Abnormal battery cell identification method and device

Abstract

The invention provides an abnormal battery cell identification method and device, and relates to the technical field of batteries, and the method comprises the steps: obtaining the working condition data of each battery cell in a to-be-detected battery under different working conditions, carrying out the feature extraction of the working condition data corresponding to each working condition, so as to obtain K groups of feature data corresponding to each working condition, determining an optimal bandwidth of the K groups of feature data corresponding to each working condition, constructing a kernel density estimation model based on the optimal bandwidth and the K groups of feature data, performing grid division in a space formed by the K groups of feature data corresponding to each working condition, and calculating a kernel density value of each grid point based on the kernel density estimation model, according to the characteristic data of each battery cell in the K groups of characteristic data and the nuclear density value of each grid point, the abnormal battery cell in the to-be-detected battery is identified, the battery cell characteristic data is combined with the space nuclear density distribution, the detection of the abnormal battery cell under the complex working condition is realized, and the accuracy, the adaptability and the real-time performance are considered.

Description

A method and apparatus for identifying abnormal battery cells Technical Field

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

[0002] With the rapid development of electric vehicles, energy storage systems, and other fields, battery safety has become increasingly important. For example, lithium-ion batteries are widely used due to their high energy density and long cycle life. However, the safety of lithium-ion batteries is significantly affected by the consistency of individual cells. If some cells experience abnormal voltage due to manufacturing defects, aging, or misuse, it may lead to localized overheating, thermal runaway, or even a chain reaction resulting in an explosion. Therefore, quickly identifying abnormal cells in a battery is crucial to ensuring the safety of the battery system.

[0003] Abnormal cell behavior manifests as voltage deviations from the normal range, abnormal temperature increases, or sudden changes in internal resistance. If not detected promptly, these abnormalities can lead to thermal runaway, threatening the safety of the entire battery system. In related technologies, threshold methods are commonly used to identify abnormal cells, but these rely on fixed thresholds and are ill-suited to dynamic conditions such as fast charging and sudden load changes. Statistical methods based on the 3σ criterion assume that the data follows a normal distribution, making them poorly adaptable to common non-Gaussian distributions in real-world scenarios. Machine learning methods such as SVM or neural networks face challenges such as model dependence on large amounts of labeled data and limitations in generalization ability due to the training set. Clustering methods (such as K-means) suffer from high computational complexity and insufficient real-time performance.

[0004] In summary, the relevant technologies rely on fixed thresholds, strong distribution assumptions, large amounts of labeled data, or high computational costs, making it difficult to balance accuracy, adaptability, and real-time performance under complex operating conditions. Summary of the Invention

[0005] The problem solved by this invention is how to detect abnormal battery cells under complex operating conditions, while taking into account accuracy, adaptability and real-time performance.

[0006] To address the above problems, the present invention provides a method and apparatus for identifying abnormal battery cells.

[0007] In a first aspect, the present invention provides a method for identifying abnormal battery cells, comprising: acquiring operating condition data of each cell in a battery under test under different operating conditions, wherein the operating conditions include at least a discharge operating condition and a charging operating condition; performing feature extraction on the operating condition data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition, wherein K is an integer greater than or equal to 2; wherein each set of feature data includes feature data of each cell in the battery under test; determining the optimal bandwidth of the K sets of feature data corresponding to each operating condition, and constructing a kernel density estimation model based on the optimal bandwidth and the K sets of feature data; performing grid division within the space formed by the K sets of feature data corresponding to each operating condition, and calculating the kernel density value of each grid point based on the kernel density estimation model; and identifying abnormal cells in the battery under test according to the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

[0008] Optionally, the step of extracting features from the operating data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition includes: converting the operating data corresponding to each operating condition into a first matrix; the first matrix having time as columns and cell operation data as rows; performing decentralization processing on the first matrix to obtain a decentralized first matrix; and performing feature extraction based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition.

[0009] Optionally, the process of decentralizing the first matrix includes: determining the average operating data of each cell under the corresponding operating condition based on the first matrix; and subtracting the average operating data of the corresponding cell from the operating data of each cell at different times in the first matrix to obtain the decentralized first matrix.

