Lithium battery state anomaly early warning method and system, computer device and storage medium

Abstract

This invention relates to the field of lithium battery state management technology, and provides a method, system, computer device, and storage medium for early warning of abnormal lithium battery states. The method includes: segmenting continuous charging segments into a charging segment dataset based on the battery's runtime sequence data to be analyzed; dividing each continuous charging segment in the charging segment dataset into current-stable segments to generate a current-stable segment dataset; extracting and analyzing the peak cell capacity features of each current-stable segment corresponding to each continuous charging segment in the current-stable segment dataset to obtain the feature vector of each individual cell within each continuous charging segment; and executing corresponding state abnormality early warning based on the state clustering results obtained by clustering the feature vectors of all individual cells within all continuous charging segments. This invention eliminates the dependence on constant current throughout the charging process, effectively improving the accuracy and efficiency of online battery state detection and abnormality early warning under complex and variable operating conditions.

Description

Technical Field

[0001] This invention relates to the field of lithium battery status management technology, and in particular to a method, system, computer device, and storage medium for early warning of abnormal lithium battery status. Background Technology

[0002] With the development of new energy technologies and the widespread application of lithium batteries in small power applications, the state safety of lithium batteries has become a key concern in the industry. In particular, lithium batteries in small power devices such as sweepers and electric vehicles are prone to aging and internal short circuits during daily use, which may lead to safety hazards such as thermal runaway during charging or operation, and even cause fires in confined spaces, threatening personal and property safety.

[0003] Existing lithium battery anomaly warning methods mainly include threshold-based judgment or model-based analysis: 1) Threshold-based anomaly warning methods based on voltage, current, and temperature are simple and easy to implement, but the warnings are delayed and cannot achieve early fault diagnosis; 2) Model-based analysis methods, which use battery internal resistance and capacity for health status assessment, mainly focus on macroscopic life assessment and are not sensitive to latent faults such as sudden internal short circuits, and it is difficult to guarantee the accuracy of real-time online estimation of battery internal resistance; at the same time, state anomaly assessment techniques based on electrochemical models or equivalent circuit models are not only complex in model design, but also have relatively harsh application conditions, poor universality, and are difficult to apply to battery state monitoring under variable operating conditions. In other words, although existing lithium battery state anomaly warning methods can meet the needs of lithium battery state analysis in some scenarios to a certain extent, they all rely heavily on constant current charging conditions throughout the process and have limitations in applicability. They cannot capture significant state characteristics from complex non-constant current charging processes, and are even less able to meet the lithium battery state management needs in small power lithium battery application scenarios with complex and variable operating conditions such as frequent start-stop, load fluctuation, temperature changes, and intelligent fast charging, and cannot guarantee the overall safety of small power lithium battery systems. Summary of the Invention

[0004] The purpose of this invention is to provide a method for early warning of abnormal lithium battery status, which at least partially solves the technical problem that the existing methods for early warning of abnormal lithium battery status have demanding application conditions, rely heavily on constant current charging conditions throughout the process, and cannot meet the requirements for online accurate monitoring and reliable early warning of abnormal lithium battery status under complex and variable charging scenarios.

[0005] To achieve the above objectives, it is necessary to provide a method, system, computer device, and storage medium for early warning of abnormal lithium battery status.

[0006] In a first aspect, embodiments of the present invention provide a method for early warning of abnormal lithium battery status, the method comprising: The acquired battery runtime sequence data to be analyzed is segmented into continuous charging segments to obtain a charging segment dataset; the charging segment dataset includes multiple continuous charging segments. The continuous charging segments in the charging segment dataset are divided into current-stable segments to generate a current-stable segment dataset; the current-stable segment dataset includes multiple current-stable segments. For each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, the peak cell capacity feature extraction and analysis are performed to obtain the feature vector of each individual cell in each continuous charging segment. Cluster analysis is performed on the feature vectors of all individual cells within all continuous charging segments to obtain state clustering results, and corresponding state anomaly warnings are executed based on the state clustering results.

[0007] Furthermore, the step of segmenting the acquired battery runtime sequence data to be analyzed into continuous charging segments to obtain a charging segment dataset includes: Based on the battery ID and timestamp in the battery runtime sequence data to be analyzed, the battery runtime sequence data to be analyzed is sorted in ascending order to obtain the first dataset; Based on the battery pack current and operating mode in the battery runtime sequence data to be analyzed, the first dataset is filtered for charging data to obtain the second dataset. Based on the timestamps, each data point in the second dataset is extended forward to obtain the third dataset; The third dataset is segmented into continuous charging segments according to preset segmentation conditions to obtain the charging segment dataset. The preset segmentation conditions include that the working mode of the preceding data point in adjacent data points is charging mode and the following data point meets preset change conditions. The preset change conditions are one of non-charging mode, battery ID change, and data point time difference exceeding a preset time difference threshold.

[0008] Further, the step of dividing each continuous charging segment in the charging segment dataset into current-stable segments to generate a current-stable segment dataset includes: Anomaly processing is performed on the voltage values ​​of each individual cell within each data point in each of the continuous charging segments to obtain the corresponding filtered voltage values. Based on the timestamp of each data point in each of the continuous charging segments, adjacent data time difference analysis is performed to obtain the corresponding duration. Based on the duration of each data point in each continuous charging segment and the battery pack current, capacity change analysis is performed to obtain the corresponding capacity increment and cumulative capacity increment. The charging segment dataset is expanded based on the capacity increment, cumulative capacity increment, duration, and filtered voltage value of each individual cell corresponding to each data point in each continuous charging segment to obtain the expanded charging segment dataset. The current mutation points of each continuous charging segment in the extended charging segment dataset are identified, and the corresponding continuous charging segments are divided into current stable segments according to the obtained mutation point sets to generate the current stable segment dataset.

[0009] Further, the step of extracting and analyzing the peak cell capacity features of each current stable segment corresponding to each continuous charging segment in the current stable segment dataset to obtain the feature vector of each individual cell within each continuous charging segment includes: Cell differential capacity analysis is performed on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, and the current stable segment dataset is expanded according to the differential capacity value of each data point to obtain a feature analysis dataset. Peak features are extracted from the differential capacity values ​​of the same cell within each current stability segment in the feature analysis dataset to obtain the peak feature set of each cell within the corresponding current stability segment; the peak features in the peak feature set include peak voltage, peak temperature, peak height, peak width, and peak area; The peak feature sets of the same single cell within each of the continuous charging segments are merged to obtain the corresponding segment cell peak feature set. Based on the number of peaks and the peak feature corresponding to the largest peak area in the segment cell peak feature set, the feature vector of the corresponding single cell is obtained.

