Battery cell anomaly detection method, device, apparatus, and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for detecting abnormal battery cells. The method involves acquiring individual cell voltage data of each cell in a battery pack over a continuous time period; determining a parameter sequence corresponding to each cell based on the individual cell voltage data over the continuous time period; the parameter sequence characterizing the changing trend of the individual cell voltage data of the corresponding cell over the continuous time period; performing multidimensional curve feature extraction processing on the parameter sequence of each cell to determine a multidimensional curve feature corresponding to each cell; the multidimensional curve feature reflecting the changing trend of the parameter sequence of the corresponding cell over the continuous time period; and detecting whether any abnormality exists in each cell based on the multidimensional curve feature corresponding to each cell. Using the technical solution of this application can improve the detection capability of battery abnormalities.

Description

Technical Field

[0001] This application relates to the field of battery management technology, and in particular to a method, apparatus, device and storage medium for detecting abnormalities in battery cells. Background Technology

[0002] A power battery is a power supply device that provides power to electric vehicles such as electric cars and electric trains. Each electric vehicle typically has a power battery pack, which includes multiple individual battery cells. During the use of electric vehicles, abnormalities inevitably occur in the power battery pack, making it necessary to inspect the battery cells within the power battery pack for any abnormalities.

[0003] In related technologies, when detecting abnormalities in battery cells, a fixed upper or lower voltage limit is usually set. When the voltage of a single battery cell exceeds the upper or lower voltage limit, it is judged as abnormal.

[0004] However, the upper or lower voltage limits set in the above technologies are difficult to set scientifically. If they are too high, they will lead to missed detections, and if they are too low, they will lead to frequent false alarms, resulting in insufficient detection capability for battery abnormalities. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for detecting abnormal battery cells, which addresses the shortcomings of existing technologies where thresholds are difficult to set scientifically, resulting in insufficient detection capabilities for battery abnormalities. It enables the detection of abnormal battery cells by introducing DI features and multidimensional curve analysis over a long time scale, amplifying the subtle abnormal change trends of battery voltage signals, and detecting abnormal battery cells in thermal runaway electrical devices in advance, thereby improving the detection capability of abnormal battery cells.

[0006] This application provides a method for detecting abnormalities in battery cells, including: Acquire the individual cell voltage data of each cell in the battery pack over a continuous time period; Based on the individual cell voltage data of each cell over a continuous time period, the parameter sequence corresponding to each cell is determined; the above parameter sequence is used to characterize the changing trend of the individual cell voltage data of the corresponding cell over a continuous time period. Multidimensional curve feature extraction is performed on the parameter sequence of each cell to determine the multidimensional curve features corresponding to each cell; the multidimensional curve features are used to reflect the changing trend of the parameter sequence of the corresponding cell in a continuous time period. Based on the multidimensional curve characteristics of each cell, detect whether there are any abnormalities in each cell.

[0007] According to the cell anomaly detection method provided in this application, the above-mentioned method for determining the parameter sequence corresponding to each cell based on the individual cell voltage data within a continuous time period includes: Based on the individual cell voltage data and at least one preset target order combination, determine the parameter sequence of each cell under each target order combination; the aforementioned target order combination is used to amplify the degree of deviation of the individual cell voltage data. Accordingly, the above-mentioned multidimensional curve feature extraction process is performed on the parameter sequence of each cell to determine the multidimensional curve features corresponding to each cell, including: For each target order combination, multidimensional curve feature extraction is performed on the parameter sequence of each cell under the target order combination to determine the multidimensional curve features corresponding to each cell under the target order combination.

[0008] According to the cell anomaly detection method provided in this application, the above-mentioned detection of whether each cell has an anomaly based on the multidimensional curve characteristics corresponding to each cell includes: For each target order combination, the BIRCH algorithm is used to cluster the multidimensional curve features of all cells under the target order combination to obtain multiple clusters under the target order combination; each cluster includes at least one cell and the multidimensional curve features of the cell. Based on the multidimensional curve characteristics of the cells in each cluster under the target order combination, and the cluster-related information corresponding to each cluster, calculate the anomaly score of each cell under the target order combination; Based on the abnormal score of each cell under the target order combination, detect whether there are any abnormalities in each cell.

[0009] According to the cell anomaly detection method provided in this application, the aforementioned cluster-related information includes cluster size, cluster density, and cluster center point. The method calculates the anomaly score for each cell under the target order combination based on the multidimensional curve characteristics of the cells in each cluster under the target order combination and the cluster-related information corresponding to each cluster, including: For any first cell among all cells, calculate the distance between the multidimensional curve feature of the first cell under the target order combination and the origin, and obtain the origin distance score. Calculate the cluster size factor based on the cluster size of the cluster to which the first cell belongs; The cluster density factor is calculated based on the cluster density of the cluster to which the first cell belongs and the maximum cluster density among all clusters. Calculate the distance between the multidimensional curve features of the first battery cell and the cluster center points of other clusters to obtain the distance scores of other clusters; the other clusters mentioned above are the clusters other than the cluster to which the first battery cell belongs. Calculate the anomaly score of the first cell under the target order combination based on at least one of the origin distance score, cluster size factor, cluster density factor, and other cluster distance scores corresponding to the first cell under the target order combination.

[0010] According to the cell anomaly detection method provided in this application, the above-mentioned detection of whether each cell has an anomaly based on the anomaly score of each cell under the target order combination includes: For each cell, obtain the cell's anomaly score under all target order combinations, and determine the anomaly degree index corresponding to the cell based on each anomaly score; the higher the anomaly degree index, the higher the anomaly degree of the corresponding cell. Based on the abnormality index corresponding to each cell, detect whether there are any abnormalities in each cell.

[0011] According to the cell anomaly detection method provided in this application, the above-mentioned determination of the anomaly degree index corresponding to the cell based on the anomaly scores of the cell includes: Identify at least one candidate abnormal score that is greater than the abnormal score threshold from all abnormal scores of the battery cell. The anomaly index corresponding to the cell is determined based on the combination of target orders corresponding to the candidate anomaly scores of the cell.

[0012] According to the cell anomaly detection method provided in this application, the method further includes: The abnormal scores of each cell are sorted and visualized. And / or, sort the anomaly index of each cell and output the visualization.

[0013] According to the cell anomaly detection method provided in this application, the determination of the above-mentioned target order combination includes: Acquire sample data of multiple candidate order combinations and multiple thermal runaway electrical devices; the sample data includes the sample cell voltage data of each sample cell in the sample battery pack of the thermal runaway electrical device within a continuous time period, and each sample cell includes cells with abnormalities. For each candidate order combination, a sensitivity score is calculated based on the sample data. The sensitivity score is used to characterize the degree to which the candidate order combination distinguishes between normal voltage data and abnormal voltage data. Based on the sensitivity scores of each candidate order combination, the target order combination is selected from the candidate order combinations.