[0010] Optionally, the step of extracting features based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition includes: determining the corresponding covariance matrix based on the decentralized first matrix; determining the correlation coefficient between the operating data of each cell at different times based on the covariance matrix; and taking the first K operating data of each cell in the decentralized first matrix as the K sets of feature data in descending order of the correlation coefficient.

[0011] Optionally, the step of extracting features based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition includes: determining the time-frequency domain features corresponding to each operating data in the decentralized first matrix based on the decentralized first matrix; and taking the first K operating data of each cell in the decentralized first matrix as the K sets of feature data in descending order of the time-frequency domain features.

[0012] Optionally, determining the optimal bandwidth for the K sets of feature data corresponding to each operating condition includes: determining the standard deviation of all operating data in the battery to be tested; determining the adjustment coefficient based on the Gaussian kernel function; and multiplying the standard deviation by the adjustment coefficient to obtain the optimal bandwidth.

[0013] Optionally, constructing the kernel density estimation model based on the optimal bandwidth and the K sets of feature data includes: constructing the kernel density estimation model according to the following expression: Where n is the number of cells in the battery to be tested. Indicates the first Operating data of each battery cell Let represent the average operating data of n battery cells, and h be the optimal bandwidth. This represents the kernel density value.

[0014] Optionally, identifying abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point includes: determining outlier regions based on the kernel density value of each grid point, and determining candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the outlier regions; and / or determining the point corresponding to the mean of the kernel density values ​​of all grid points based on the kernel density value of each grid point, and determining candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the point corresponding to the mean of the kernel density values; determining whether the feature data of the candidate abnormal cells meets preset conditions; if the feature data of the candidate abnormal cells meets the preset conditions, then determining that the candidate abnormal cells are abnormal cells under the current operating conditions.

[0015] Optionally, determining the outlier region based on the kernel density value of each grid point includes: determining the continuous region formed by grid points whose kernel density values ​​are less than a preset kernel density threshold as the outlier region, wherein the preset kernel density threshold is the kernel density value of all grid points that is located at a predetermined quantile after being arranged in ascending order.

[0016] Optionally, the method further includes: determining the kernel density estimate corresponding to the feature data of each cell in the K sets of feature data according to the kernel density estimation model; determining candidate abnormal cells based on the relationship between the kernel density estimate corresponding to the feature data of each cell in the K sets of feature data and a preset threshold; determining whether the feature data of the candidate abnormal cells meets a preset condition; and determining that the candidate abnormal cells are abnormal cells under the current operating conditions if the feature data of the candidate abnormal cells meet the preset condition.

[0017] Optionally, it further includes: if the battery cell is determined to be an abnormal battery cell under operating conditions exceeding a preset number, then the battery cell is determined to be an abnormal battery cell.

[0018] Optionally, acquiring the operating condition data of each cell in the battery under test under different operating conditions includes: dividing the operating data of each cell in the battery under test into multiple operating condition data for each cell according to different operating conditions.

[0019] Secondly, the present invention provides an abnormal cell identification device, characterized in that it comprises: a working condition data acquisition module, used to acquire working condition data of each cell in a battery under test under different working conditions, wherein the working conditions include at least a discharge working condition and a charging working condition; a feature extraction module, used to extract features from the working condition data corresponding to each working condition to obtain K sets of feature data corresponding to each working condition, where K is an integer greater than or equal to 2; wherein each set of feature data includes feature data of each cell in the battery under test; a kernel density estimation model construction module, used to determine the optimal bandwidth of the K sets of feature data corresponding to each working condition, and construct a kernel density estimation model based on the optimal bandwidth and the K sets of feature data; a kernel density value calculation module, used to perform grid division within the space formed by the K sets of feature data corresponding to each working condition, and calculate the kernel density value of each grid point obtained based on the kernel density estimation model; and an abnormal cell identification module, used to identify abnormal cells in the battery under test according to the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

[0020] Thirdly, the present invention provides an electronic device, including a memory and a processor; the memory is used to store a computer program; the processor is used to implement the abnormal battery cell identification method as described in the first aspect when the computer program is executed.

[0021] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the abnormal battery cell identification method as described in the first aspect.

