Method and apparatus for detecting an abnormal battery cell

Abstract

The application discloses a kind of detection method and device of abnormal battery, the method includes: obtaining a plurality of battery to be detected, according to the formation data of each battery First sample set is generated;The first sample set is detected based on isolated forest algorithm, and the outlier score of each data sample in the first sample set is obtained;If the number of target data sample in the first sample set is greater than the preset number, the second sample set consisting of the target data sample is clustered based on the preset clustering algorithm Analysis;According to the clustering analysis result, determine the abnormal battery from each battery;Wherein, the formation data is obtained after the formation treatment to the battery, and the outlier score of target data sample is higher than the preset score, to combine the abnormal detection process and clustering analysis process, improve the detection accuracy of abnormal battery, and then improve the reliability of battery.

Description

Technical Field

[0001] This application relates to the field of battery manufacturing technology, and more specifically, to a method and apparatus for detecting abnormal battery cells. Background Technology

[0002] With the widespread application of lithium-ion batteries, safety incidents characterized by thermal runaway occur frequently. Cell defects introduced during battery manufacturing are one of the potential causes of thermal runaway. For example, manufacturing defects leading to torn cell tabs can cause uneven charging and discharging, resulting in secondary cracks in the cells. The diffusion length of electrons and ions in the solid phase increases with the appearance of cracks, reducing performance and potentially leading to fire. If defective cells can be screened out during battery manufacturing to prevent them from entering the usage stage, the possibility of thermal runaway can be reduced.

[0003] In existing technologies, battery cells are typically photographed using industrial cameras. The captured images are then transmitted to a dedicated image processing system via digital signals. The system determines whether the battery cell is abnormal based on data such as the image's color, pixels, and brightness. However, this method is purely visual inspection, and the accuracy of the inspection is low because the difference between some minor defects and images of normal battery cells is negligible.

[0004] Therefore, improving the detection accuracy of abnormal battery cells is a technical problem that needs to be solved. Summary of the Invention

[0005] This application provides a method and apparatus for detecting abnormal battery cells, in order to improve the accuracy of detecting abnormal battery cells.

[0006] In a first aspect, a method for detecting abnormal battery cells is provided. The method includes: acquiring multiple battery cells to be detected; generating a first sample set based on the formation data of each battery cell; performing anomaly detection on the first sample set based on the isolated forest algorithm to obtain anomaly scores for each data sample in the first sample set; if the number of target data samples in the first sample set is greater than a preset number, performing cluster analysis on a second sample set composed of the target data samples based on a preset clustering algorithm; and determining abnormal battery cells from each battery cell based on the cluster analysis results; wherein the formation data is obtained after forming the battery cells, and the anomaly scores of the target data samples are higher than the preset scores.

[0007] Secondly, a device for detecting abnormal battery cells is provided. The device includes: a generation module for acquiring multiple battery cells to be detected and generating a first sample set based on the formation data of each battery cell; a detection module for performing anomaly detection on the first sample set based on an isolated forest algorithm to obtain anomaly scores for each data sample in the first sample set; a clustering module for performing cluster analysis on a second sample set composed of the target data samples based on a preset clustering algorithm if the number of target data samples in the first sample set is greater than a preset number; and a determination module for determining abnormal battery cells from among the battery cells based on the cluster analysis results; wherein the formation data is obtained after the battery cells have undergone formation processing, and the anomaly scores of the target data samples are higher than preset scores.

[0008] By applying the above technical solution, multiple battery cells to be tested are obtained, and a first sample set is generated based on the formation data of each battery cell. Anomaly detection is performed on the first sample set based on the isolated forest algorithm to obtain the outlier score of each data sample in the first sample set. If the number of target data samples in the first sample set is greater than a preset number, cluster analysis is performed on a second sample set composed of the target data samples based on a preset clustering algorithm. Based on the cluster analysis results, abnormal battery cells are determined from each battery cell. This combines the anomaly detection process and the cluster analysis process, improving the detection accuracy of abnormal battery cells and thus improving the reliability of the battery. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating a method for detecting abnormal battery cells according to an embodiment of the present invention is shown. Figure 2 A flowchart illustrating a method for detecting abnormal battery cells according to another embodiment of the present invention is shown. Figure 3 A schematic diagram of the structure of a detection device for abnormal battery cells proposed in an embodiment of the present invention is shown. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] This application provides a method for detecting abnormal battery cells, such as... Figure 1 As shown, the method includes the following steps: Step S101: Obtain multiple battery cells to be tested, and generate a first sample set based on the formation data of each battery cell.