[0010] Furthermore, the step of performing cell differential capacity analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset includes: The voltage values ​​of each individual cell within each data point in each current stabilization segment are subjected to monotonicity processing to obtain the corresponding processed voltage values. Gaussian smoothing is performed on the processed voltage value and the cumulative capacity increment of each individual cell in each data point within each current stabilization segment to obtain the corresponding smoothed voltage value and smoothed capacity value. The gradient approximation is performed based on the smoothed voltage and smoothed capacity values ​​of each individual cell at each data point within each current stabilization segment to obtain the corresponding initial differential capacity. The initial differential capacity is then subjected to amplitude limiting and smoothing processing in sequence to obtain the corresponding differential capacity value.

[0011] Furthermore, the feature vector includes the peak voltage, peak temperature, peak height, and peak width corresponding to the number of peaks and the maximum peak area; The step of performing cluster analysis on the feature vectors of all individual cells within all continuous charging segments to obtain the state clustering results includes: The feature vectors of each individual cell within each continuous charging segment are standardized to obtain the corresponding standardized feature vectors. Obtain the time interval between the current state analysis time and the previous full cluster analysis time, and determine whether the time interval is less than the preset full clustering cycle time. When the time interval is less than the preset full clustering cycle duration, the standardized feature vectors corresponding to the current state analysis time are matched and analyzed with the cluster structure obtained at the previous full clustering analysis time to obtain the state clustering result. When the time interval is equal to the preset full clustering cycle duration, density clustering analysis is performed on all the standardized feature vectors corresponding to the current state analysis time and the full standardized feature vectors corresponding to the previous full clustering analysis time to obtain the state clustering result.

[0012] Furthermore, the step of executing the corresponding state anomaly warning based on the state clustering result includes: Based on the clustering cluster to which the standardized feature vector of each individual cell in each continuous charging segment belongs in the state clustering results, an abnormal feature vector set is obtained; Based on the sample point information corresponding to each abnormal feature vector in the abnormal feature vector set, a corresponding abnormal warning output list is generated; the sample point information includes battery ID, number of battery cells, continuous charging segment number, number of cycles, and timestamp. Based on the various abnormal warning output lists, execute the corresponding status abnormal warning.

[0013] Secondly, embodiments of the present invention provide a lithium battery abnormality early warning system, the system comprising: The charging segment segmentation module is used to segment the acquired battery runtime sequence data to be analyzed into continuous charging segments to obtain a charging segment dataset; the charging segment dataset includes multiple continuous charging segments. The current stability analysis module is used to divide each continuous charging segment in the charging segment dataset into current stable segments and generate a current stable segment dataset; the current stable segment dataset includes multiple current stable segments. The state feature extraction module is used to perform cell capacity peak feature extraction and analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, and obtain the feature vector of each individual cell in each continuous charging segment. The anomaly analysis and early warning module is used to perform cluster analysis on the feature vectors of all individual cells in all the continuous charging segments to obtain state clustering results, and execute corresponding state anomaly early warnings based on the state clustering results.

[0014] Thirdly, embodiments of the present invention also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described lithium battery state abnormality early warning method.

[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described lithium battery state abnormality early warning method.

[0016] This invention provides a method, system, computer device, and storage medium for early warning of abnormal lithium battery status. The method involves segmenting the acquired battery runtime sequence data into continuous charging segments to obtain a charging segment dataset. After dividing each continuous charging segment in the charging segment dataset into current-stable segments to generate a current-stable segment dataset, the method performs cell capacity peak feature extraction and analysis on each current-stable segment corresponding to each continuous charging segment in the current-stable segment dataset to obtain the feature vector of each individual cell within each continuous charging segment. Furthermore, the method performs cluster analysis on the feature vectors of all individual cells within all continuous charging segments to obtain state clustering results. Based on the state clustering results, the method executes a corresponding technical solution for early warning of abnormal battery status. Compared with existing technologies, this lithium battery anomaly early warning method identifies local quasi-steady-state stages by dividing the collected battery operation sequence data into continuous charging segments and current-stable segments. Then, based on the peak cell capacity characteristics analysis of the current-stable segments, it extracts local quasi-steady-state features. This mechanism effectively captures significant features of changes in the battery's internal health state during complex non-constant-current charging processes and performs cluster analysis accordingly, promptly identifying potentially risky batteries. By overcoming the limitation of lithium battery state analysis relying heavily on information collected under constant-current charging conditions, it effectively improves the accuracy and efficiency of online battery state detection and anomaly early warning under complex charging conditions. This provides reliable technical support for lithium battery state management in small-power lithium battery application scenarios with complex and variable conditions such as frequent start-stop, load fluctuations, temperature changes, and intelligent fast charging, thereby ensuring the overall safety of small-power lithium battery systems. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the lithium battery state abnormality early warning method in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the lithium battery abnormality early warning system in an embodiment of the present invention; Figure 3 This is an internal structural diagram of the computer device in an embodiment of the present invention; The attached figures are labeled as follows: 1. Charging segment division module; 2. Current stability analysis module; 3. State feature extraction module; 4. Anomaly analysis and early warning module. Detailed Implementation

[0018] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Obviously, the embodiments described below are only part of the embodiments of this invention and are used to illustrate the invention, but are not intended to limit the scope of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] The lithium battery state anomaly early warning method provided by this invention can be understood as addressing the application limitations of existing lithium battery state anomaly early warning methods. It proposes a method that transforms the conditional dependence on constant current throughout the charging process into local quasi-steady-state identification, extracting the most significant local feature segments from the complex non-constant current charging process for state analysis. This improves the accuracy of battery anomaly detection and ensures the identification of potentially risky batteries before safety accidents occur. The method combines the changing characteristics of the charging capacity-voltage curve (QV curve) with a clustering algorithm for battery state analysis and anomaly early warning. The following embodiments will provide a detailed description of the lithium battery state anomaly early warning method of this invention.

[0020] In one embodiment, such as Figure 1 As shown, a method for early warning of abnormal lithium battery status is provided, including: S11. The acquired battery runtime sequence data to be analyzed is segmented into continuous charging segments to obtain a charging segment dataset. The battery runtime sequence data to be analyzed can be understood as the time-series dataset that can meet the data volume requirements of a single analysis by periodically collecting the operating status data of all lithium batteries involved in the target battery project management (such as battery management projects for electric vehicles, floor scrubbers, etc.) in actual applications. In order to ensure the comprehensiveness and reliability of battery status analysis, this embodiment preferably sets the battery runtime sequence data to be analyzed to include battery ID, timestamp, battery pack current, working mode, number of cycles, maximum cell temperature, and voltage value of each individual cell. All of these can be obtained through the corresponding project information storage database. The project information storage database can be a cloud database or a database deployed in a terminal server. No specific limitation is made here.