[0014] According to the cell anomaly detection method provided in this application, the above-mentioned method for determining the parameter sequence corresponding to each cell based on the individual cell voltage data within a continuous time period includes: Get the set sliding window size and the set sliding step size; Based on the sliding window size and sliding step size, the individual cell voltage data of all cells in a continuous time period are divided on the time axis to obtain multiple time windows; each time window includes a portion of the individual cell voltage data of all cells. Based on a subset of individual cell voltage data from all cells within each time window, determine the parameters of each cell in each time window; Based on the parameters of each cell across the entire time window, determine the parameter sequence corresponding to each cell.

[0015] This application also provides a battery cell anomaly detection device, including the following modules: The acquisition module is used to acquire the individual cell voltage data of each cell in the battery pack over a continuous time period. The parameter determination module is used to determine the parameter sequence corresponding to each battery cell based on the individual cell voltage data of each cell within a continuous time period; the above parameter sequence is used to characterize the changing trend of the individual cell voltage data of the corresponding battery cell within a continuous time period. The feature extraction module is used to perform multidimensional curve feature extraction processing on the parameter sequence of each battery cell to determine the multidimensional curve features corresponding to each battery cell; the multidimensional curve features are used to reflect the changing trend of the parameter sequence of the corresponding battery cell in a continuous time period. The cell anomaly detection module is used to detect whether there are any anomalies in each cell based on the multi-dimensional curve characteristics corresponding to each cell.

[0016] This application also provides an electronic 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 any of the above-described cell anomaly detection methods.

[0017] This application also provides an electrical device, including a battery pack and electronic equipment.

[0018] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the cell anomaly detection method as described above.

[0019] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the cell anomaly detection method as described above.

[0020] The cell anomaly detection method, apparatus, device, and storage medium provided in this application acquire individual cell voltage data of each cell in a battery pack over a continuous time period. Based on the individual cell voltage data, a parameter sequence corresponding to each cell is determined. Multidimensional curve feature extraction processing is performed on the parameter sequence of each cell to determine the corresponding multidimensional curve features. Then, the presence of anomalies in each cell is detected based on these multidimensional curve features. The parameter sequence characterizes the changing trend of the individual cell voltage data over a continuous time period, and the multidimensional curve features reflect the changing trend of the parameter sequence over a continuous time period. This method detects cell anomalies by introducing sequence features characterizing the changing trend of individual cell voltage data and long-term multidimensional curve analysis features. This amplifies minute abnormal changes in battery voltage signals, enabling early detection of abnormal cells in thermal runaway devices and improving the detection capability of abnormal cells. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the cell anomaly detection method provided in this application.

[0023] Figure 2 This is a schematic diagram of the battery cell anomaly detection device provided in this application.

[0024] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions 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, 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.

[0026] In the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone, where A and B can be singular or plural. In the textual description of this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0027] Currently, the detection of abnormalities in power batteries mainly relies on the direct monitoring of the voltage of individual cells, and mainly includes the following cell abnormality detection methods: 1. Threshold comparison method: Set a fixed upper or lower voltage limit. When the voltage of a single cell exceeds the threshold, it is judged as abnormal. The disadvantages of this method are: the threshold is difficult to set scientifically. Too high a threshold will lead to missed alarms, while too low a threshold will lead to frequent false alarms. Normal voltage fluctuations under actual vehicle operating conditions can easily trigger false alarms, and it cannot identify small, gradual anomalies within the threshold.

[0028] 2. Simple Statistical Method: This method calculates the mean and standard deviation of the voltage of all battery cells at the same time point. Cells whose voltage deviation from the mean exceeds N (greater than 0) times the standard deviation are considered abnormal. The disadvantages of this method are: it is overly sensitive to instantaneous voltage fluctuations, resulting in a high false alarm rate under dynamic conditions such as rapid acceleration and deceleration of the vehicle. Furthermore, it analyzes only a single time segment, ignoring the long-term evolution trend of voltage and failing to capture early abnormal characteristics.

[0029] 3. Single-feature clustering method: This method extracts single or a small number of statistical features, such as mean voltage and standard deviation, for clustering to achieve anomaly detection. The disadvantages of this method are: low feature dimensionality, insufficient information, and difficulty in comprehensively depicting the complex state of the battery cell. Normal battery cells and cells in the early stages of anomalies are not easily distinguishable in a single feature space, leading to poor clustering results and a high false negative rate.

[0030] Based on this, the embodiments of this application provide a method, apparatus, device and storage medium for detecting abnormalities in battery cells, which can solve the above-mentioned technical problems. Specifically, it can effectively utilize the multi-dimensional dynamic evolution information of voltage signals over a long time scale, improve the sensitivity to minute abnormal signals and the robustness to complex working conditions in real vehicles, and achieve higher accuracy and stronger detection capabilities.

[0031] It should be noted that the execution subject in the embodiments of this application can be a battery cell anomaly detection device, an electronic device, or other devices or equipment, etc., and is not specifically limited here. The electronic device here can be a terminal or a server; the following embodiments will use an electronic device as the execution subject for illustration.

[0032] The cell anomaly detection method provided in this application will be described in detail below through several specific embodiments. It is understood that these specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0033] Example 1 Figure 1 This is a flowchart illustrating the cell anomaly detection method provided in this application, as shown below. Figure 1 As shown, the method includes the following steps: Step 102: Obtain the individual cell voltage data of each cell in the battery pack over a continuous time period.

[0034] Here, the battery pack can be a power battery pack in an electrical device that requires anomaly detection. The electrical device can be, for example, a vehicle, a robot, or a person / cargo transport device. The battery pack includes multiple cells, and the specific number and arrangement of the cells can be set according to the actual situation.

[0035] During battery pack operation, individual cell voltage data can be collected over a continuous time period. This data is then cleaned, and the voltage columns of all cells are identified to obtain the individual cell voltage data for each cell within that continuous time period. Each cell's individual voltage data includes the voltage collected at various time points within that continuous time period. This continuous time period can include multiple consecutive time points, and its duration can be set according to actual conditions; there is no limitation here. Furthermore, this continuous time period can include the time before or at the time of battery failure in the electrical device. The individual cell voltage data within this continuous time period can also include charging and discharging data, thus enabling the collection of individual cell voltage data under various operating conditions for long-term and multi-condition analysis.

[0036] Step 104: Determine the parameter sequence for each cell based on the individual cell voltage data over a continuous time period; the parameter sequence is used to characterize the changing trend of the individual cell voltage data over a continuous time period.

[0037] In this step, the parameter sequence of the battery cell can be a sequence of DI values ​​calculated from the cell's individual cell voltage data, or it can be a parameter sequence calculated from the cell's individual cell voltage data combined with other data. This parameter sequence of the battery cell can reflect the changing trend of the cell's individual cell voltage data over a continuous time period, and more specifically, it can reflect the changing trend of the deviation of the cell's individual cell voltage data from the average voltage of all cells in the battery pack over a continuous time period.