[0022] The beneficial effects of the abnormal cell identification method and device of the present invention are as follows: It acquires operating condition data of each cell in the battery under test under different operating conditions, including at least discharge and charging conditions. By acquiring the operating condition data of each cell under different operating conditions, it achieves a refined analysis of the dynamic operation process of the cells, laying the foundation for subsequent accurate anomaly detection under different operating conditions. Feature extraction is performed on the operating condition data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition. Each set of feature data includes feature data of each cell in the battery under test. Through feature extraction, the transformation from raw time series data to key feature information is realized, effectively compressing the amount of data and highlighting key feature information related to anomalies. The optimal bandwidth of the K sets of feature data corresponding to each operating condition is determined, and a kernel density estimation model is constructed based on the optimal bandwidth and the K sets of feature data. By determining the optimal bandwidth and constructing the kernel density estimation model, adaptive fitting to the true distribution of feature data is achieved. A grid is created within the space defined by K sets of feature data corresponding to each operating condition. Based on a kernel density estimation model, the kernel density value of each grid point is calculated. Through grid division and kernel density value calculation, intuitive and quantitative localization of abnormal low-density areas is achieved. Abnormal cells in the battery under test are identified based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point. By combining cell feature data with spatial kernel density distribution, accurate identification of abnormal cells under various operating conditions is achieved without the need for pre-labeled data, exhibiting good real-time performance and adaptive fitting to the true distribution of feature data, demonstrating good adaptability. Attached Figure Description

[0023] Figure 1 is a flowchart of an abnormal battery cell identification method according to an embodiment of the present invention; Figure 2 is a schematic diagram of the curve corresponding to the voltage time series data of each battery cell according to an embodiment; Figure 3 is a schematic diagram of the distribution of kernel density values ​​of each grid point according to an embodiment; Figure 4 is a flowchart of feature extraction of operating condition data corresponding to each operating condition according to an embodiment; Figure 5 is a schematic diagram of the initial matrix according to an embodiment; Figure 6 is a schematic diagram of the first matrix according to an embodiment; Figure 7 is a schematic diagram of two sets of feature data according to an embodiment; Figure 8 is a schematic diagram of the distribution of kernel density estimates corresponding to two feature data of each battery cell according to an embodiment; Figure 9 is a schematic diagram of the distribution of abnormal battery cell feature data according to an embodiment; Figure 10 is a structural schematic diagram of an abnormal battery cell identification device according to an embodiment of the present invention; Figure 11 is a structural schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0025] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0026] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0027] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0028] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0029] As shown in Figure 1, an embodiment of the present invention provides a method for identifying abnormal battery cells, including: step S100: acquiring the operating condition data of each battery cell in the battery to be tested under different operating conditions.

[0030] Specifically, each operating condition data corresponds to one operating condition, which includes at least discharge operating condition and charging operating condition, and may also include static operating condition.

[0031] Specifically, the battery to be tested can be a lithium-ion battery, a sodium-ion battery, etc.

[0032] In some embodiments, the operating data of each cell in the battery under test is divided into multiple operating condition data for each cell according to different operating conditions, so that the operating condition data of each cell under different operating conditions can be obtained. The operating data of the battery under test can be divided into multiple operating condition data according to different operating conditions in the following way: Charging condition: When the current I > a certain positive threshold, the battery system absorbs energy from the outside, and at this time, the voltage generally rises.

[0033] Discharge condition: When the current I is less than a certain negative threshold, the battery system releases energy to the outside, and at this time, the voltage generally drops.

[0034] Idle condition: When the current |I| ≤ a certain threshold (e.g., 6A), the battery system is in an idle state.

[0035] In addition, for the operating condition data segmented in the above manner, the state of charge (SOC) of the battery system can be used for auxiliary verification. For example, for the operating condition data of the charging condition, check whether its SOC is generally increasing; for the operating condition data of the discharging condition, check whether its SOC is generally decreasing; and for the operating condition data of the resting condition, check whether its SOC is stable.

[0036] Specifically, the operating data of the battery under test includes at least the operating time-series data of each cell in the battery under test. The operating time-series data includes voltage time-series data, temperature time-series data, or internal resistance time-series data. In addition, since operating condition segmentation is also required, the operating data of the battery under test may also include the current time-series data of each cell in the battery under test. As shown in Figure 2, Figure 2 shows a schematic diagram of the curves corresponding to the voltage time-series data of each cell.