[0021] In practical applications, after the lithium battery status anomaly early warning method is initiated, it first loads the basic configuration information related to the target battery project management. Based on this basic configuration information, it establishes a connection with the project information storage database to obtain the required battery runtime sequence data to be analyzed. The specific content of the basic configuration information in this embodiment varies depending on the actual application scenario and may include the data source environment type, target server type, connection parameters for different servers, etc. The connection parameters may include host address, port number, username, password, and database name, etc. The specific process for obtaining the battery runtime sequence data to be analyzed is as follows: Based on this basic configuration information, it is possible to identify whether the data source environment is a local development environment or a production environment. Then, based on the data source environment, the corresponding target server type can be obtained, and based on the target server type, the corresponding server connection parameters can be obtained.

[0022] Based on the server connection parameters obtained above, a secure session is established with the project information storage database. Relevant battery operating status data is then retrieved from the corresponding form in the project information storage database based on the required data content for subsequent analysis. For example, if the project information storage database is a MySQL database, a secure session can be established using `mysql.connector.connect`, and then SQL query statements (e.g., `SELECT ...`) can be used. FROM battery_info_table) retrieves the runtime sequence data of the battery to be analyzed, which can be denoted as: Each data point in the battery runtime sequence data to be analyzed , can be represented as: In the formula, The index of the data points in the battery runtime sequence data to be analyzed is... , This represents the total number of data points. , , , , and The first The data points include the battery ID, timestamp (absolute time), battery pack current (unit: mA), operating mode (e.g., charging, resting, discharging), number of cycles, and maximum cell temperature (unit: °C). For the first The first data point The voltage value of each individual battery cell (unit: mV), and , This refers to the number of individual battery cells.

[0023] It should be noted that if the battery operating status data obtained directly from the corresponding form in the project information storage database does not conform to the definition of the above data points in actual application, a data field extraction mapping can be added to analyze the obtained raw data to obtain the battery operating sequence data to be analyzed in the required structure. For the specific implementation process, please refer to the existing related acquisition technology implementation of the data items involved in the battery operating sequence data to be analyzed, which will not be detailed here.

[0024] Considering that the battery runtime sequence data obtained through the above methods and steps is not entirely valid data from the charging process, in order to ensure comprehensive analysis of the battery state during charging while conserving data analysis resources and improving data analysis efficiency, this embodiment preferably performs valid charging data filtering, expansion, and charging segment division processing on the runtime sequence data of each battery pack sequentially to obtain a charging segment dataset including multiple continuous charging segments; specifically, the step of segmenting the acquired battery runtime sequence data to obtain the charging segment dataset includes: Based on the battery ID and timestamp in the battery runtime sequence data to be analyzed, the data is sorted in ascending order to obtain the first dataset; that is, the battery runtime sequence data to be analyzed. All data points are sorted first by battery ID in ascending order, and then by timestamp in ascending order for data points with the same battery ID (same battery). The first dataset, denoted as [dataset name], is a sequential arrangement of all operational status data for the same battery. .

[0025] Based on the battery pack current and operating mode in the battery runtime sequence data to be analyzed, the first dataset is filtered for charging data to obtain the second dataset; the second dataset can be understood as being derived from the first dataset based on the battery pack current and operating mode. The dataset consists of all charging data obtained through data point filtering; in practical applications, it can be derived from the first dataset. By selecting all data points whose operating mode is charging mode and whose battery pack current is greater than or equal to a preset current threshold (e.g., 100mA), a data subset can be obtained. , : , The preset current threshold is, i.e. This is the second dataset required.

[0026] Based on timestamps, the data points in the second dataset are extended forward to obtain the third dataset. The third dataset is derived by extending the data points of each data point in the second dataset forward over time. This third dataset captures the state data before each charge cycle, facilitating subsequent anomaly observation and analysis. The specific acquisition process is as follows: For each data point in the second dataset According to its first dataset The index position in the table is expanded forward to multiple data points (e.g., 5), and the timestamps corresponding to these newly expanded data points are... If the time difference between the timestamps does not exceed a preset time difference value (e.g., 60 seconds), then... Corresponding extended set For example, suppose for The first element In the first dataset The index position is the 100th data point. ,like timestamp and If the timestamps differ by a preset time difference value, an extended set can be obtained. ;like First dataset If the index position is the second data point, then only the following can be selected: Expand the set.

[0027] Each The resulting extended sets are merged into a large dataset, and duplicates are removed to obtain the third dataset, which is used for subsequent charging segment segmentation. Assume... The 1st, 2nd, 100th, ... The expanded sets obtained by each element are respectively , … , will this The sets are merged into one large dataset, and then combined with... The third dataset is obtained by performing a union of the sets and removing duplicate elements. Finally, reset the element indices, starting the marking from 1, to obtain: , For the third dataset, the first Data points, For the third dataset The total number of data points in the dataset.

[0028] The third dataset is segmented into continuous charging segments based on preset segmentation criteria to obtain a charging segment dataset. The preset segmentation criteria include that the preceding data point in an adjacent data point is in charging mode and the following data point meets preset change conditions. The preset change conditions are one of the following: non-charging mode, battery ID change, or the time difference between data points exceeding a preset time difference threshold. The charging segment dataset is a collection composed of continuous charging segments from the third dataset. The specific process for obtaining the charging segment dataset is as follows: Based on the above preset segmentation conditions, the following segmentation indicator function is defined: In the formula, = = in, and The third dataset contains the first... The data point and the first -1 data point battery ID; For the third dataset, the first The change in each data point; and The third dataset contains the first... The data point and the first -1 timestamp of a data point; For the third dataset, the first The time difference between data points; The preset time difference threshold for adjacent data points is preferably set to 300s; and The third dataset contains the first... The data point and the first Working mode with -1 data point; For the third dataset, the first The variable indicating whether the nth data point is a dividing point between different charging segments is a variable; a value of 1 indicates that the nth data point is a dividing point between different charging segments. The data point is the dividing point (i.e., from the th data point) Each data point marks the beginning of a new charging segment, and a value of 0 indicates that it is not a segmentation point.

[0029] Due to the third dataset DE The dataset contains all charging data and pre-charging state data. Finding pre-charging state data (resting or discharging), or finding data with a long duration, indicates a new charging cycle. Therefore, based on the aforementioned preset segmentation conditions, the set of segmentation points in the third dataset can be accurately obtained. Then based on the set of split points The segmentation points in the dataset divide the third dataset into several segments. In principle, each segment obtained at this point can be used as a charging segment for subsequent analysis. However, considering that the number of data points in a segment is too small to represent a complete charging process, all the aforementioned segments are filtered for validity. Segments with fewer than a preset number of data points are deleted, resulting in the final charging segment dataset. , For the first The battery data sequence corresponding to each continuous charging segment, and each continuous charging segment This includes several items related to the aforementioned Data points with the same data point format.

