors to determine K clusters, where K is a positive integer greater than 0; clustering is performed based on the K clusters to construct multiple initial data clusters; the multiple initial data clusters are traversed for optimization and verification to construct the multiple charging mode clusters.

[0059] Furthermore, the battery health analysis module 12 is used to perform the following methods:

[0060] Based on the low-dimensional feature vector, fuzzy boundary division is performed, and a candidate threshold for the number of clusters is set; the low-dimensional feature vector is contour calculated according to the candidate threshold for the number of clusters to generate multiple contour coefficients, and the contour coefficients correspond to the number of clusters; the multiple contour coefficients are arranged in descending order, and the first contour coefficient is extracted as the K number of clusters.

[0061] Furthermore, the battery health analysis module 12 is used to perform the following methods:

[0062] Data center analysis is performed based on the multiple initial data clusters to determine multiple data cluster centers; data points are identified by traversing the multiple initial data clusters, and multiple cluster sample points are extracted based on the identification results; according to the multiple initial data clusters, the multiple cluster sample points are matched with the multiple data cluster centers to generate multiple data point combinations, each of the multiple data point combinations having one and only one data cluster center and multiple cluster sample points; verification is performed based on the multiple data cluster centers and the multiple cluster sample points, and the multiple initial data clusters are optimized based on the verification results to construct the multiple charging mode clusters.

[0063] Furthermore, the battery health analysis module 12 is used to perform the following methods:

[0064] Based on the multiple data point combinations, the average distance between the centers of the multiple data clusters and the multiple cluster sample points is calculated, generating multiple average distance values; the total average distance is calculated, and if the average distance value within any data point combination exceeds Q times the total average distance, the data point combination is identified as a loose cluster, where Q is a positive integer greater than 1; when the loose cluster exists, clustering optimization is performed on the multiple data point combinations, the multiple initial data clusters are updated, and the multiple charging mode clusters are constructed.

[0065] Furthermore, the aging assessment module 14 is used to perform the following methods:

[0066] Based on the target charging mode cluster, the battery in the battery swapping cabinet is evaluated for charging over multiple charging cycles to obtain health status estimates for multiple charging cycles. These health status estimates are then serialized according to the charging time sequence to construct a historical health status sequence. Data regression training is performed according to the target charging mode cluster to set a baseline aging curve. The historical health status sequence is compared with the baseline aging curve, and dynamic time warping is performed based on the comparison results to calculate battery aging rate data. Attenuation is evaluated based on the battery aging rate data to construct a battery aging trajectory. Based on the battery aging trajectory, the battery life of the battery in the battery swapping cabinet is predicted to obtain a predicted remaining useful life value, which is then added to the battery aging evaluation result.

[0067] Furthermore, the aging assessment module 14 is used to perform the following methods:

[0068] Starting from the last data point in the health status history sequence, and using the battery aging rate data as the attenuation coefficient, the battery attenuation decrease data is obtained by traversing the health status history sequence from the starting point and performing charging cycle increment calculations based on the attenuation coefficient. The battery attenuation decrease data is then visualized to construct the battery aging trajectory.

[0069] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A battery aging evaluation method for a battery swap cabinet based on charging mode clustering, characterized in that, The method includes: Based on the charging cycle of the battery swapping cabinet, a charging cycle analysis of the battery swapping cabinet is performed to construct multi-dimensional charging characteristics; Based on the multidimensional charging features, cluster learning is performed to construct multiple charging mode clusters. Based on the multiple charging mode clusters, battery health analysis is performed on the battery swapping cabinet battery to obtain battery health status parameters. The real-time charging data of the battery in the battery swapping cabinet is retrieved and combined with the battery health status parameters to match and determine the target charging mode cluster. The battery aging assessment of the battery swapping cabinet is performed according to the target charging mode cluster, and the battery aging assessment results are generated. The method for performing battery aging assessment on the battery swapping cabinet battery according to the target charging mode cluster and generating battery aging assessment results includes: Based on the target charging mode cluster, the battery swapping cabinet battery is evaluated for charging in multiple charging cycles to obtain health status estimates for multiple charging cycles. The health status estimates of the multiple charging cycles are serialized according to the charging time sequence to construct a health status history sequence. Data regression training is performed according to the target charging mode cluster, and a baseline aging curve is set. The health status history sequence is compared with the baseline aging curve, and dynamic time normalization is performed based on the comparison results to calculate the battery aging rate data. Based on the battery aging rate data, a degradation assessment is performed to construct a battery aging trajectory. Based on the battery aging trajectory, the battery life of the battery swapping cabinet is predicted to obtain the predicted value of the remaining useful life of the battery, and the predicted value of the remaining useful life of the battery is added to the battery aging assessment result. The method for assessing battery degradation based on the battery aging rate data and constructing a battery aging trajectory includes: The last data point in the health status history sequence is taken as the starting point, and the battery aging rate data is taken as the decay coefficient. Based on the attenuation coefficient, the charging cycle is increased by traversing the historical sequence of the health state from the starting point to obtain battery attenuation data. The battery degradation data is visualized to construct the battery aging trajectory. 2.The battery aging evaluation method based on charging mode clustering of the battery swap cabinet according to claim 1, wherein, The charging cycle analysis of the batteries in the battery swapping cabinet is performed according to the charging cycle of the cabinet to construct multi-dimensional charging characteristics. The methods include: The battery in the battery swapping cabinet is analyzed according to the charging cycle of the battery swapping cabinet to extract charging time sequence data, which includes initial voltage sequence, initial current sequence and initial temperature sequence. Interference analysis is performed by traversing the initial voltage sequence, the initial current sequence, and the initial temperature sequence to identify abnormal abrupt changes; Based on the anomalous mutation points, data is removed from the initial voltage sequence, the initial current sequence, and the initial temperature sequence to generate a data missing sequence; The missing data sequence is filled in using a linear interpolation algorithm to generate voltage, current, and temperature sequences. The voltage sequence, current sequence, and temperature sequence are subjected to charging cycle analysis to construct charging cycle state parameters; Based on the charging cycle state parameters, a multi-stage analysis is performed to construct the multi-dimensional charging characteristics. 3.The battery aging evaluation method based on charging mode clustering of the battery swap station according to claim 2, wherein, The method for constructing the multidimensional charging characteristics based on the charging cycle state parameters through multi-stage analysis includes: Based on the charging cycle state parameters and the voltage sequence, the constant voltage charging stage is determined by dividing the stage into phases. The battery in the battery swapping cabinet is analyzed in the time domain according to the constant voltage charging stage to obtain the first time domain feature. Based on the charging cycle state parameters and the current sequence, the constant current charging stage is determined by dividing the process into stages. Perform time-domain analysis on the battery swapping cabinet battery according to the constant current charging stage to obtain the second time-domain characteristics; Based on the charging cycle state parameters and the temperature sequence, the constant temperature charging stage is determined by dividing the stage into phases. A time-domain analysis of the battery in the battery swapping cabinet is performed according to the constant temperature charging stage to obtain the third time-domain feature. The first time-domain feature, the second time-domain feature, and the third time-domain feature are correlated and integrated to construct the multidimensional charging feature. 4.The battery aging evaluation method based on charging mode clustering of the battery swap station according to claim 1, wherein, Clustering learning is performed based on the aforementioned multidimensional charging features to construct multiple charging pattern clusters. The method includes: The multidimensional charging features are subjected to dimensionality reduction processing to obtain a low-dimensional feature vector; Clustering learning is performed based on the low-dimensional feature vectors to determine the number of K clusters, where K is a positive integer greater than 0; Based on the K clustering numbers, clustering is performed to construct multiple initial data clusters; The multiple initial data clusters are traversed for optimization and verification to construct the multiple charging mode clusters.