[0037] Specifically, the time interval corresponding to each operating condition data should be at least greater than or equal to 30 minutes to ensure that each segmented operating condition data has statistical significance, fully covers the battery dynamic process, and effectively filters short-term interference.

[0038] Step S200: Extract features from the operating data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition; wherein, each set of feature data includes the feature data of each cell in the battery to be tested.

[0039] Specifically, the operating condition data is the runtime sequence data, and the K sets of feature data are the K feature data corresponding to each battery cell. K can be a positive integer greater than or equal to 2, specifically 2 to 5. Therefore, feature extraction reduces the amount of subsequent data processing, and the extracted feature data can highlight the consistency and behavior patterns of the battery cells, thus providing data support for the subsequent identification of abnormal battery cells.

[0040] Step S300: Determine the optimal bandwidth of the K sets of feature data corresponding to each working condition, and construct a kernel density estimation model based on the optimal bandwidth and the K sets of feature data.

[0041] Specifically, the optimal bandwidth determines the accuracy and smoothness of the kernel density estimation model's fit to the data distribution. As the bandwidth increases, the peak values ​​of the data gradually become slower, the number of peaks decreases, and the variance of the data increases. Therefore, it is necessary to determine the optimal bandwidth for the K sets of feature data corresponding to each working condition.

[0042] Specifically, after determining the optimal bandwidth, a non-parametric probability density estimation method can be used to construct a kernel density estimation model. Specifically, each data point (i.e., the feature value of a battery cell) is regarded as the center of a probability density kernel (such as a Gaussian kernel), and the determined optimal bandwidth is used as the smoothing parameter of the kernel function. The kernel density estimation model forms a smooth and continuous probability density surface by superimposing the kernel functions corresponding to all data points. This surface can represent the clustering pattern of all normal battery cells in the multidimensional space formed by K sets of features, while those points that deviate from the main clustering area and fall in the area of ​​extremely low probability density are naturally identified as abnormal candidate objects to be detected, thereby realizing accurate and quantitative identification of abnormal battery cells.

[0043] Step S400: Divide the space into grids within the K sets of feature data corresponding to each working condition, and calculate the kernel density value of each grid point based on the kernel density estimation model.

[0044] Specifically, taking K=2 as an example, for each operating condition, each cell has two feature data, namely feature F1 and feature F2. A value range is defined for feature F1 and feature F2, for example, F1 from 0 to 0.3 and F2 from 0 to 0.5. Then, this two-dimensional region is divided into a regular grid array. For example, the F1 direction is divided into 100 equal parts and the F2 direction is divided into 100 equal parts, thereby generating 10,000 grid points. Each grid point represents a specific (F1, F2) feature combination.

[0045] Specifically, based on the constructed kernel density estimation model, the kernel density value of each grid point is calculated. For example, assuming a grid point is (F1... i F2 i ), grid points (F1) i F2 i Substituting this into the kernel density estimation model, we can obtain the grid point (F1). i F2 i The kernel density values ​​of the grid points are shown in Figure 3. Figure 3 shows a schematic diagram of the distribution of kernel density values ​​at each grid point.

[0046] Step S500: Identify abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

[0047] Following the above steps, the kernel density values ​​of all grid points can be calculated, resulting in a continuous kernel density distribution surface covering the entire feature space. Regions with extremely low values ​​on the surface, i.e., grid point sets far from all dense data point areas, are identified as outlier regions. For a battery cell, if the coordinates corresponding to its feature data fall within these low-density grid regions, it can be identified as a candidate abnormal battery cell. This method does not rely on a fixed threshold, adapts to dynamic distribution, avoids threshold failure caused by battery aging and temperature changes, and reduces the false alarm rate.