[0030] This embodiment first sorts the battery data in ascending order based on battery ID and timestamp to ensure the continuity of each battery data in the time dimension. Then, it filters the charging data based on battery pack current and operating mode to effectively exclude data from charging scenarios irrelevant to the charging conditions, such as discharging and resting. Finally, it performs forward expansion and charging segmentation on the charging data points to obtain the state data of each charging segment corresponding to an independent charging event. This makes it easier to capture data fluctuations in complex charging processes and provides a reliable data foundation for subsequent battery life cycle state management.

[0031] S12. Divide each continuous charging segment in the charging segment dataset into current-stable segments to generate a current-stable segment dataset; the current-stable segment dataset includes multiple current-stable segments; wherein, the current-stable segment dataset can be understood as a dataset formed by further dividing each continuous charging segment in the charging segment dataset into multiple current-stable segments based on the identification of current abrupt change points; specifically, the step of dividing each continuous charging segment in the charging segment dataset into current-stable segments to generate a current-stable segment dataset includes: Anomaly processing is performed on the voltage values ​​of each individual cell at each data point in each continuous charging segment to obtain the corresponding filtered voltage values. Anomaly processing can be understood as the process of validating the voltage values ​​of each individual cell at each data point and uniformly assigning invalid values.

[0032] Assumption For charging fragment dataset The Middle In the first continuous charging segment Data points Inner The voltage value of each individual battery cell; among which... , The total number of data points in the charging segment data, i.e. ; , For the first The number of data points within a continuous charging segment, i.e. ; , Let be the number of individual battery cells; then, the voltage value of each individual cell is filtered according to the following formula: In the formula, The minimum voltage value for a single battery cell is preferably set to 500mV; The minimum voltage of a single battery cell is preferably set to 5000mV; The default invalid voltage value is set. For charging fragment dataset The Middle In the first continuous charging segment Within the data point, the first The filtered voltage value of a single battery cell.

[0033] Based on the timestamp of each data point in each continuous charging segment, adjacent data time difference analysis is performed to obtain the corresponding duration; that is, the duration of each data point in each continuous charging segment can be expressed as: =0 In the formula, and These are the charging segment data points. In the first continuous charging segment + 1 data point and the first Timestamps of each data point; For the charging segment dataset, the first In the first continuous charging segment The duration of each data point.

[0034] Capacity change analysis is performed based on the duration and battery pack current of each data point in each continuous charging segment to obtain the corresponding capacity increment and cumulative capacity increment. The capacity change analysis can be understood as first calculating the capacity increment between adjacent data points, and then calculating the cumulative capacity value from the first data point to the current data point in the corresponding continuous charging segment. The calculation process for the corresponding capacity increment and cumulative capacity increment is as follows: Assumption Represents the first charging segment in the dataset. In the first continuous charging segment The battery pack current (unit: mA) for each data point corresponds to the capacity increment as follows: Where 3600 is the coefficient for converting milliampere-seconds (mAs) to milliampere-hours (mAh); For the charging segment dataset, the first In the first continuous charging segment Capacity increment at each data point; Based on the capacity increment at each data point obtained above, the corresponding cumulative capacity increment can be calculated using the following formula: in, For the charging segment dataset, the first In the first continuous charging segment Cumulative capacity increment at each data point; For the first In the first continuous charging segment Capacity increment at each data point.

[0035] The charging segment dataset is expanded based on the capacity increment, cumulative capacity increment, duration, and filtered voltage value of each individual cell for each data point in each continuous charging segment, resulting in an expanded charging segment dataset; that is, the expanded charging segment dataset. Each data point in each continuous charging segment can be represented as: In the formula, Expanding the dataset for charging segments The Middle In the first continuous charging segment Each data point represents a portion of the charging segment dataset. The Middle In the first continuous charging segment Data points The expanded version is a data point format.

[0036] The current abrupt change points in each continuous charging segment of the extended charging segment dataset are identified, and the corresponding continuous charging segments are divided into current stable segments based on the obtained abrupt change point sets to generate the current stable segment dataset. The current abrupt change point identification can be understood as determining the current abrupt change point based on the magnitude of the change in battery pack current between adjacent data points within the same continuous charging segment. The specific judgment process includes: First, calculate the current change at each data point within the same continuous charging segment, expressed as: In the formula, and Expand the dataset for charging segments respectively The Middle In the first continuous charging segment The data point and the first - Battery pack current at one data point; Expand the dataset for charging segments. In the first continuous charging segment The change in current at each data point.

[0037] The current change at each data point in the continuous charging segment is checked based on the following conditions. If the conditions are met, the data point is considered to be a current segment switching point, i.e., a current abrupt change point. Otherwise, the data point is not a current abrupt change point: In the formula, The threshold value for current change can be set based on actual application requirements. In this embodiment, it is preferably set to 1000mA.

[0038] The above method yields the set of mutation points for each continuous charging segment. Using each mutation point in the set as a boundary, each continuous charging segment can be divided into several subsets, with each subset corresponding to a current-stable time-series data set. This results in the initial partitioned dataset, which can be represented as: in, For the initial partitioning of the dataset The Middle Each continuous charging segment is equivalent to expanding the dataset with charging segments. The Middle One continuous charging segment; For the initial partitioning of the dataset The Middle In the first continuous charging segment One current segment; For the initial partitioning of the dataset The Middle In the first continuous charging segment Within the current segment, the first 1 data point, and The data format is the same as the aforementioned middle The data formats are the same.

[0039] The initial dataset obtained above can theoretically be used as a dataset for current-stable segments. However, considering that in practical applications, some current segments may have excessively low average current and short durations, making it impossible to accurately analyze the peak characteristics of cell capacity, this embodiment preferably performs an effectiveness analysis on each current segment obtained above based on the following effective detection conditions (simultaneously satisfied) to ensure the effectiveness of each current-stable segment: 1) Average current is greater than the minimum current threshold: =( - ) In the formula, and The initial partitioning of the dataset is as follows: The Middle In the first continuous charging segment The start and end timestamps of each current segment; For the initial partitioning of the dataset The Middle In the first continuous charging segment The total duration of each current segment; ) is the initial partitioned dataset The Middle In the first continuous charging segment Within the current segment, the first Each data point corresponds to a time Battery pack current at the location; Minimum current threshold; 2) The total duration of the current segment is greater than the shortest duration: In the formula, This represents the shortest duration.