[0038] Taking the DI value sequence of a battery cell as an example, DI stands for Dimensionless Indicator, a diagnostic feature sensitive to battery anomalies and faults. Calculated using a specific formula, it can sensitively capture and amplify minute trends in the battery's voltage signal, such as amplifying weak abnormal signals while maintaining good stability to eliminate interference. Based on this, this embodiment calculates the DI value sequence of each battery cell using its individual cell voltage data over a continuous time period. The DI value sequence of each cell reflects the minute trends in the individual cell voltage data over time, which can be used for cell anomaly detection.

[0039] Specifically, after obtaining the individual voltage data of each battery cell within a continuous time period, for each cell, its individual voltage data can be divided into multiple voltage data segments. A DI value can be calculated from each voltage data segment, and the calculated DI values ​​can be arranged in order to obtain the DI value sequence corresponding to that cell. Alternatively, for all battery cells, the individual voltage data of all cells can be divided into multiple voltage data segments, and a DI value can be calculated from each voltage data segment. The DI values ​​of each cell can then be arranged in order to obtain the DI value sequence corresponding to that cell.

[0040] Step 106: Perform multidimensional curve feature extraction processing on the parameter sequence of each cell to determine the multidimensional curve features corresponding to each cell; the multidimensional curve features are used to reflect the changing trend of the parameter sequence of the corresponding cell in a continuous time period.

[0041] In this step, we continue to use the parameter sequence of a battery cell, which is its DI value sequence, as an example. After obtaining the DI value sequence of each battery cell, six types of curve features can be extracted for each battery cell's DI value sequence to form the multidimensional curve features of that battery cell. The six types of curve features extracted here include: discrete features, linear trend features, curvature features, rate of change features, local trend features, and peak-valley features.

[0042] For the discrete characteristics corresponding to the DI value sequence of each cell, it can include the standard deviation corresponding to the DI value sequence, which is used to reflect the overall fluctuation of the DI value of the cell. Typically, cells with abnormalities (referred to as abnormal cells) have greater fluctuations.

[0043] For the linear trend characteristics corresponding to the DI value sequence of each cell, it can include the slope corresponding to the DI value sequence. Specifically, the slope can be obtained by linearly fitting the DI value sequence, which is used to reflect the direction of long-term changes in DI value.

[0044] For the curvature characteristics corresponding to the DI value sequence of each cell, it can include the mean of the second derivative corresponding to the DI value sequence. Specifically, the mean of the second derivative can be obtained by solving the second derivative and its mean of the DI value sequence, which is used to measure the degree of nonlinear curvature of the curve corresponding to the DI value sequence.

[0045] For the rate of change characteristic of the DI value sequence corresponding to each cell, it can include the mean of the first difference corresponding to the DI value sequence. Specifically, the mean of the first difference can be obtained by solving the first difference and its mean of the DI value sequence, which is used to reflect the instantaneous rate of change of the DI value.

[0046] For the local trend characteristics corresponding to the DI value sequence of each cell, it can include the slope variance corresponding to the DI value sequence. Specifically, the slope variance can be obtained by segmenting the DI value sequence and calculating the slope of each segment, as well as calculating the variance of each slope, which can be used to identify abnormal patterns in local areas.

[0047] For the peak and valley characteristics corresponding to the DI value sequence of each cell, it can include the number of peaks and valleys corresponding to the DI value sequence. Specifically, the number of peaks and valleys in the DI value sequence can be calculated by a peak finding algorithm to measure the oscillation characteristics of the DI value sequence.

[0048] The above method can obtain the multi-dimensional curve features corresponding to the individual voltage data of each battery cell, which can reflect the small change trend of the parameter sequence of the battery cell in a continuous time period, so that it can distinguish between the real abnormality and normal fluctuation of the battery cell and achieve accurate detection of abnormal battery cells.

[0049] Step 108: Detect whether there are any abnormalities in each battery cell based on the multidimensional curve characteristics corresponding to each cell.

[0050] In this step, after obtaining the multidimensional curve characteristics of each cell, a score or index that characterizes the degree of abnormality of each cell can be calculated using these characteristics. The higher the score or index, the higher the degree of abnormality of the corresponding cell. Then, based on the score or index of each cell, it is determined whether each cell is abnormal. For example, if the abnormality score or index of a certain cell exceeds a threshold, then the cell is determined to be abnormal. Alternatively, the cell with the highest score or index is identified as the most abnormal cell.

[0051] In this embodiment, by acquiring the individual cell voltage data of each cell in the battery pack over a continuous time period, the parameter sequence corresponding to each cell is determined based on the individual cell voltage data. Multidimensional curve feature extraction processing is then performed on the parameter sequence of each cell to determine the corresponding multidimensional curve features. Finally, the multidimensional curve features of each cell are used to detect whether any cell exhibits anomalies. The parameter sequence characterizes the changing trend of the individual cell voltage data over a continuous time period, while the multidimensional curve features reflect the changing trend of the parameter sequence over a continuous time period. This method detects cell anomalies by introducing sequence features representing the changing trend of individual cell voltage data and long-term multidimensional curve analysis features. This amplifies minute abnormal changes in the battery voltage signal, enabling early detection of abnormal cells in thermal runaway devices and improving the detection capability of abnormal cells.

[0052] Example 2 For example, in step 104 above, the parameter sequence corresponding to each cell is determined based on the individual cell voltage data over a continuous time period, including: Get the set sliding window size and the set sliding step size; Based on the sliding window size and sliding step size, the individual cell voltage data of all cells in a continuous time period are divided on the time axis to obtain multiple time windows; each time window includes a portion of the individual cell voltage data of all cells. Based on a subset of individual cell voltage data from all cells within each time window, determine the parameters of each cell in each time window; Based on the parameters of each cell across the entire time window, determine the parameter sequence corresponding to each cell.

[0053] Specifically, the required sliding window size (window_size) and sliding step size (step_size) can be pre-defined according to actual conditions. The individual cell voltage data of all battery cells are arranged in time sequence. Then, using the aforementioned sliding window, the window is slid along the time axis with the sliding step size, thereby dividing the individual cell voltage data of all battery cells into multiple time windows. Each time window includes a portion of the individual cell voltage data of all battery cells within the corresponding sliding window size. The individual cell voltage data within each time window may or may not overlap.