5. The battery aging assessment method for battery swapping cabinets based on charging mode clustering as described in claim 4, characterized in that, Clustering learning is performed based on the low-dimensional feature vectors to determine the number of K clusters. The method includes: Based on the low-dimensional feature vector, fuzzy boundary division is performed, and a candidate threshold for the number of clusters is set. The low-dimensional feature vector is contoured according to the candidate threshold for the number of clusters to generate multiple contour coefficients, and the contour coefficients are related to the number of clusters. The multiple silhouette coefficients are sorted in descending order, and the first-order silhouette coefficient is extracted as the number of K clusters.

6. The battery aging assessment method for battery swapping cabinets based on charging pattern clustering as described in claim 4, characterized in that, The method involves iterating through the multiple initial data clusters for optimization and verification, and constructing the multiple charging mode clusters, including: Data center analysis is performed based on the multiple initial data clusters to determine multiple data cluster centers; The multiple initial data clusters are traversed to identify data points, and multiple cluster sample points are extracted based on the identification results. According to the multiple initial data clusters, the multiple cluster sample points are matched with the multiple data cluster centers to generate multiple data point combinations. Each data point combination has one and only one data cluster center and multiple cluster sample points. Based on the verification of the multiple data cluster centers and the multiple cluster sample points, the multiple initial data clusters are optimized according to the verification results to construct the multiple charging mode clusters.

7. The battery aging assessment method for battery swapping cabinets based on charging pattern clustering as described in claim 6, characterized in that, The method involves verifying the multiple data cluster centers and the multiple cluster sample points, optimizing the multiple initial data clusters based on the verification results, and constructing the multiple charging mode clusters. Based on the combination of the multiple data points, the average distance between the centers of the multiple data clusters and the sample points of the multiple clusters is calculated, and multiple average distance values ​​are generated. Calculate the total average distance. If the average distance within any data point combination exceeds Q times the total average distance, then the data point combination is identified as a loose cluster, where Q is a positive integer greater than 1. When the loose clusters exist, cluster optimization is performed on the combination of the multiple data points, the multiple initial data clusters are updated, and the multiple charging mode clusters are constructed.

8. A battery aging assessment system for battery swapping cabinets based on charging mode clustering, characterized in that, The system is used to implement the battery aging assessment method for battery swapping cabinets based on charging pattern clustering as described in any one of claims 1-7, the system comprising: Charging cycle analysis module: Performs charging cycle analysis on the batteries in the battery swapping cabinet according to the charging cycle of the battery swapping cabinet, and constructs multi-dimensional charging characteristics; Battery health analysis module: Based on the multi-dimensional charging features, cluster learning is performed to construct multiple charging mode clusters. Based on the multiple charging mode clusters, battery health analysis is performed on the battery in the battery swapping cabinet to obtain battery health status parameters. Matching and Determination Module: Retrieves real-time charging data of the battery in the battery swapping cabinet and combines it with the battery health status parameters to match and determine the target charging mode cluster. Aging assessment module: Performs battery aging assessment on the battery in the battery swapping cabinet according to the target charging mode cluster, and generates battery aging assessment results.

Images