[0048] In this embodiment, operating condition data of each cell in the battery under test is acquired under different operating conditions, including at least discharge and charging conditions. By acquiring the operating condition data of each cell under different operating conditions, a refined analysis of the dynamic operation process is achieved, laying the foundation for accurate anomaly detection under different operating conditions. Feature extraction is performed on the operating condition data corresponding to each operating condition to obtain K sets of feature data for each operating condition. Each set of feature data includes feature data of each cell in the battery under test. Through feature extraction, the transformation from raw time series data to key characterization information is realized, effectively compressing the data scale and highlighting key information related to anomalies. The optimal bandwidth of the K sets of feature data corresponding to each operating condition is determined, and a kernel density estimation model is constructed based on the optimal bandwidth and the K sets of feature data. By determining the optimal bandwidth and constructing the kernel density estimation model, adaptive fitting of the true distribution of feature data is achieved. The space formed by the K sets of feature data corresponding to each operating condition is divided into grids, and the kernel density value of each grid point is calculated based on the kernel density estimation model. Through grid division and calculation of kernel density values ​​of grid points, intuitive and quantitative localization of abnormal low-density areas is achieved. Based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point, abnormal cells in the battery under test are identified. By combining the cell feature data with the spatial kernel density distribution, accurate identification of abnormal cells is achieved.

[0049] Optionally, as shown in Figure 4, feature extraction is performed on the operating data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition, including the following steps: Step S410: Convert the operating data corresponding to each operating condition into a first matrix; wherein, the first matrix has time as the column and the cell operation data as the row.

[0050] In some embodiments, the operating condition data is the voltage time series data of each battery cell. First, the voltage time series data of each battery cell is converted into an initial matrix with time as the row and battery cell voltage as the column, as shown in Figure 5, where volt_0, volt_1, volt_2, volt_i (i=3 to n-1), volt_n is the voltage of n battery cells, and t is the time. Then, the initial matrix with time as the column and battery cell voltage as the row is transposed to obtain a first matrix with time as the column and battery cell voltage as the row, as shown in Figure 6. Step S420: Decentralize the first matrix to obtain the decentralized first matrix.

[0051] In some embodiments, decentralizing the first matrix includes: determining the average operating data (e.g., average voltage) of each cell at all times based on the first matrix, i.e., calculating the average of all data points in the first matrix; subtracting the average operating data of the corresponding cell (e.g., subtracting the average voltage from the voltage value of each cell) from the operating data of each cell at different times in the first matrix to obtain the decentralized first matrix, i.e., subtracting the average from the value of each data point in the first matrix.

[0052] Before feature extraction, the data in the first matrix needs to be decentered. This involves subtracting the mean from the data in the first matrix to move the origin of the coordinate system of each data point in the first matrix to the center (mean) of all data. This ensures that the position of each data point is no longer affected by the overall translation, but is determined only by the relative relationship between the data point and other points. This provides accurate data support for accurately identifying abnormal cells that consistently deviate from the average level of the group during the dynamic process.

[0053] Step S430: Perform feature extraction based on the decentralized first matrix to obtain K sets of feature data corresponding to each working condition.

[0054] In some embodiments, feature extraction may include feature extraction based on the covariance matrix, and may also include time-frequency domain feature extraction, such as extracting features such as mean, variance, FFT spectral energy, wavelet entropy, etc., to enhance anomaly detection capabilities.

[0055] In some embodiments, for the case of K=2, two sets of feature data are obtained, which include two feature data for each cell, as shown in Figure 7.

[0056] In some embodiments, step S430 involves feature extraction based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition, including: Step S431: Determining the corresponding covariance matrix based on the decentralized first matrix. Taking the cell voltage as an example, since the rows of the first matrix represent time and the columns represent cell voltage, each element in the corresponding covariance matrix represents the covariance Cov(volt_i_1', volt_i_2') between cell voltages at different times, where volt_i_1' is the decentralized voltage of cell i at time 1 and volt_i_2' is the decentralized voltage of cell i at time 2. Step S432: Determining the correlation coefficient between the operating data of each cell at different times based on the covariance matrix. Taking voltage as an example, the correlation coefficient is Cov(volt_i_1',volt_i_2') / (σ_1*σ_2), where σ_1 is the standard deviation of the decentralized voltage of cell i at time 1, and σ_2 is the standard deviation of the decentralized voltage of cell i at time 2.

[0057] Step S433: According to the correlation coefficient from largest to smallest, take the first K operating data of each cell in the decentralized first matrix as K sets of feature data, that is, select the top K data from the sorted list.