[0054] Continuing with the example of a battery cell's parameter sequence being its DI value sequence, after obtaining each time window, we can obtain a portion of the individual cell voltage data for all cells within each time window. Then, using the DI value calculation formula, we can calculate the DI value of each cell in each time window. An example of the DI value calculation formula is as follows: ; in, This represents the DI value calculated for the battery cell within the time window. This represents the individual cell voltage data within a time window. v Functions for performing simple signal processing, including any one of the following: central moment processing, raw moment processing, and absolute value processing. Central moment processing refers to... , This represents the average voltage data of all cells within a time window; the origin moment processing refers to... Whether to use absolute value processing refers to... . This represents the probability density distribution of the battery cell within a time window. l and m These represent the order, both being integers between 1 and 10, and are used to amplify the deviation of the individual cell voltage data.

[0055] They represent l The sum of the order central moments / raw moments / expected value of the absolute value m The expected value of the absolute value of the central moment / origin moment / absolute value.

[0056] Using the above DI value calculation formula, a DI value can be calculated for each cell in each time window. Then, by arranging the DI values ​​of each cell in time sequence, the DI value sequence corresponding to each cell can be obtained.

[0057] In this embodiment, the individual voltage data of all battery cells are divided into time windows by using a sliding window size and a sliding step size to obtain multiple sets of data. The parameters of each individual battery cell are calculated using each set of data to obtain the parameter sequence of each battery cell. This enables anomaly analysis of the battery cells over a long time scale and under multiple operating conditions.

[0058] Example 3 For example, in step 104 above, the parameter sequence corresponding to each cell is determined based on the individual cell voltage data over a continuous time period, including: Based on the individual cell voltage data and at least one preset target order combination, determine the parameter sequence of each cell under each target order combination; the aforementioned target order combination is used to amplify the degree of deviation of the individual cell voltage data. Accordingly, in step 106 above, multidimensional curve feature extraction processing is performed on the parameter sequence of each cell to determine the multidimensional curve features corresponding to each cell, including: For each target order combination, multidimensional curve feature extraction is performed on the parameter sequence of each cell under the target order combination to determine the multidimensional curve features corresponding to each cell under the target order combination.

[0059] Continuing with the example of a battery cell's parameter sequence being its DI value sequence, as shown in the DI value calculation formula above, when calculating the DI value of a battery cell in each time window, it can be obtained from the individual cell voltage data and probability density distribution within the time window, the individual cell voltage data of all cells within the time window, and so on. l and m The order is calculated together from a set of parameters. Based on this, one or more target order combinations that provide the best amplification effect on the deviation of the cell voltage data can be determined in advance using sample data from thermal runaway electrical devices. Each target order combination includes two sets of parameters. l and m The order of the composition, such as the combination of each target order including ( l 1, m 1) and ( l 2, m 2) The four order values ​​in these two sets of orders can be different for different target order combinations.

[0060] After obtaining one or more target order combinations, for each order in each target order combination, the DI value of the battery cell in each time window under that order can be calculated separately. Then, the DI values ​​of the battery cell under that order are combined to obtain the first DI value sequence of the battery cell. By combining the two first DI value sequences of the battery cell under the target order combination (e.g., splicing them together), the DI value sequence of the battery cell under the target order combination can be obtained.

[0061] The above method yields the DI value sequence for each battery cell under each target order combination. Then, multi-dimensional curve feature extraction is performed on the DI value sequence for each battery cell under each target order combination to obtain the multi-dimensional curve features for each battery cell under each target order combination. Taking the six types of curve features extracted above as an example, 6-dimensional curve features can be extracted from the corresponding first DI value sequence for each order group in each target order combination. This allows for the extraction of 12-dimensional curve features for each battery cell under each target order combination, meaning each dimension includes two features under different order groups. This enables horizontal comparison between features of the same dimension, effectively improving the ability to identify abnormal battery cells.

[0062] In this embodiment, the parameter sequence of the battery cell under each order combination is determined by combining the individual cell voltage data with the selected order combination of the amplification factor. In this way, the multidimensional features extracted by the parameter sequences of different order combinations can better distinguish abnormal battery cells and effectively improve the detection capability of abnormal battery cells.

[0063] Example 4 For example, the methods for determining the above-mentioned combination of target orders include: Acquire sample data of multiple candidate order combinations and multiple thermal runaway electrical devices; the sample data includes the sample cell voltage data of each sample cell in the sample battery pack of the thermal runaway electrical device within a continuous time period, and each sample cell includes cells with abnormalities. For each candidate order combination, a sensitivity score is calculated based on the sample data. The sensitivity score is used to characterize the degree to which the candidate order combination distinguishes between normal voltage data and abnormal voltage data. Based on the sensitivity scores of each candidate order combination, the target order combination is selected from the candidate order combinations.

[0064] Continuing with the example of a battery cell whose parameter sequence is its DI value sequence, as mentioned above, the formula for calculating the DI value includes... l and m Given two orders, both of which are integers between 1 and 10, we can construct all possible combinations using permutations and combinations. l and m Combine, and then remove from it. l and m For cells with equal orders, multiple sets of orders are obtained. Then, using the individual cell voltage data, the standard deviation of the DI value sequence composed of all cells under each order is calculated. If the calculated standard deviation is greater than a preset threshold (the specific value can be set according to actual conditions), the order is considered valid and can be retained; otherwise, the order is filtered out / deleted, resulting in multiple valid orders. Then, any two orders from the valid multiple orders are combined to obtain multiple order combinations, all of which are recorded as candidate order combinations.

[0065] Simultaneously, sample data from multiple thermal runaway electrical devices can be acquired. These multiple sample data can be from the same or multiple thermal runaway electrical devices, including sample battery packs. Each sample data includes the individual cell voltage data of each sample cell in the sample battery pack of the corresponding thermal runaway electrical device over a continuous time period. Each sample data includes data from the thermal runaway electrical device under normal battery conditions as well as data from the thermal runaway electrical device under abnormal battery conditions; that is, it includes both normal and abnormal cells.

[0066] After obtaining the sample data and candidate order combinations, for each candidate order combination, the Fisher score can be used to calculate its sensitivity score for each sample data. This sensitivity score reflects the degree to which the candidate order combination can distinguish between normal and abnormal voltage data in the sample data, i.e., its anomaly detection capability. A higher sensitivity score indicates greater sensitivity in distinguishing between normal and abnormal voltage data, meaning it is easier to differentiate between them.

[0067] Based on this, after obtaining the sensitivity scores corresponding to each candidate order combination, the sensitivity scores of each candidate order combination can be sorted in descending order, and one or more candidate order combinations with the highest ranking can be selected as the target order combination for subsequent calculation of parameter sequence set to extract multidimensional curve features.

[0068] In this embodiment, the sensitivity score of each candidate order combination is calculated using multiple thermal runaway sample data, and the target order combination to be used for subsequent cell anomaly detection is selected accordingly. By using the selected optimal order combination for anomaly cell detection, the abnormal cells can be better distinguished, and the detection capability is stronger.