[0013] Multiple cells to be tested can be from the same production batch, and the cells to be tested can be identified based on their cell serial numbers. Each cell undergoes a formation process to obtain formation data, and a first sample set is produced based on this data. Formation is a crucial step in lithium-ion battery production, specifically an energy conversion process that activates the active materials in the cell. Optionally, if each cell has already undergone formation, the corresponding formation data can be obtained from the cell production MES (Manufacturing Execution System).

[0014] In some embodiments of this application, generating a first sample set based on the formation data of each of the battery cells includes: An initial dataset is generated based on the transformation data described above; The initial dataset is cleaned according to preset data cleaning rules to obtain the first sample set.

[0015] In this embodiment, an initial dataset is first generated based on the various transformation data. Since the initial dataset may contain some abnormal data that affects the detection accuracy, in order to remove these abnormal data and improve detection efficiency, the initial dataset is cleaned according to a preset data cleaning rule to obtain a first sample set, thereby improving detection efficiency and accuracy. Furthermore, those skilled in the art can adopt different preset data cleaning rules according to actual needs, which does not affect the scope of protection of this application.

[0016] In some embodiments of this application, the step of cleaning the initial dataset according to preset data cleaning rules includes: Remove empty data and data that do not meet the preset design indicators from the initial dataset; Data exceeding the upper and lower limits of the preset box plot are considered abnormal noise data and are deleted from the initial dataset. The data in the initial dataset are normalized based on a preset data normalization algorithm.

[0017] In this embodiment, empty data, data that does not meet the preset design indicators, and abnormal noise data will affect the detection accuracy. Therefore, these data are deleted from the initial dataset to improve data processing efficiency and detection accuracy. Then, based on a preset data normalization algorithm, the data in the initial dataset are normalized to improve detection efficiency.

[0018] A box plot is a statistical chart used to display the dispersion of a set of data. A preset box plot is created, and data exceeding the upper and lower limits of the preset box plot are considered outlier noise. The upper limit is the maximum value within the non-outlier range, i.e., upper limit = Q1 + 1.5. IQR; the lower limit is the minimum value within the non-outlier range, i.e., lower limit = Q³ - 1.5 IQR; Q1 (also known as the lower quartile) = the 25th percentile of all values ​​in the sample, arranged in ascending order. Q3 (also known as the upper quartile) = the 75th percentile of all values ​​in the sample, arranged in ascending order. IQR (also known as the interquartile range) = Q3 - Q1.

[0019] Optionally, the preset data normalization algorithm is any one of the maximum-minimum normalization algorithm and the standard normalization algorithm.

[0020] Step S102: Perform anomaly detection on the first sample set based on the isolated forest algorithm to obtain the outlier score of each data sample in the first sample set.

[0021] In this embodiment, the Isolation Forest algorithm is an anomaly detection algorithm that starts from outliers, divides the data according to specified rules, and determines the outlier score based on the number of divisions. Outliers are defined as points that are easily isolated. By performing anomaly detection on the first sample set using the Isolation Forest algorithm, the outlier score of each data sample can be determined. The magnitude of this outlier score indicates whether a data sample is an outlier.

[0022] Step S103: If the number of target data samples in the first sample set is greater than a preset number, perform cluster analysis on the second sample set composed of the target data samples based on a preset clustering algorithm.

[0023] In this embodiment, after determining the outlier score of each data sample, each outlier score is compared with a preset score. If the outlier score is greater than the preset score, the corresponding data sample is the target data sample. Then, it is determined whether the number of target data samples is greater than a preset number. If it is, the target data samples are grouped into a second sample set, and cluster analysis is performed on the second sample set based on a preset clustering algorithm to further filter the second sample set. Optionally, the preset clustering algorithm can be any of the following algorithms: DBSCAN (Density-Based Spatial Clustering of Applications with Noise), OPTICS (Ordering Points to identify the clustering structure), DPC (clustering by fastsearch and find of density peaks).

[0024] Step S104: Identify abnormal cells from each of the cells based on the cluster analysis results.

[0025] After performing cluster analysis, the cluster analysis results are obtained, and the abnormal cells in each cell are identified from the cluster analysis results.

[0026] In some embodiments of this application, after obtaining the outlier scores of each data sample in the first sample set, the method further includes: If the number of target data samples is not greater than the preset number, it is determined that there are no abnormal cells among the cells.