[0058] In other embodiments, step S330 involves feature extraction based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition. This includes: determining the time-frequency domain features corresponding to each operating data in the decentralized first matrix based on the decentralized first matrix; and selecting the first K voltage data of each cell in the decentralized first matrix as K sets of feature data in descending order of the time-frequency domain features. The time-frequency domain features may include at least one of the following: mean, variance, FFT spectral energy, wavelet entropy, etc.

[0059] In this optional embodiment, by converting voltage time-series data into a matrix and decentralizing it, a data foundation is provided for feature extraction. Feature extraction based on this foundation provides comprehensive and reliable data support for subsequent accurate identification of abnormal battery cells.

[0060] Optionally, the optimal bandwidth for the K sets of feature data corresponding to each operating condition is determined, including: determining the standard deviation of the operating data of all cells in the battery to be tested; determining the adjustment coefficient based on the Gaussian kernel function; and multiplying the standard deviation by the adjustment coefficient to obtain the optimal bandwidth.

[0061] The optimal bandwidth h is determined using the following formula: Where h is the optimal bandwidth and n is the number of cells in the lithium battery; Standard deviation , The data includes the operating data (voltage, temperature, or internal resistance) of cell i in the battery to be tested. This represents the average of the operating data of all cells in the battery under test. The adjustment coefficient is the optimal bandwidth adjustment coefficient based on the Gaussian kernel, derived from Silverman's empirical rule.

[0062] In this optional embodiment, by using the standard deviation of the operating data of all cells combined with the Gaussian kernel function to adjust the coefficients, an efficient and stable estimation of the optimal bandwidth is achieved.

[0063] Optionally, based on the optimal bandwidth and K sets of feature data, a kernel density estimation model is constructed, including: constructing the kernel density estimation model according to the following expression: Where n is the number of cells in the lithium battery; Indicates the first Operating data of each battery cell Let represent the average operating data of n battery cells, and h be the optimal bandwidth. This represents the kernel density value.

[0064] In this optional embodiment, a non-parametric kernel density estimation method is used to adaptively fit the true distribution of cell voltage data with optimal bandwidth, thereby accurately identifying abnormal cells that deviate from the main distribution and effectively improving the accuracy and adaptability of detection.

[0065] Optionally, step S500, which identifies abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point, includes the following steps: Step S510: Determine outlier regions based on the kernel density value of each grid point, and determine candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the outlier regions; and / or, determine the point corresponding to the mean of the kernel density values ​​of all grid points based on the kernel density value of each grid point, and determine candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the point corresponding to the mean of the kernel density values.

[0066] Specifically, determining outlier regions based on the kernel density value of each grid point includes: determining a continuous region consisting of grid points whose kernel density values ​​are less than a preset kernel density threshold as an outlier region, wherein the preset kernel density threshold is the kernel density value that is located at a predetermined quantile (e.g., 1%) after all grid points are arranged in ascending order.

[0067] Specifically, based on the positional relationship between the feature data of each cell in the K sets of feature data and the outlier region, candidate abnormal cells are determined, including: taking each feature data in the K sets of feature data as a data point, determining the shortest Euclidean distance between each data point and the outlier region, and taking the cell outside the outlier region corresponding to the data point with the smallest shortest Euclidean distance as a candidate abnormal cell.

[0068] Specifically, based on the positional relationship between the feature data of each cell in the K sets of feature data and the point corresponding to the mean of the kernel density value, candidate abnormal cells are determined, including: treating each feature data in the K sets of feature data as a data point, and taking the cell corresponding to the data point farthest from the mean point corresponding to the mean of the kernel density value as a candidate abnormal cell.

[0069] In other embodiments, the kernel density estimate corresponding to the feature data of each cell in the K sets of feature data can be determined according to the kernel density estimation model, as shown in Figure 8. Figure 8 shows a schematic diagram of the distribution of the kernel density estimates corresponding to the two feature data of each cell. Candidate abnormal cells are determined based on the relationship between the kernel density estimates corresponding to the feature data of each cell in the K sets of feature data and a preset threshold. For example, the preset threshold is determined using the 3σ criterion based on the kernel density estimates corresponding to the feature data of each cell in the K sets of feature data, thereby identifying candidate abnormal cells.