[0069] Example 5 For example, in step 108 above, based on the multidimensional curve characteristics corresponding to each cell, the detection of whether there are abnormalities in each cell includes: For each target order combination, the BIRCH algorithm is used to cluster the multidimensional curve features of all cells under the target order combination to obtain multiple clusters under the target order combination; each cluster includes at least one cell and the multidimensional curve features of the cell. Based on the multidimensional curve characteristics of the cells in each cluster under the target order combination, and the cluster-related information corresponding to each cluster, calculate the anomaly score of each cell under the target order combination; Based on the abnormal score of each cell under the target order combination, detect whether there are any abnormalities in each cell.

[0070] BIRCH stands for Balanced Iterative Reducing and Clustering using Hierarchies. It's an incremental hierarchical clustering algorithm that compresses data and identifies dense regions by constructing a clustering feature tree. Suitable for large-scale datasets, it uses a CF-Tree (Clustering Feature Tree) data structure. This CF-Tree contains multiple nodes, each storing the clustering features of its subordinate data points. The BIRCH algorithm and the multidimensional curve features extracted based on DI value sequences are insensitive to noise and outliers, thus enabling stable abnormal cell detection even under complex real-world vehicle conditions.

[0071] After determining multiple target order combinations and obtaining the multidimensional curve features of each cell under each target order combination, for each target order combination, the vectors corresponding to the multidimensional curve features of all cells under that target order combination can be combined into a feature matrix X, and then the feature matrix X is standardized. Standardization here refers to converting the feature matrix X into a matrix with a mean of 0 and a standard deviation of 1, which can eliminate the differences in dimensions and scales between curve features of different dimensions. The standardized feature matrix is ​​then input into the BIRCH algorithm for clustering. The BIRCH algorithm groups cells with similar curve features into the same cluster by constructing a CF-Tree, while cells with abnormal behavior may form small clusters or outliers, thus dividing multiple cells into multiple clusters. Each cluster can include one or more cells and their multidimensional curve features. Simultaneously, the clustering process obtains cluster-related information for each cluster, including the cluster size, cluster density, and cluster centroid.

[0072] For each cell, the anomaly score of the cell under the target order combination can be calculated by using its multidimensional curve characteristics, the cluster-related information of its cluster, and the cluster-related information of other clusters. The higher the anomaly score, the higher the probability of the cell being abnormal.

[0073] Optionally, the above calculation of the anomaly score for each cell under the target order combination, based on the multidimensional curve characteristics of the cells in each cluster under the target order combination and the cluster-related information corresponding to each cluster, includes: For any first cell among all cells, calculate the distance between the multidimensional curve feature of the first cell under the target order combination and the origin, and obtain the origin distance score. Calculate the cluster size factor based on the cluster size of the cluster to which the first cell belongs; The cluster density factor is calculated based on the cluster density of the cluster to which the first cell belongs and the maximum cluster density among all clusters. Calculate the distance between the multidimensional curve features of the first battery cell and the cluster center points of other clusters to obtain the distance scores of other clusters; the other clusters mentioned above are the clusters other than the cluster to which the first battery cell belongs. Calculate the anomaly score of the first cell under the target order combination based on at least one of the origin distance score, cluster size factor, cluster density factor, and other cluster distance scores corresponding to the first cell under the target order combination.

[0074] Specifically, for any one of the battery cells, denoted as the first cell, the Euclidean distance from the feature point of the first cell (i.e., the vector corresponding to the multidimensional curve feature of the first cell) to the origin (i.e., the zero vector) in the ideal normal state can be calculated. The calculated Euclidean distance is then normalized to obtain the OriginScore corresponding to the first cell. The greater the calculated Euclidean distance, the higher the corresponding OriginScore.

[0075] Simultaneously, by taking the logarithm of the cluster size of the first cell, adding the logarithm to 1, and then taking the reciprocal, the cluster size factor SizeScore corresponding to the first cell can be obtained. The calculation formula is: SizeScore = 1 / (log(cluster size) + 1). It can be seen that cells in the same cluster have the same cluster size factor. The smaller the cluster to which the first cell belongs, the higher its corresponding cluster size factor.

[0076] After obtaining clusters through clustering, for each cluster, the distance between cells can be calculated using the multidimensional curve features of each cell. The average of these distances is then applied to obtain the cluster density. The cluster densities of all clusters are then sorted to find the highest density. For each cell within a cluster, the ratio of its current cluster density to the highest density is calculated. Subtracting this ratio from 1 yields the cluster density factor (DensityScore) for that cell. The formula is: DensityScore = 1 - (Current Cluster Density / Highest Cluster Density). It can be seen that cells within the same cluster have the same cluster density factor. The more dispersed the points within a cell's cluster (i.e., the lower the cluster density), the higher its cluster density factor.

[0077] After obtaining the clusters through clustering, the vector corresponding to the cluster center point of each cluster can also be obtained. For the first battery cell, the Euclidean distance between the feature point of the first battery cell (i.e., the vector corresponding to the multidimensional curve feature of the first battery cell) and the vector corresponding to the cluster center point of its nearest other cluster can be calculated. The calculated Euclidean distance is then normalized to obtain the OtherClusterScore corresponding to the first battery cell. The greater the calculated Euclidean distance, the more isolated the cell is, and the higher its corresponding OtherClusterScore.

[0078] Using the above method, the origin distance score, cluster size factor, cluster density factor, and other cluster distance scores corresponding to any first cell under the target order combination can be obtained. Then, any one of these values ​​can be used as the abnormal score corresponding to the first cell, or multiple values ​​can be selected and weighted to obtain the abnormal score corresponding to the first cell. For example, here we preferentially select to weight the four values ​​of each first cell to obtain the abnormal score corresponding to the first cell.

[0079] The above methods can be used to obtain the abnormal score of each cell under each target order combination. Then, the abnormal scores of each cell under all target order combinations can be used to detect whether any cell is abnormal. For example, the abnormal scores of each cell under all target order combinations can be summed to obtain the total abnormal score for each cell, and the abnormal cells can be identified by the total abnormal score of each cell; or, based on the abnormal scores of each cell, a portion of the target order combinations can be selected, and the abnormal cells can be identified by the magnitude of the selected portion of the target order combinations for each cell; or, based on the abnormal scores of each cell and the corresponding target order combinations, the abnormal cells can be identified together; or, based on the abnormal scores of each cell, the maximum abnormal score or the target order combination corresponding to the maximum abnormal score can be obtained, and the abnormal cells can be identified by the magnitude of the maximum abnormal score or the target order combination corresponding to the maximum abnormal score; or, based on the abnormal scores of each cell, other methods can be used to identify abnormal cells, which are not limited here.