[0040] Based on the above effective detection conditions, from each of the above... By extracting the effective current segments and summarizing all effective current segments as current-stabilized segments, the required current-stabilized segment dataset can be obtained, which can be represented as: in, For current-stabilized segment datasets The Middle One continuous charging segment; For current-stabilized segment datasets The Middle In the first continuous charging segment A current-stabilized segment; For current-stabilized segment datasets The Middle In the first continuous charging segment Within the current stabilization segment, the first 1 data point, and The data format is the same as the aforementioned middle The data format is also the same.

[0041] This embodiment not only effectively avoids the adverse effects of individual cell anomalies on subsequent voltage and capacity change analysis by cleaning the individual cell voltage of each data point in each continuous charging segment, but also expands the data items based on duration, capacity increment, and cumulative capacity increment to provide a comprehensive and reliable data foundation for subsequent voltage and capacity change analysis. Furthermore, it divides the current into stable segments based on abrupt change point identification, ensuring that the current within each minimum time-series data segment used for individual cell status remains relatively stable, thereby avoiding interference from other factors and ensuring the reliability and high consistency of the voltage difference analysis of individual cells under stable current.

[0042] S13. Perform cell capacity peak feature extraction and analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset to obtain the feature vector of each individual cell in each continuous charging segment; wherein, cell capacity peak feature extraction can be understood as the analysis of the voltage capacity change characteristics (differential capacity) of each individual cell in each current stable segment; specifically, the steps of performing cell capacity peak feature extraction and analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset to obtain the feature vector of each individual cell in each continuous charging segment include: Cell differential capacity analysis is performed on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset. Based on the obtained differential capacity values ​​of each data point, the current stable segment dataset is expanded to obtain a feature analysis dataset. Cell differential capacity analysis can be understood as processing the individual cell voltage values ​​at each data point within each current stable segment to ensure monotonically increasing voltage, and then calculating the differential capacity based on the smoothed voltage and capacity cumulative increments. Specifically, the steps for performing cell differential capacity analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset include: Monotonicity processing is applied to the voltage values ​​of each individual cell within each data point in each current stabilization segment to obtain the corresponding processed voltage value. This monotonicity processing can be understood as applying validity processing to the current voltage value based on the cumulative maximum voltage of each individual cell. The resulting processed voltage value can be expressed as: In the formula, For current-stabilized segment datasets The Middle In the first continuous charging segment The first current stabilization segment Within the data point, the first The voltage value of each individual battery cell; and These are the current-stabilized segment datasets. The Middle In the first continuous charging segment The first current stabilization segment Within the data point, the first The maximum cumulative voltage and the processed voltage value of each individual cell.

[0043] Gaussian smoothing is applied to the processed voltage and cumulative capacity increments of each individual cell at each data point within each current stabilization segment to obtain the corresponding smoothed voltage and capacity values. The Gaussian smoothing process can be implemented based on the following formula: In the formula, The standard deviation is expressed as The Gaussian kernel function; * indicates the convolution operation; for The corresponding smoothed voltage value; For current-stabilized segment datasets The Middle In the first continuous charging segment Within the current stabilization segment, the first Cumulative capacity increment of each data point; for The corresponding smooth capacity value.

[0044] Based on the smoothed voltage and capacity values ​​of each individual cell at each data point within each current stabilization segment, gradient approximation is performed to obtain the corresponding initial differential capacity. This initial differential capacity is then sequentially limited and smoothed to obtain the corresponding differential capacity value. The initial differential capacity can be understood as the dQ / dV value, which can be obtained based on the following gradient approximation formula: In the formula, For current-stabilized segment datasets The Middle In the first continuous charging segment Within the current stabilization segment, the first The initial differential capacity of each individual battery cell; It is a discrete gradient operator.

[0045] It should be noted that the process of obtaining the initial differential capacity based on the above formula can be implemented with reference to relevant existing technologies, and will not be described in detail here. To avoid division-by-zero errors and outliers in the initial differential capacity, and to ensure the numerical stability of the differential capacity used for subsequent analysis, this embodiment preferably processes the obtained initial differential capacity based on the following limiting constraints: in, The maximum value threshold is preferably set to [value]. ; For the sign function; For the initial differential capacity The limited differential capacity obtained by applying the limiting process.

[0046] After obtaining the limited differential capacity using the above method, Gaussian smoothing is then applied to smooth the derivative: in, The standard deviation is expressed as The Gaussian kernel function, and It can be set to 1 or a smaller value; For current-stabilized segment datasets The Middle In the first continuous charging segment Within the current stabilization segment, the first The differential capacity value of a single battery cell.

[0047] This implementation effectively solves problems such as noise interference, data anomalies, and monotonic distortion in the original voltage / capacity data during cell charging by processing the monotonicity of the individual cell voltage value and smoothing the cumulative increment of voltage value and capacity, as well as the method of obtaining the differential capacity value of each individual cell based on gradient approximation calculation, amplitude limiting, and smoothing. This ensures that the obtained differential capacity value can reflect the electrochemical characteristics of the cell in a true, smooth, and stable manner, providing reliable data support for the effectiveness, accuracy, and robustness of subsequent extraction of state characteristics based on differential capacity curves.

[0048] After obtaining the differential capacity values ​​of each individual cell at each data point in each current-stabilized segment using the above methods and steps, the current-stabilized segment dataset is... Each data point in The differential capacity values ​​of all individual cells are used as data augmentation terms to expand the corresponding data points, resulting in a feature analysis dataset that facilitates peak feature extraction of the differential capacity curve for each individual cell. That is, each data point in the feature analysis dataset. It can be represented as: In the formula, For feature analysis datasets The Middle In the first continuous charging segment Within the current stabilization segment, the first One data point; and Feature analysis datasets The Middle In the first continuous charging segment Within the current stabilization segment, the first Battery ID and timestamp for each data point.

[0049] Peak features are extracted from the differential capacity values ​​of the same cell within each current-stabilized segment of the feature analysis dataset to obtain the peak feature set for each cell within the corresponding current-stabilized segment. Peak feature extraction can be understood as using a peak detection algorithm to identify the characteristic peaks of the differential capacity curves corresponding to each cell within each current-stabilized segment, and then extracting features from each detected peak. The resulting peak features include peak voltage, peak temperature, peak height, peak width, and peak area. The specific process is as follows: For each current steady segment in the feature analysis dataset The differential capacity curve (dQ / dV-Voltage) composed of the differential capacity values ​​of each individual cell is used to identify characteristic peaks using a peak detection algorithm, resulting in the corresponding set of peak values. in, For feature analysis of the dataset, the first In the first continuous charging segment Within the current stabilization segment, the first Differential capacity curve of individual battery cells The corresponding set of peak values; The minimum peak height threshold is preferably set to [value]. ; To minimize the peak width, it is preferably set to ; For minimum protrusion, it is preferably set to ; The minimum distance between peaks is preferably set to ; For peak detection algorithms, existing technologies can be referenced for implementation.