[0070] Step S520: Determine whether the feature data of the candidate abnormal battery cell meets the preset conditions.

[0071] Step S530: If the characteristic data of the candidate abnormal cell meets the preset conditions, then the candidate abnormal cell is determined to be an abnormal cell under the current operating conditions.

[0072] In some embodiments, the preset conditions may be: (1) the feature data of the candidate abnormal cell exceeds the range of a predetermined standard deviation multiple (e.g., 2 times) of the distribution of the feature data of the normal cell; (2) the feature data of the candidate abnormal cell exceeds the range of a predetermined multiple quantile (e.g., 10 times the quantile) of the distribution of the feature data of the normal cell; the above-mentioned distribution of the feature data of the normal cell refers to the distribution of the feature data of other cells besides the candidate abnormal cell. If both of the above conditions are met, the candidate abnormal cell is determined to be an abnormal cell under the current operating condition. As shown in Figure 9, Figure 9 shows a schematic diagram of the distribution of the feature data of the abnormal cell that meets both of the above conditions.

[0073] Furthermore, if a battery cell is identified as an abnormal battery cell under operating conditions exceeding a preset number, then the battery cell is determined to be an abnormal battery cell.

[0074] In this optional embodiment, compared to a single outlier detection method, this embodiment uses dual statistical verification (2σ criterion and quantile criterion) and multi-condition confirmation to accurately identify abnormal cells and avoid triggering fault alarms due to instantaneous noise.

[0075] As shown in Figure 10, an abnormal battery cell identification device 1000 provided in this embodiment of the invention includes: a working condition data acquisition module 1010, used to divide the operating data of the battery under test into multiple working condition data according to different working conditions; each working condition data corresponds to one working condition, the working condition includes at least a discharge working condition and a charging working condition, and the operating data of the battery under test includes at least the operating time series data of each cell in the battery under test; a feature extraction module 1020, used to extract features from the working condition data corresponding to each working condition to obtain K sets of feature data corresponding to each working condition; wherein, each set of feature data includes the operating time series data of each cell in the battery under test. The system includes: a cell feature data module; a kernel density estimation model construction module 1030, used to determine the optimal bandwidth of K sets of feature data corresponding to each operating condition, and to construct a kernel density estimation model based on the optimal bandwidth and the K sets of feature data; a kernel density value calculation module 1040, used to perform grid division within the space formed by the K sets of feature data corresponding to each operating condition, and to calculate the kernel density value of each grid point based on the kernel density estimation model; and an abnormal cell identification module 1050, used to identify abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

[0076] Optionally, the feature extraction module 1020 is further configured to: convert the operating condition data corresponding to each operating condition into a first matrix; the first matrix is ​​composed of time as columns and cell operation data as rows; perform decentralization processing on the first matrix to obtain a decentralized first matrix; and perform feature extraction based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition.

[0077] Optionally, the process of decentralizing the first matrix includes: determining the average operating data of each cell at all times based on the first matrix; and subtracting the average operating data of the corresponding cell from the operating data of each cell at different times in the first matrix to obtain the decentralized first matrix.

[0078] Optionally, the step of extracting features based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition includes: determining the corresponding covariance matrix based on the decentralized first matrix; determining the correlation coefficient between the operating data of each cell at different times based on the covariance matrix; and taking the first K operating data of each cell in the decentralized first matrix as K sets of feature data in descending order of the correlation coefficient.

[0079] Optionally, the step of extracting features based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition includes: determining the time-frequency domain features corresponding to each operating data in the decentralized first matrix based on the decentralized first matrix; and taking the first K operating data of each cell in the decentralized first matrix as K sets of feature data in descending order of the time-frequency domain features.

[0080] Optionally, the kernel density estimation model construction module 1030 is further configured to: determine the standard deviation of all operating data in the battery to be tested; determine the adjustment coefficient based on the Gaussian kernel function; and multiply the standard deviation by the adjustment coefficient to obtain the optimal bandwidth.