[0080] In this embodiment, the BIRCH algorithm is used to cluster the multidimensional curve features of all battery cells under the same order combination. The clustering results are then used to calculate the anomaly score for each cell to detect abnormal cells, enabling faster and more accurate detection. Furthermore, by calculating scores for each cell in four dimensions—distance from the origin, cluster size, cluster density, and distance to other clusters—the anomaly score is determined. This method yields more reasonable and accurate anomaly scores, leading to more accurate and reasonable identification of abnormal cells, effectively avoiding false alarms caused by instantaneous voltage fluctuations.

[0081] Example 6 For example, the above method of detecting whether each cell is abnormal based on its abnormal score under the target order combination includes: For each cell, obtain the cell's anomaly score under all target order combinations, and determine the anomaly degree index corresponding to the cell based on each anomaly score; the higher the anomaly degree index, the higher the anomaly degree of the corresponding cell. Based on the abnormality index corresponding to each cell, detect whether there are any abnormalities in each cell.

[0082] After obtaining the anomaly score for each cell under all target order combinations, the anomaly scores for each cell can be summed to obtain the anomaly severity index for each cell. This anomaly severity index is proportional to the anomaly severity of the cell and is a scalar value.

[0083] Alternatively, the abnormality scores of each cell can be summed with the corresponding target order to obtain the abnormality index for each cell.

[0084] Alternatively, as an option, the above-mentioned determination of the abnormality level index corresponding to the battery cell based on each abnormality score includes: Identify at least one candidate abnormal score that is greater than the abnormal score threshold from all abnormal scores of the battery cell. The anomaly index corresponding to the cell is determined based on the combination of target orders corresponding to the candidate anomaly scores of the cell.

[0085] For each battery cell, the abnormal scores of that cell can be sorted in descending order, and the top-ranked abnormal scores can be selected as candidate abnormal scores. There can be one or more candidate abnormal scores. Then, the target order combination corresponding to the candidate abnormal scores is found among the target order combinations, and this is taken as the abnormality degree (Degree) of that cell. The abnormality degrees of that cell are then accumulated, i.e., the abnormality degrees of the candidate abnormal scores in the target order combinations are accumulated. l and m The value of is used to obtain the abnormality index corresponding to the battery cell. Among them, l and m These represent the two orders in the target order combination. The values ​​of the two orders can be the same or different. They are mainly used to amplify the deviation of the individual cell voltage data.

[0086] After obtaining the anomaly index for each battery cell, the anomaly indexes of each cell can be sorted in descending order. The cell with the highest anomaly index can be considered the most abnormal cell, or several cells with the highest anomaly index can be considered abnormal cells. Alternatively, cells with an anomaly index greater than the anomaly index threshold (the specific threshold is set according to the actual situation) can be considered abnormal cells. Other methods can also be used to identify abnormal cells. Typically, the anomaly index of an abnormal cell is higher than that of a normal cell.

[0087] In this embodiment, the anomaly index of each cell is determined by the anomaly score of each cell under each target order combination, and anomaly cells are detected accordingly. Detecting anomaly cells using a scalar anomaly index allows for more intuitive identification. Furthermore, for each cell, a subset of order combinations with higher anomaly scores are selected to calculate the cell's anomaly index, resulting in a more accurate calculated index.

[0088] Example 7 For example, the above methods also include: The abnormal scores of each cell are sorted and visualized. And / or, sort the anomaly index of each cell and output the visualization.

[0089] After obtaining the anomaly scores and anomaly severity indices for each cell, the system can sort the anomaly scores of each cell and output visualization results (such as bar charts, line graphs, etc.), and sort the anomaly severity indices of each cell and output visualization results. For example, the final output is an Excel report containing the anomaly scores, anomaly score rankings, and anomaly severity index rankings for each cell under various target order combinations, and provides a 3D PCA (Principal Component Analysis) visualization and a bar chart of anomaly severity index rankings to intuitively show the distribution of abnormal cells.

[0090] In this embodiment, by displaying the abnormality score ranking and the abnormality degree index ranking of each cell, abnormal cells can be identified at a glance, and the interpretability is stronger, making it easier for relevant technical personnel to understand the outlier patterns and decision-making basis of abnormal cells.

[0091] As can be seen from the above embodiments, this application introduces the dimensionless index DI to calculate the DI value of the cell voltage signal, thereby extracting multidimensional curve features and identifying abnormal cells. This amplifies the subtle abnormal changes in the battery voltage signal, enabling accurate capture of early abnormal signs, such as detecting abnormal cells in vehicles involved in thermal runaway accidents 5 days or even earlier. Furthermore, the multidimensional curve features extracted using the DI value are more stable, effectively filtering out normal fluctuations and noise interference under real-world vehicle conditions, thus enhancing anti-interference capabilities and reducing false alarm rates. Moreover, by extracting six types of curve features from the DI value sequence to characterize the cell's voltage dynamic behavior, the cell state can be comprehensively depicted from multiple dimensions, including discreteness, trend, curvature, rate of change, local patterns, and oscillation characteristics. Compared to a single statistical measure, this approach more comprehensively and sensitively captures the entire evolutionary information from minor anomalies to severe faults. Simultaneously, before calculating the DI value sequence of each cell under each target order combination, the system can automatically filter out the target order combinations most sensitive to voltage signal changes through traversal and validity checks. This allows for automatic clustering, scoring, and sorting, ultimately outputting an intuitive result report. This reduces manual intervention, enhances adaptability to different battery systems and operating conditions, and lowers the cost and technical threshold of manual analysis. Furthermore, by employing the BIRCH algorithm for hierarchical clustering and comprehensively calculating the cell's anomaly score from four dimensions—origin distance, cluster size, cluster density, and distance to other clusters—it can effectively identify various anomaly patterns such as "small cluster islands" and "cluster edge points." This effectively avoids false alarms caused by instantaneous voltage fluctuations and significantly improves the detection accuracy of abnormal cells. Moreover, this application's embodiments can perform long-term scale analysis on data from several days prior to an accident, including various operating conditions such as charging and discharging, avoiding the limitations of relying solely on instantaneous or short-term data.

[0092] It should be noted that the cell anomaly detection method involved in the above embodiments may include at least one of Embodiments 1 to 7. For example, Embodiment 1 can be implemented as an independent embodiment, Embodiment 1+2 can be implemented as an independent embodiment, Embodiment 1+2+3 can be implemented as an independent embodiment, Embodiment 1+4 can be implemented as an independent embodiment, Embodiment 1+2+4 can be implemented as an independent embodiment, Embodiment 1+2+3+4 can be implemented as an independent embodiment, Embodiment 1+4+5 can be implemented as an independent embodiment, Embodiment 1+2+4+5 can be implemented as an independent embodiment, Embodiment 1+2+3+4+5 can be implemented as an independent embodiment, and Embodiment 1+4+5+6 can be implemented as... Examples 1+2+4+5+6, 1+2+3+4+5+6, 1+7, 1+2+7, 1+2+3+7, 1+2+3+7, 1+2+3+4+7, 1+2+3+4+5+7, and 1+2+3+4+5+7 can all be implemented as independent embodiments.