[0050] For each peak in the peak set By performing feature extraction, the corresponding peak features can be obtained: 1) Peak potential: In the formula, For feature analysis of the dataset, the first In the first continuous charging segment Within the current stabilization segment, the first Peak values ​​of individual battery cells The first in One peak; for The corresponding timestamp; for The corresponding voltage value; 2) Temperature corresponding to the peak value: In the formula, for The corresponding maximum temperature of the battery cell; 3) Peak height: In the formula, for The corresponding differential capacity value; 4) Peak width: ; 5) Peak area (approximated using trapezoidal integral): In the formula, for The corresponding peak area; For timestamps The corresponding differential capacity value; and Each is a timestamp +1 and timestamp The corresponding number The voltage value of each individual battery cell.

[0051] The peak feature sets of the same single cell within each continuous charging segment are merged to obtain the corresponding segment cell peak feature set. Based on the number of peaks and the peak feature corresponding to the largest peak area in the segment cell peak feature set, the feature vector of the corresponding single cell is obtained. In practical applications, the peak feature sets of the same single cell belonging to the same continuous charging segment but different current stable segments are merged into a large set. Then, all peak features are sorted in descending order based on peak area to obtain the required segment cell peak feature set. Next, the number of peaks in the segment cell peak feature set is statistically analyzed and merged with the first peak feature in the segment cell peak feature set to generate the feature vector of the corresponding single cell within the continuous charging segment, expressed as: in, For feature analysis of the dataset, the first Within the first continuous charging segment Peak number of individual cells; superscript The peak feature concentration location of the segment cell is the corresponding value of the first. In this embodiment, the peak feature with the largest peak area among multiple current stable segments of the same continuous charging segment is considered to be the most significant. For feature analysis of the dataset, the first Within the first continuous charging segment The feature vectors of individual battery cells can be used for subsequent cluster analysis.

[0052] This embodiment uses a combination of peak potential, temperature, peak height, peak width, and peak area to construct a feature vector for single-cell state analysis. This can more accurately describe the peak shape in the QV curve and more accurately reflect the battery usage under this continuous charging segment, providing a sample space with higher sensitivity and higher correlation for subsequent clustering models.

[0053] S14. Perform cluster analysis on the feature vectors of all individual cells within all continuous charging segments to obtain state clustering results, and execute corresponding state anomaly warnings based on the state clustering results. The state clustering results can be understood as the state classification results of each individual cell in the battery pack obtained by unsupervised learning of the feature vectors of all individual cells within all continuous charging segments using a density clustering algorithm. Specifically, the steps for performing cluster analysis on the feature vectors of all individual cells within all continuous charging segments to obtain state clustering results include: The feature vectors of each individual cell within each continuous charging segment are standardized to obtain the corresponding standardized feature vectors. The standardization process can employ existing standardization techniques, and Z-score standardization is preferably used in this embodiment. The specific process of obtaining each standardized feature vector will not be detailed here.

[0054] The time interval between the current state analysis time and the previous full cluster analysis time is obtained, and it is determined whether the time interval is less than the preset full clustering cycle time. The current state analysis time can be understood as the time when the state analysis is performed based on the collected battery runtime sequence data to be analyzed, and the corresponding previous full cluster analysis time is the time when all battery full life cycle charging data were executed sequentially for cluster analysis. The preset full clustering cycle time can be set based on actual application needs. For example, when the frequency of acquiring the battery runtime sequence data to be analyzed is once every 6 hours, the preset full clustering cycle time can be set to a longer period such as 1 to 3 months. There is no specific limitation here.

[0055] When the interval between current state analysis times is less than the preset full clustering period, the standardized feature vectors corresponding to the current state analysis time are matched with the cluster structure obtained at the previous full clustering analysis time to obtain the state clustering result. The cluster structure obtained at the previous full clustering analysis time can be understood as the neighborhood relationship (core point, boundary point group, etc.) constructed by the previous full clustering. In practical applications, if the battery runtime sequence data to be analyzed is collected every 6 hours, that is, the classification of new samples is performed every 6 hours, the newly added feature vector sample points captured in the last 6 hours are input into the neighborhood relationship constructed by the previous full clustering for matching and judgment. It is calculated whether these new sample points fall within the neighborhood of the existing clusters and are assigned clustering label values ​​to obtain the required state clustering result. It should be noted that this operation does not dynamically update the cluster structure. It only assigns each new sample to an existing cluster or marks it as noise. The samples marked as noise are judged as cell state abnormalities for subsequent abnormal warning operations.

[0056] When the time interval is equal to the preset full clustering period, density clustering analysis is performed on the standardized feature vector corresponding to the current state analysis time and the full standardized feature vector corresponding to the previous full clustering analysis time to obtain the state clustering result; that is, a full clustering analysis needs to be performed every preset full clustering period to ensure the adaptability of the classification result to data changes. In practical applications, if the preset full clustering cycle is set to 3 months, then full DBSCAN clustering of all feature vector sample points needs to be performed again every 3 months to identify all high-density regions in the sample space and update the cluster structure, which can cope with changes in the abnormal pattern structure of the battery group. For battery project stages where the abnormal pattern evolves rapidly (such as battery groups that have just been put into operation), a shorter interval can be set, such as 1 month. For battery project stages where the abnormal pattern has stabilized (such as battery groups that have been in service for more than 2 years or battery groups that have begun to enter a stable phase of retiring batteries), a longer interval can be set. For battery groups that are about to end their service life, a reduced feature space can be constructed by sampling historical normal samples based on server resources, and then clustering can be performed, or full clustering can be skipped. It should be noted that for the first execution of the abnormal status warning analysis, it is necessary to form a feature space by combining all standardized feature vectors corresponding to the full battery runtime sequence data in the project information storage database, and perform comprehensive cluster analysis. Subsequent cluster analysis processes can be implemented using the above method. The implementation process of DBSCAN clustering in this embodiment can refer to relevant existing technologies, and will not be described in detail here.

[0057] The state anomaly analysis strategy combining full clustering and incremental matching provided in this embodiment can not only accurately characterize the real state differences of different cells under charging conditions through clustering results, thereby improving the sensitivity and accuracy of abnormal cell identification, but also effectively reduce the consumption of computing and memory resources for state analysis, ensuring the real-time performance and efficiency of battery state anomaly analysis.