[0027] In this embodiment, if the number of target data samples is not greater than the preset number, it means that the number of abnormal points in the first sample set is too small to determine the existence of abnormal cells. Therefore, it is determined that there are no abnormal cells in each cell, thereby ensuring the reliability of the detection process.

[0028] In some embodiments of this application, the abnormal cell is a cell with a tab defect, and the formation data includes the cell number, K value, the starting temperature of the first pre-charge stage, the starting temperature of the second pre-charge stage, and the capacity change of the second pre-charge stage.

[0029] In this embodiment, the abnormal cell is a cell with tab defects, such as tab tearing or poor tab soldering. To accurately detect whether a cell has tab defects, the formation data needs to be correlated with the tab defects; that is, data from the formation data that is significantly affected by tab defects should be selected. Therefore, the formation data includes the cell number, K-value, the starting temperature of the first pre-charge stage, the starting temperature of the second pre-charge stage, and the capacity change during the second pre-charge stage, thereby improving the detection accuracy.

[0030] K-value represents the voltage drop of the battery cell per unit time and is an indicator used to measure the self-discharge rate of lithium batteries. The first and second pre-charge stages are part of the cell processing in the formation process.

[0031] Those skilled in the art may add or remove the types of formation data according to actual needs, and different formation data do not affect the scope of protection of this application.

[0032] By applying the above technical solution, multiple battery cells to be tested are obtained, and a first sample set is generated based on the formation data of each battery cell. Anomaly detection is performed on the first sample set based on the isolated forest algorithm to obtain the outlier score of each data sample in the first sample set. If the number of target data samples in the first sample set is greater than a preset number, cluster analysis is performed on a second sample set composed of the target data samples based on a preset clustering algorithm. Abnormal battery cells are determined from each battery cell based on the cluster analysis results. The formation data is obtained after the battery cells have undergone formation processing, and the outlier score of the target data sample is higher than the preset score. This combines the anomaly detection process and the cluster analysis process, improving the detection accuracy of abnormal battery cells and thus improving the reliability of the battery.

[0033] To further illustrate the technical concept of this invention, the technical solution of this invention will now be described in conjunction with specific application scenarios.

[0034] This application provides a method for detecting abnormal battery cells, used to detect whether the battery cell has tab defects, such as... Figure 2 As shown, it includes the following steps: Step S201: Obtain multiple cells to be tested, and generate an initial dataset based on the formation data of each cell.

[0035] The cells to be tested are from the same batch. The formation data of each cell is obtained from the cell production MES. The formation data includes: cell number N, K value, starting temperature T1 of the first pre-charge stage, starting temperature T2 of the second pre-charge stage, and capacity change Q of the second pre-charge stage, as shown in Table 1.

[0036] Table 1

[0037] Initial dataset ,in, .

[0038] Step S202: Clean the initial dataset according to the preset data cleaning rules to obtain the first sample set.

[0039] Specifically, empty data and data that do not meet the preset design indicators are deleted from the initial dataset; data exceeding the upper and lower limits of the preset box plot are considered abnormal noise data and are deleted from the initial dataset; the data in the initial dataset are normalized based on the maximum-minimum normalization algorithm to obtain the first sample set. .

[0040] The preset design parameters include 0.001≤K≤0.23mV / h, 20℃≤T1≤45℃, and 20℃≤T2≤45℃. Furthermore, the specific process of the maximum-minimum normalization algorithm is existing technology and will not be elaborated here.

[0041] Step S203: Perform anomaly detection on the first sample set based on the isolated forest algorithm to obtain the outlier score of each data sample in the first sample set.

[0042] Configure the relevant parameters for the Isolation Forest algorithm: =80%; ; =4.

[0043] in, X represents the number of samples drawn from X to train each base estimator. Contamination is the proportion of outliers in the dataset, used to define the threshold for the decision function during fitting. The number of attributes extracted from X to train each tree.

[0044] Specifically, step S203 includes the following steps: Step S2031, Given a dataset, Where n represents the number of data samples collected, and d represents the number of features in each data sample. Randomly selected from Each sample point constitutes subset of Place it in the root node.

[0045] Step 2032: From Randomly select one dimension from the dimensions. Randomly generate a cutting point in the current data. ,in, .