[0093] The following describes the cell anomaly detection device provided in this application. The cell anomaly detection device described below can be referred to in correspondence with the cell anomaly detection method described above.

[0094] Figure 2 This is a schematic diagram of the battery cell anomaly detection device provided in this application. See also: Figure 2 As shown, the device may include: The acquisition module 210 is used to acquire the individual cell voltage data of each cell in the battery pack over a continuous time period. The parameter determination module 220 is used to determine the parameter sequence corresponding to each battery cell based on the individual cell voltage data of each cell in a continuous time period; the parameter sequence is used to characterize the changing trend of the individual cell voltage data of the corresponding battery cell in a continuous time period. The feature extraction module 230 is used to perform multidimensional curve feature extraction processing on the parameter sequence of each battery cell to determine the multidimensional curve features corresponding to each battery cell; the multidimensional curve features are used to reflect the changing trend of the parameter sequence of the corresponding battery cell in a continuous time period. The cell anomaly detection module 240 is used to detect whether there is an anomaly in each cell based on the multi-dimensional curve characteristics corresponding to each cell.

[0095] In one embodiment, the parameter determination module 220 is specifically used to determine the parameter sequence of each cell under each target order combination based on the individual cell voltage data of each cell and at least one preset target order combination; the target order combination is used to amplify the deviation of the individual cell voltage data. Accordingly, the feature extraction module 230 is specifically used to perform multi-dimensional curve feature extraction processing on the parameter sequence of each cell under the target order combination for each target order combination, and to determine the multi-dimensional curve features corresponding to each cell under the target order combination.

[0096] In one embodiment, the aforementioned cell anomaly detection module 240 is specifically used to perform clustering processing on the multidimensional curve features of all cells under the target order combination using the BIRCH algorithm for each target order combination, obtaining multiple clusters under the target order combination; each cluster includes at least one cell and the multidimensional curve features of the cell; based on the multidimensional curve features of the cells in each cluster under the target order combination and the cluster-related information corresponding to each cluster, calculate the anomaly score corresponding to each cell under the target order combination; and detect whether each cell has an anomaly based on the anomaly score of each cell under the target order combination.

[0097] In one embodiment, the cluster-related information includes cluster size, cluster density, and cluster center point. The cell anomaly detection module 240 is specifically used to calculate the distance between the multidimensional curve feature of any first cell in the target order combination and the origin point for any first cell among all cells, and obtain the origin distance score; calculate the cluster size factor according to the cluster size of the cluster to which the first cell belongs; calculate the cluster density factor according to the cluster density of the cluster to which the first cell belongs and the maximum cluster density among all clusters; calculate the distance between the multidimensional curve feature of the first cell and the cluster center point of other clusters, and obtain the other cluster distance scores; the other clusters are the clusters other than the cluster to which the first cell belongs; and calculate the anomaly score corresponding to the first cell under the target order combination according to at least one of the origin distance score, cluster size factor, cluster density factor, and other cluster distance scores corresponding to the first cell under the target order combination.

[0098] In one embodiment, the cell anomaly detection module 240 is specifically used to obtain the anomaly score of each cell under all target order combinations, and determine the anomaly degree index corresponding to the cell based on the anomaly scores of each cell; the higher the anomaly degree index, the higher the anomaly degree of the corresponding cell; and detect whether each cell has an anomaly based on the anomaly degree index corresponding to each cell.

[0099] In one embodiment, the cell anomaly detection module 240 is specifically used to determine at least one candidate anomaly score greater than the anomaly score threshold from the various anomaly scores of the cell; and to determine the anomaly degree index corresponding to the cell based on the target order combination corresponding to the candidate anomaly scores of the cell.

[0100] In one embodiment, the above-mentioned apparatus further includes: The visualization output module is used to sort and visualize the abnormal scores of each cell; and / or to sort and visualize the abnormality index of each cell.

[0101] In one embodiment, the above-mentioned apparatus further includes: A screening module is used to acquire sample data of multiple candidate order combinations and multiple thermal runaway electrical devices. The sample data includes the sample cell voltage data of each sample cell in the sample battery pack of the thermal runaway electrical device within a continuous time period, and each sample cell includes cells with abnormalities. For each candidate order combination, a sensitivity score is calculated based on the sample data. The sensitivity score is used to characterize the degree to which the candidate order combination distinguishes between normal voltage data and abnormal voltage data. Based on the sensitivity scores of each candidate order combination, the target order combination is selected from the candidate order combinations.

[0102] In one embodiment, the parameter determination module 220 is specifically used to obtain a set sliding window size and a set sliding step size; divide the individual cell voltage data of all cells in a continuous time period on the time axis according to the sliding window size and sliding step size to obtain multiple time windows; each time window includes a portion of the individual cell voltage data of all cells; determine the parameters of each cell in each time window based on the portion of the individual cell voltage data of all cells in each time window; and determine the parameter sequence corresponding to each cell based on the parameters of each cell in all time windows.

[0103] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application, such as... Figure 3As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340. The processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a cell anomaly detection method. This method includes: acquiring the individual cell voltage data of each cell in the battery pack over a continuous time period; determining the parameter sequence corresponding to each cell based on the individual cell voltage data over the continuous time period; the parameter sequence characterizing the changing trend of the individual cell voltage data over the continuous time period; performing multidimensional curve feature extraction processing on the parameter sequence of each cell to determine the multidimensional curve feature corresponding to each cell; the multidimensional curve feature reflecting the changing trend of the parameter sequence of the corresponding cell over the continuous time period; and detecting whether each cell has an anomaly based on the multidimensional curve feature corresponding to each cell.

[0104] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0105] Additionally, this application provides an electrical device including a battery pack and the aforementioned electronic equipment. This electrical device may be, for example, a vehicle, or it may be a robot, a person / cargo transport device, etc.

[0106] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the cell anomaly detection method provided by the above methods. The method includes: acquiring the single-cell voltage data of each cell in the battery pack over a continuous time period; determining the parameter sequence corresponding to each cell based on the single-cell voltage data of each cell over the continuous time period; the parameter sequence being used to characterize the changing trend of the single-cell voltage data of the corresponding cell over the continuous time period; performing multi-dimensional curve feature extraction processing on the parameter sequence of each cell to determine the multi-dimensional curve feature corresponding to each cell; the multi-dimensional curve feature being used to reflect the changing trend of the parameter sequence of the corresponding cell over the continuous time period; and detecting whether each cell has an anomaly based on the multi-dimensional curve feature corresponding to each cell.