[0058] After obtaining the state clustering results, potentially risky abnormal batteries can be identified based on the proportion of cluster labels in each cluster, and corresponding state anomaly warnings can be executed for each abnormal battery. Specifically, the steps for executing corresponding state anomaly warnings based on the state clustering results include: Based on the clusters to which the standardized feature vectors of each individual cell within each continuous charging segment belong in the state clustering results, an abnormal feature vector set is obtained. In practical applications, in addition to the noise point clusters in the state clustering results, it is also necessary to calculate the sample proportion analysis of each cluster: when the sample proportion of a certain cluster is higher than a preset proportion threshold (e.g., 5%), the cluster is a cluster with highly dense and similar safety features in the battery group, and it is regarded as a normal battery; correspondingly, when the sample proportion of a certain cluster is lower than the preset proportion threshold, the cluster is an outlier or low-density area, and it is regarded as a special cluster; if the feature vector sample of a certain individual cell is marked as a noise point or a special cluster, it is considered that its QV curve features deviate from the normal group, and it is an abnormal individual cell with potential risks. Based on this, an abnormal feature vector set can be obtained; at the same time, the nearest distance between the abnormal feature vector sample point and the core point of the main cluster can also be calculated. The greater the distance, the more serious the potential risk of the sample point, that is, the abnormal risk level of the abnormal feature vector sample point is determined.

[0059] Based on the sample point information corresponding to each abnormal feature vector in the abnormal feature vector set, a corresponding abnormal warning output list is generated; wherein, the sample point information can be understood as data information that facilitates the location and analysis of battery abnormalities. In this embodiment, the sample point information is preferably set to include battery ID, number of cells, serial number of continuous charging segment, number of cycles, and timestamp, etc.

[0060] Based on the various anomaly warning output lists, corresponding status anomaly warnings are executed. This involves uploading each anomaly warning output list to the project information storage database via a secure session between the server and the database. Then, other technologies are used to send warning information to the battery holder, manufacturer, or operator, allowing them to take different measures based on the varying levels of anomaly risk. It should be noted that in practical applications, status anomaly warnings can be implemented using existing technologies, and specific limitations are not specified here.

[0061] This invention provides a method for segmenting continuous charging segments into a dataset. This involves dividing the acquired battery runtime sequence data (including battery ID, timestamp, battery pack current, operating mode, cycle count, maximum cell temperature, and voltage values ​​of individual cells) into current-stable segments. Then, a current-stable segment dataset is generated by further segmenting each continuous charging segment within the dataset into current-stable segments. Finally, peak cell capacity features are extracted and analyzed for each current-stable segment within each continuous charging segment, yielding feature vectors for each individual cell within each continuous charging segment. Cluster analysis is then performed on the feature vectors of all individual cells across all continuous charging segments to obtain state clustering results. Based on these state clustering results, a corresponding state anomaly warning system is implemented. This process is based on the sequential execution of continuous charging segment segmentation and current-stable segment division. This process involves extracting QV curve features from current-stable segments and then using the segments with the most significant features as state features for constructing the battery cluster feature space. This transforms the conditional dependence on constant current throughout charging into a local quasi-steady-state identification and anomaly analysis mechanism. It effectively captures significant features of changes in the battery's internal health state during complex, non-constant-current charging processes, promptly identifying potentially risky batteries. This significantly improves the accuracy and efficiency of online battery status detection and anomaly warning under complex charging conditions. It provides reliable technical support for lithium battery status management in small-power lithium battery applications with complex and variable conditions such as frequent start-stop cycles, load fluctuations, temperature changes, and intelligent fast charging, thereby ensuring the overall safety of small-power lithium battery systems. It also provides a technical reference for achieving online, real-time safety monitoring of large-scale battery clusters, demonstrating high practical value.

[0062] It should be noted that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0063] In one embodiment, such as Figure 2 As shown, a lithium battery abnormality early warning system is provided, the system comprising: The charging segment segmentation module 1 is used to segment the acquired battery runtime sequence data to be analyzed into continuous charging segments to obtain a charging segment dataset; the charging segment dataset includes multiple continuous charging segments. The current stability analysis module 2 is used to divide each continuous charging segment in the charging segment dataset into current stability segments and generate a current stability segment dataset; the current stability segment dataset includes multiple current stability segments; State feature extraction module 3 is used to perform cell capacity peak feature extraction and analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, and obtain the feature vector of each individual cell in each continuous charging segment. The anomaly analysis and early warning module 4 is used to perform cluster analysis on the feature vectors of all individual cells in all the continuous charging segments to obtain state clustering results, and execute corresponding state anomaly early warnings based on the state clustering results.

[0064] Specific limitations regarding the lithium battery anomaly warning system can be found in the limitations of the lithium battery anomaly warning method described above, and the corresponding technical effects are equivalent, so they will not be repeated here. Each module in the aforementioned lithium battery anomaly warning system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0065] Figure 3 An internal structural diagram of a computer device is shown in one embodiment. This computer device may specifically be a terminal or a server. Figure 3 As shown, the computer device includes a processor, memory, network interface, display, camera, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it can implement a method for early warning of abnormal lithium battery status. The display screen can be an LCD screen or an e-ink display screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device casing, or an external keyboard, touchpad, or mouse.

[0066] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. Specific computing devices may include more or fewer components than those shown in the figure, or combine certain components, or have the same component arrangement.

[0067] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0068] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0069] In summary, the lithium battery state anomaly early warning method, system, computer device, and storage medium provided by this invention are based on sequentially performing operations such as continuous charging segment segmentation, current stable segment division, extracting QV curve features from the current stable segment, and then using the segment with the most significant features as the state features of the continuous charging segment to construct the battery group feature space. This realizes the transformation of the conditional dependence on constant current throughout the charging process into a state anomaly analysis mechanism for local quasi-steady-state identification. It can effectively capture significant features of changes in the internal health state of the battery from the complex non-constant current charging process, promptly identify potentially risky batteries, and effectively improve the accuracy and efficiency of online battery state detection and anomaly early warning under complex charging conditions. It provides reliable technical support for lithium battery state management in small power lithium battery application scenarios with complex and variable conditions such as frequent start-stop, load fluctuation, temperature change, and intelligent fast charging, thereby ensuring the overall safety of small power lithium battery systems. It also provides a technical reference for realizing online and real-time safety monitoring of large-scale battery groups and has high practical value.