[0046] Step 2033: This cutting point A hyperplane is generated that divides the current data space into two subspaces: one with a specified dimension less than 1. The sample point is placed in the left child node, which is greater than or equal to Place it into the right child node.

[0047] Step 2034: Recursively repeat steps 2032 and 2033 until all leaf nodes have only one sample point or the isolated tree has reached the specified height.

[0048] Step 2035: Repeat steps 2031 to 2034 until a specified number of isolated trees are generated.

[0049] Step 2036: For each data point Let it traverse every isolated tree and calculate the points. Average height in the forest The average height of all points is normalized. Outlier score. The calculation formula is as follows:

[0050] in:

[0051] E(h(x)) is the average value of h(x) in the iTree set. , For harmonic numbers, is Euler's constant, approximately 0.5772156649; n is the number of leaf nodes; c(n) is the average value of h(x) given n, used to standardize h(x).

[0052] It's about recording. Outlier scores in an isolated tree constructed from n training samples. The value range is [0,1]. Abnormal situations are judged in the following ways: 1) The closer the value is to 1, the higher the probability that it is an outlier; 2) The closer to 0, the higher the probability that it is a normal point; 3) If most of the training samples are The values ​​are all close to 0.5, indicating that there are no obvious outliers in the entire dataset.

[0053] Step S204: Is the number of target data samples greater than the preset number? If yes, proceed to step S205; otherwise, proceed to step S206.

[0054] After determining the outlier score for each data sample, each outlier score is compared with a preset score. If the outlier score is greater than the preset score, the corresponding data sample is designated as the target data sample. Then, it is determined whether the number of target data samples is greater than a preset number. If it is, all target data samples are combined into a second sample set. If not, proceed to step S205; otherwise, proceed to step S206.

[0055] Step S205: Based on the DBSCAN algorithm, cluster analysis is performed on the second sample set composed of target data samples to identify abnormal battery cells.

[0056] Configure the relevant parameters for the DBSCAN algorithm: ; Euclidean distance was chosen as the distance metric. It describes the neighborhood distance threshold of a given sample. The distance to a certain sample is described as The threshold for the number of samples in the neighborhood.

[0057] The specific process of cluster analysis includes: Step S2051: Initialize the core object collection Initialize the number of clusters k=0, and initialize the set of unvisited samples. Cluster partitioning ; Step 2052: For To find all core objects, follow these steps: a) Find the sample using Euclidean distance of - Neighborhood Subsample Set ; b) If the number of samples in the subset satisfies , will sample Add to the core object sample collection: ; Step 2053: If the core object collection If the condition is met, the process ends; otherwise, proceed to step 2054.

[0058] Step 2054: In the core object collection In the middle, randomly select a core object. Initialize the current cluster core object queue Initialize the category index k = k + 1, and initialize the current cluster sample set. Update the set of unvisited samples. ; Step 2055: If the current cluster core object queue Then the current cluster Once generated, update the cluster partitions. Update the core object collection Proceed to step 2053. Otherwise, update the core object collection. ; Step 2056: In the current cluster core object queue Extract a core object from By neighborhood distance threshold Find all - Neighborhood Subsample Set ,make Update the current cluster sample set. Update the set of unvisited samples. , renew Proceed to step 2055.

[0059] The output result is: cluster partitioning This refers to an abnormal battery cell with a tab defect.

[0060] In step 2052, other distance metrics may also be used, such as Manhattan distance, Chebyshev distance, Minkowski distance, weighted Minkowski distance, standardized Euclidean distance, Mahalanobis distance, etc.

[0061] Step S206: Determine that there are no abnormal cells among the battery cells.

[0062] If the number of target data samples is not greater than the preset number, it means that the number of abnormal points in the first sample set is too small to determine the existence of abnormal cells. Therefore, it is determined that there are no abnormal cells in the battery cells, thus ensuring the reliability of the detection process.

[0063] By applying the above technical solutions and verifying with actual data, the accuracy is 99.01%. This combines the anomaly detection process with the cluster analysis process, improving the accuracy of detecting abnormal cells and thus improving the reliability of the battery.

[0064] This application also proposes a device for detecting abnormal battery cells, such as... Figure 3As shown, the device includes: a generation module 301, used to acquire multiple battery cells to be tested and generate a first sample set based on the formation data of each battery cell; a detection module 302, used to perform anomaly detection on the first sample set based on the isolated forest algorithm to obtain the outlier score of each data sample in the first sample set; a clustering module 303, used to perform cluster analysis on a second sample set composed of the target data samples based on a preset clustering algorithm if the number of target data samples in the first sample set is greater than a preset number; and a determination module 304, used to determine the abnormal battery cells from each battery cell based on the cluster analysis results; wherein, the formation data is obtained after the battery cells have undergone formation processing, and the outlier score of the target data sample is higher than a preset score.