[0107] Furthermore, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the cell anomaly detection method provided by the above methods. The method includes: acquiring single-cell voltage data of each cell in a battery pack over a continuous time period; determining a parameter sequence corresponding to each cell based on the single-cell voltage data of each cell over the continuous time period; the parameter sequence being used to characterize the changing trend of the single-cell voltage data of the corresponding cell over the continuous time period; performing multi-dimensional curve feature extraction processing on the parameter sequence of each cell to determine the multi-dimensional curve feature corresponding to each cell; the multi-dimensional curve feature being used to reflect the changing trend of the parameter sequence of the corresponding cell over the continuous time period; and detecting whether each cell has an anomaly based on the multi-dimensional curve feature corresponding to each cell.

[0108] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0110] 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 abnormalities in battery cells, characterized in that, include: Acquire the individual cell voltage data of each cell in the battery pack over a continuous time period; Based on the individual cell voltage data of each cell within the continuous time period, a parameter sequence corresponding to each cell is determined; the parameter sequence is used to characterize the changing trend of the individual cell voltage data of the corresponding cell within the continuous time period. Multidimensional curve feature extraction processing is performed on the parameter sequence of each battery cell to determine the multidimensional curve features corresponding to each battery cell; The multidimensional curve feature is used to reflect the changing trend of the parameter sequence of the corresponding battery cell within the continuous time period. Based on the multidimensional curve characteristics corresponding to each of the battery cells, detect whether there are any abnormalities in each of the battery cells.

2. The cell anomaly detection method according to claim 1, characterized in that, The step of determining the parameter sequence corresponding to each battery cell based on the individual cell voltage data within the continuous time period includes: Based on the individual cell voltage data of each cell and at least one preset target order combination, determine the parameter sequence of each cell under each target order combination; the target order combination is used to amplify the deviation of the individual cell voltage data. Accordingly, the step of performing multidimensional curve feature extraction processing on the parameter sequence of each battery cell to determine the multidimensional curve features corresponding to each battery cell includes: For each target order combination, multidimensional curve feature extraction is performed on the parameter sequence of each cell under the target order combination to determine the multidimensional curve features corresponding to each cell under the target order combination.

3. The cell anomaly detection method according to claim 2, characterized in that, The step of detecting whether each of the battery cells is abnormal based on the multidimensional curve characteristics corresponding to each of the battery cells includes: For each target order combination, the balanced iterative reduction hierarchical clustering BIRCH algorithm is used to cluster the multidimensional curve features of all cells under the target order combination to obtain multiple clusters under the target order combination; each cluster includes at least one cell and the multidimensional curve features of the cell. Based on the multidimensional curve characteristics of the cells in each cluster under the target order combination, and the cluster-related information corresponding to each cluster, calculate the anomaly score corresponding to each cell under the target order combination; Based on the abnormal score of each cell under the target order combination, detect whether each cell has an abnormality.

4. The cell anomaly detection method according to claim 3, characterized in that, The cluster-related information includes cluster size, cluster density, and cluster center point. The calculation of the anomaly score for each cell under the target order combination, based on the multidimensional curve characteristics of the cells in each cluster under the target order combination and the cluster-related information corresponding to each cluster, includes: For any first cell among all the said cells, calculate the distance between the multidimensional curve feature of the first cell under the target order combination and the origin, and obtain the origin distance score; Calculate the cluster size factor based on the cluster size of the cluster to which the first battery cell belongs; The cluster density factor is calculated based on the cluster density of the cluster to which the first cell belongs and the maximum cluster density among all the cluster densities. Calculate the distance between the multidimensional curve features of the first battery cell and the cluster center points of other clusters to obtain the distance scores of other clusters; the other clusters are the clusters other than the cluster to which the first battery cell belongs. Based on at least one of the origin distance score, cluster size factor, cluster density factor, and other cluster distance scores corresponding to the first cell under the target order combination, calculate the anomaly score corresponding to the first cell under the target order combination.

5. The cell anomaly detection method according to claim 3 or 4, characterized in that, The step of detecting whether each of the battery cells is abnormal based on the abnormal score of each of the battery cells under the target order combination includes: For each of the battery cells, the abnormality score of the battery cell under all target order combinations is obtained, and the abnormality index corresponding to the battery cell is determined based on each abnormality score of the battery cell; the higher the abnormality index, the higher the abnormality of the corresponding battery cell. Based on the abnormality index corresponding to each of the battery cells, it is determined whether each of the battery cells is abnormal.

6. The cell anomaly detection method according to claim 5, characterized in that, The step of determining the abnormality index corresponding to the battery cell based on each abnormality score of the battery cell includes: From the abnormal scores of the battery cell, at least one candidate abnormal score that is greater than the abnormal score threshold is determined; The abnormality index corresponding to the battery cell is determined based on the target order combination corresponding to the candidate abnormality scores of the battery cell.

7. The cell anomaly detection method according to any one of claims 2 to 6, characterized in that, The methods for determining the target order combination include: Acquire sample data of multiple candidate order combinations and multiple thermal runaway electrical devices; the sample data includes sample cell voltage data of each sample cell in the sample battery pack of the thermal runaway electrical device within a continuous time period, and each sample cell includes cells with abnormalities; For each candidate order combination, a sensitivity score is calculated based on the sample data. The sensitivity score is used to characterize the degree to which the candidate order combination distinguishes between normal voltage data and abnormal voltage data. Based on the sensitivity scores of each candidate order combination, the target order combination is selected from the candidate order combinations.

8. The cell anomaly detection method according to claim 1, characterized in that, The step of determining the parameter sequence corresponding to each battery cell based on the individual cell voltage data within the continuous time period includes: Get the set sliding window size and the set sliding step size; According to the sliding window size and the sliding step size, the individual voltage data of all cells in the continuous time period are divided on the time axis to obtain multiple time windows; each time window includes a portion of the individual voltage data of all cells. Based on a portion of the individual cell voltage data within each time window, the parameters of each cell in each time window are determined. Based on the parameters of each cell across the entire time window, determine the parameter sequence corresponding to each cell.

9. A battery cell anomaly detection device, characterized in that, include: The acquisition module is used to acquire the individual cell voltage data of each cell in the battery pack over a continuous time period. The parameter determination module is used to determine the parameter sequence corresponding to each battery cell based on the individual cell voltage data of each battery cell in the continuous time period; the parameter sequence is used to characterize the changing trend of the individual cell voltage data of the corresponding battery cell in the continuous time period. The feature extraction module is used to perform multidimensional curve feature extraction processing on the parameter sequence of each battery cell to determine the multidimensional curve features corresponding to each battery cell; the multidimensional curve features are used to reflect the changing trend of the parameter sequence of the corresponding battery cell in the continuous time period. The cell anomaly detection module is used to detect whether there is an anomaly in each of the cells based on the multidimensional curve characteristics corresponding to each cell.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the cell anomaly detection method as described in any one of claims 1 to 8.

11. A non-transitory 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 cell anomaly detection method as described in any one of claims 1 to 8.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the cell anomaly detection method as described in any one of claims 1 to 8.

Images