[0070] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0071] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for early warning of abnormal lithium battery status, characterized in that, The method includes: The acquired battery runtime sequence data to be analyzed is segmented into continuous charging segments to obtain a charging segment dataset; the charging segment dataset includes multiple continuous charging segments. The continuous charging segments in the charging segment dataset are divided into current-stable segments to generate a current-stable segment dataset; the current-stable segment dataset includes multiple current-stable segments. For each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, the peak cell capacity feature extraction and analysis are performed to obtain the feature vector of each individual cell in each continuous charging segment. Cluster analysis is performed on the feature vectors of all individual cells within all continuous charging segments to obtain state clustering results, and corresponding state anomaly warnings are executed based on the state clustering results.

2. The lithium battery abnormality early warning method as described in claim 1, characterized in that, The step of segmenting the acquired battery runtime sequence data to obtain a charging segment dataset includes: Based on the battery ID and timestamp in the battery runtime sequence data to be analyzed, the battery runtime sequence data to be analyzed is sorted in ascending order to obtain the first dataset; Based on the battery pack current and operating mode in the battery runtime sequence data to be analyzed, the first dataset is filtered for charging data to obtain the second dataset. Based on the timestamps, each data point in the second dataset is extended forward to obtain the third dataset; The third dataset is segmented into continuous charging segments according to preset segmentation conditions to obtain the charging segment dataset. The preset segmentation conditions include that the working mode of the preceding data point in adjacent data points is charging mode and the following data point meets preset change conditions. The preset change conditions are one of non-charging mode, battery ID change, and data point time difference exceeding a preset time difference threshold.

3. The lithium battery abnormality early warning method as described in claim 1, characterized in that, The step of dividing each continuous charging segment in the charging segment dataset into current-stable segments to generate a current-stable segment dataset includes: Anomaly processing is performed on the voltage values ​​of each individual cell within each data point in each of the continuous charging segments to obtain the corresponding filtered voltage values. Based on the timestamp of each data point in each of the continuous charging segments, adjacent data time difference analysis is performed to obtain the corresponding duration. Based on the duration of each data point in each continuous charging segment and the battery pack current, capacity change analysis is performed to obtain the corresponding capacity increment and cumulative capacity increment. The charging segment dataset is expanded based on the capacity increment, cumulative capacity increment, duration, and filtered voltage value of each individual cell corresponding to each data point in each continuous charging segment to obtain the expanded charging segment dataset. The current mutation points of each continuous charging segment in the extended charging segment dataset are identified, and the corresponding continuous charging segments are divided into current stable segments according to the obtained mutation point sets to generate the current stable segment dataset.

4. The lithium battery abnormality early warning method as described in claim 3, characterized in that, The step of performing cell capacity peak feature extraction and analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset to obtain the feature vector of each individual cell in each continuous charging segment includes: Cell differential capacity analysis is performed on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, and the current stable segment dataset is expanded according to the differential capacity value of each data point to obtain a feature analysis dataset. Peak features are extracted from the differential capacity values ​​of the same cell within each current stability segment in the feature analysis dataset to obtain the peak feature set of each cell within the corresponding current stability segment; the peak features in the peak feature set include peak voltage, peak temperature, peak height, peak width, and peak area; The peak feature sets of the same single cell within each of the continuous charging segments are merged to obtain the corresponding segment cell peak feature set. Based on the number of peaks and the peak feature corresponding to the largest peak area in the segment cell peak feature set, the feature vector of the corresponding single cell is obtained.

5. The lithium battery abnormality early warning method as described in claim 4, characterized in that, The step of performing cell differential capacity analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset includes: The voltage values ​​of each individual cell within each data point in each current stabilization segment are subjected to monotonicity processing to obtain the corresponding processed voltage values. Gaussian smoothing is performed on the processed voltage value and the cumulative capacity increment of each individual cell in each data point within each current stabilization segment to obtain the corresponding smoothed voltage value and smoothed capacity value. The gradient approximation is performed based on the smoothed voltage and smoothed capacity values ​​of each individual cell at each data point within each current stabilization segment to obtain the corresponding initial differential capacity. The initial differential capacity is then subjected to amplitude limiting and smoothing processing in sequence to obtain the corresponding differential capacity value.

6. The lithium battery state anomaly early warning method as described in claim 1, characterized in that, The feature vector includes the peak voltage, peak temperature, peak height, and peak width corresponding to the number of peaks and the maximum peak area; The step of performing cluster analysis on the feature vectors of all individual cells within all continuous charging segments to obtain the state clustering results includes: The feature vectors of each individual cell within each continuous charging segment are standardized to obtain the corresponding standardized feature vectors. Obtain the time interval between the current state analysis time and the previous full cluster analysis time, and determine whether the time interval is less than the preset full clustering cycle time. When the time interval is less than the preset full clustering cycle duration, the standardized feature vectors corresponding to the current state analysis time are matched and analyzed with the cluster structure obtained at the previous full clustering analysis time to obtain the state clustering result. When the time interval is equal to the preset full clustering cycle duration, density clustering analysis is performed on all the standardized feature vectors corresponding to the current state analysis time and the full standardized feature vectors corresponding to the previous full clustering analysis time to obtain the state clustering result.

7. The lithium battery abnormality early warning method as described in claim 1, characterized in that, The step of executing the corresponding state anomaly warning based on the state clustering result includes: Based on the clustering cluster to which the standardized feature vector of each individual cell in each continuous charging segment belongs in the state clustering results, an abnormal feature vector set is obtained; Based on the sample point information corresponding to each abnormal feature vector in the abnormal feature vector set, a corresponding abnormal warning output list is generated; the sample point information includes battery ID, number of battery cells, continuous charging segment number, number of cycles, and timestamp. Based on the various abnormal warning output lists, execute the corresponding status abnormal warning.

8. A lithium battery abnormality early warning system, characterized in that, The system includes: The charging segment segmentation module is used to segment the acquired battery runtime sequence data to be analyzed into continuous charging segments to obtain a charging segment dataset; the charging segment dataset includes multiple continuous charging segments. The current stability analysis module is used to divide each continuous charging segment in the charging segment dataset into current stable segments and generate a current stable segment dataset; the current stable segment dataset includes multiple current stable segments. The state feature extraction module is used to perform cell capacity peak feature extraction and analysis on each current stable segment corresponding to each continuous charging segment in the current stable segment dataset, and obtain the feature vector of each individual cell in each continuous charging segment. The anomaly analysis and early warning module is used to perform cluster analysis on the feature vectors of all individual cells in all the continuous charging segments to obtain state clustering results, and execute corresponding state anomaly early warnings based on the state clustering results.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the lithium battery state abnormality early warning method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the lithium battery state abnormality early warning method according to any one of claims 1 to 7.

Images