[0065] In specific application scenarios, the determining module 304 is further configured to: if the number of the target data samples is not greater than the preset number, determine that there is no abnormal cell among the cells.

[0066] In a specific application scenario, the generation module 301 is specifically used to: generate an initial dataset based on the transformed data; and perform data cleaning on the initial dataset according to a preset data cleaning rule to obtain the first sample set.

[0067] In specific application scenarios, the generation module 301 is also specifically used for: deleting empty data and data that do not conform to preset design indicators from the initial dataset; deleting data that exceeds the upper and lower limits of the preset box plot as abnormal noise data from the initial dataset; and normalizing each data in the initial dataset based on a preset data normalization algorithm.

[0068] In specific application scenarios, the abnormal cell is a cell with a tab defect, and the formation data includes the cell number, K value, the starting temperature of the first pre-charge stage, the starting temperature of the second pre-charge stage, and the capacity change of the second pre-charge stage.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting abnormal battery cells, characterized in that, The method includes: Multiple battery cells to be tested are acquired, and a first sample set is generated based on the formation data of each battery cell. Anomaly detection is performed on the first sample set based on the isolated forest algorithm to obtain the outlier score of each data sample in the first sample set, including: step S2031, Given a dataset, Where n represents the number of data samples collected, and d represents the number of features in each data sample. Randomly selected from Each sample point constitutes subset of Insert into the root node; Step 2032: From Randomly select one dimension from the dimensions. Randomly generate a cutting point in the current data. ,in, Step 2033: The cutting point A hyperplane is generated that divides the current data space into two subspaces: one with a specified dimension less than 1. The sample point is placed in the left child node, which is greater than or equal to Add the right child node; Step 2034: Recursively call steps 2032 and 2033 until all leaf nodes have only one sample point or the isolated tree has reached the specified height; Step 2035: Loop through steps 2031 to 2034 until the specified number of isolated trees are generated; Step 2036: For each data point Let it traverse every isolated tree and calculate the points. Average height in the forest The average height of all points is normalized to obtain the outlier score; If the outlier score is greater than the preset score, the corresponding data sample is the target data sample; if the number of target data samples in the first sample set is greater than the preset number, cluster analysis is performed on the second sample set composed of the target data samples based on the preset clustering algorithm. The specific process of cluster analysis includes: step S2051, initializing the core object set. Initialize the number of clusters k=0, and initialize the set of unvisited samples. Cluster partitioning Step 2052: For To find all core objects, follow these steps: a) Find the samples using Euclidean distance. of - Neighborhood Subsample Set b) If the number of samples in the subset satisfies , sample Add to the core object sample collection: , MinPts The distance to a certain sample is described as ε The threshold for the number of samples in the neighborhood; Step 2053: If the core object set If yes, then the process ends; otherwise, proceed to step 2054. Step 2054: In the core object collection... In the middle, randomly select a core object. Initialize the current cluster core object queue Initialize the category index k = k + 1, and initialize the current cluster sample set. Update the unvisited sample set Step 2055: If the current cluster core object queue Then the current cluster Once generated, update the cluster partitions. Update the core object collection Proceed to step 2053; otherwise, update the core object set. Step 2056: In the current cluster core object queue Extract a core object from Through neighborhood distance threshold Find all - Neighborhood Subsample Set ,make Update the current cluster sample set Update the unvisited sample set ,renew Proceed to step 2055; the output is: cluster partitioning. This refers to an abnormal battery cell with a tab defect; Abnormal battery cells were identified from each of the battery cells based on the cluster analysis results; The formation data is obtained after the formation process of the battery cell, and the outlier score of the target data sample is higher than the preset score.

2. The method as described in claim 1, characterized in that, After obtaining the outlier scores of each data sample in the first sample set, the method further includes: If the number of target data samples is not greater than the preset number, it is determined that there are no abnormal cells among the cells.

3. The method as described in claim 1, characterized in that, The step of generating a first sample set based on the formation data of each of the battery cells includes: An initial dataset is generated based on the transformation data described above; The initial dataset is cleaned according to preset data cleaning rules to obtain the first sample set.

4. The method as described in claim 3, characterized in that, The step of cleaning the initial dataset according to preset data cleaning rules includes: Remove empty data and data that do not meet the preset design indicators from the initial dataset; Data exceeding the upper and lower limits of the preset box plot are considered abnormal noise data and are deleted from the initial dataset. The data in the initial dataset are normalized based on a preset data normalization algorithm.

5. The method as described in claim 1, characterized in that, The abnormal cell is a cell with a tab defect. The formation data includes the cell number, K value, the starting temperature of the first pre-charge stage, the starting temperature of the second pre-charge stage, and the capacity change during the second pre-charge stage.

6. A device for detecting abnormal battery cells, characterized in that, The device includes: The generation module is used to acquire multiple cells to be tested and generate a first sample set based on the formation data of each cell; The detection module is used to perform anomaly detection on the first sample set based on the isolated forest algorithm, and obtain the outlier score of each data sample in the first sample set, including step S2031. Given a dataset, Where n represents the number of data samples collected, and d represents the number of features in each data sample. Randomly selected from Each sample point constitutes subset of Insert into the root node; Step 2032: From Randomly select one dimension from the dimensions. Randomly generate a cutting point in the current data. ,in, Step 2033: The cutting point A hyperplane is generated that divides the current data space into two subspaces: one with a specified dimension less than 1. The sample point is placed in the left child node, which is greater than or equal to Add the right child node; Step 2034: Recursively call steps 2032 and 2033 until all leaf nodes have only one sample point or the isolated tree has reached the specified height; Step 2035: Loop through steps 2031 to 2034 until the specified number of isolated trees are generated; Step 2036: For each data point Let it traverse every isolated tree and calculate the points. Average height in the forest The average height of all points is normalized to obtain the outlier score; The clustering module is used to identify target data samples if the outlier score is greater than a preset score; if the number of target data samples in the first sample set is greater than a preset number, clustering analysis is performed on a second sample set composed of the target data samples based on a preset clustering algorithm. The specific process of clustering analysis includes: Step S2051, initializing the core object set. Initialize the number of clusters k=0, and initialize the set of unvisited samples. Cluster partitioning Step 2052: For To find all core objects, follow these steps: a) Find the samples using Euclidean distance. of - Neighborhood Subsample Set b) If the number of samples in the subset satisfies , sample Add to the core object sample collection: , MinPts The distance to a certain sample is described as ε The threshold for the number of samples in the neighborhood; Step 2053: If the core object set If yes, then the process ends; otherwise, proceed to step 2054. Step 2054: In the core object collection... In the middle, randomly select a core object. Initialize the current cluster core object queue Initialize the category index k = k + 1, and initialize the current cluster sample set. Update the unvisited sample set Step 2055: If the current cluster core object queue Then the current cluster Once generated, update the cluster partitions. Update the core object collection Proceed to step 2053; otherwise, update the core object set. Step 2056: In the current cluster core object queue Extract a core object from Through neighborhood distance threshold Find all - Neighborhood Subsample Set ,make Update the current cluster sample set Update the unvisited sample set ,renew Proceed to step 2055; the output is: cluster partitioning. This refers to an abnormal battery cell with a tab defect; A determination module is used to identify abnormal cells from each of the cells based on cluster analysis results; The formation data is obtained after the formation process of the battery cell, and the outlier score of the target data sample is higher than the preset score.

7. The apparatus as claimed in claim 6, characterized in that, The determining module is further configured to: If the number of target data samples is not greater than the preset number, it is determined that there are no abnormal cells among the cells.

8. The apparatus as claimed in claim 6, characterized in that, The generation module is specifically used for: An initial dataset is generated based on the transformation data described above; The initial dataset is cleaned according to preset data cleaning rules to obtain the first sample set.

9. The apparatus as claimed in claim 8, characterized in that, The generation module is also specifically used for: Remove empty data and data that do not meet the preset design indicators from the initial dataset; Data exceeding the upper and lower limits of the preset box plot are considered abnormal noise data and are deleted from the initial dataset. The data in the initial dataset are normalized based on a preset data normalization algorithm.

10. The apparatus as claimed in claim 6, characterized in that, The abnormal cell is a cell with a tab defect. The formation data includes the cell number, K value, the starting temperature of the first pre-charge stage, the starting temperature of the second pre-charge stage, and the capacity change during the second pre-charge stage.

Images