[0081] Optionally, the kernel density estimation model construction module 1030 is further configured to: construct a kernel density estimation model based on the following expression: Where n is the number of cells in the lithium battery; Indicates the first Operating data of each battery cell Let represent the average operating data of n battery cells, and h be the optimal bandwidth. This represents the kernel density value.

[0082] Optionally, the abnormal cell identification module 1050 is further configured to: determine outlier regions based on the kernel density value of each grid point, and determine candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the outlier regions; and / or, determine the point corresponding to the mean of the kernel density values ​​of all grid points based on the kernel density value of each grid point, and determine candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the point corresponding to the mean of the kernel density values; determine whether the feature data of the candidate abnormal cell meets a preset condition; if the feature data of the candidate abnormal cell meets the preset condition, then determine that the candidate abnormal cell is an abnormal cell under the current operating condition.

[0083] Optionally, determining the outlier region based on the kernel density value of each grid point includes: determining the continuous region formed by grid points whose kernel density values ​​are less than a preset kernel density threshold as the outlier region, wherein the preset kernel density threshold is the kernel density value of all grid points that is located at a predetermined quantile after being arranged in ascending order.

[0084] Optionally, the abnormal cell identification module 1050 is further configured to: determine the kernel density estimate value corresponding to the feature data of each cell in the K sets of feature data according to the kernel density estimation model; determine candidate abnormal cells according to the relationship between the kernel density estimate value corresponding to the feature data of each cell in the K sets of feature data and a preset threshold; determine whether the feature data of the candidate abnormal cell meets a preset condition; if the feature data of the candidate abnormal cell meets the preset condition, then determine that the candidate abnormal cell is an abnormal cell under the current operating condition.

[0085] Optionally, the abnormal cell identification module 1050 is further configured to: if the cell is identified as an abnormal cell under operating conditions exceeding a preset number, then determine that the cell is an abnormal cell.

[0086] Optionally, the operating condition data acquisition module 1010 is further configured to: divide the operating data of each cell in the battery under test into multiple operating condition data for each cell according to different operating conditions.

[0087] As shown in Figure 11, an electronic device 1100 provided in this embodiment of the invention includes a memory 1110 and a processor 1120; the memory 1110 is used to store a computer program; the processor 1120 is used to implement the abnormal battery cell identification method as described above when the computer program is executed.

[0088] Alternatively, an electronic device 1100 includes a memory 1110 and a processor 1120 coupled to the memory 1110; the memory 1110 is configured to store a computer program; the processor 1120 is configured to perform the following operations when executing the computer program: acquiring operating condition data of each cell in a battery under test under different operating conditions, the operating conditions including at least a discharge condition and a charging condition; extracting features from the operating condition data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition; wherein each set of feature data includes feature data of each cell in the battery under test; determining the optimal bandwidth of the K sets of feature data corresponding to each operating condition, and constructing a kernel density estimation model based on the optimal bandwidth and the K sets of feature data; dividing the space formed by the K sets of feature data corresponding to each operating condition into a grid, and calculating the kernel density value of each grid point based on the kernel density estimation model; identifying abnormal cells in the battery under test according to the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

[0089] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the abnormal battery cell identification method described above.

[0090] Alternatively, a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following operations: acquiring operating condition data of each cell in the battery under test under different operating conditions, the operating conditions including at least discharge and charging conditions; extracting features from the operating condition data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition; wherein each set of feature data includes feature data of each cell in the battery under test; determining the optimal bandwidth of the K sets of feature data corresponding to each operating condition, and constructing a kernel density estimation model based on the optimal bandwidth and the K sets of feature data; dividing the space formed by the K sets of feature data corresponding to each operating condition into a grid, and calculating the kernel density value of each grid point based on the kernel density estimation model; identifying abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

[0091] The present invention will now be described an electronic device 1100 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. Electronic device 1100 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 1100 can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0092] Electronic device 1100 includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) or a computer program loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0093] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. In this application, 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 units can be selected to achieve the purpose of the embodiments of the present invention according to actual needs. Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units.

[0094] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for identifying abnormal battery cells, characterized in that, include: Acquire operating condition data of each cell in the battery under test under different operating conditions, wherein the operating conditions include at least discharge conditions and charging conditions; Feature extraction is performed on the operating condition data corresponding to each operating condition to obtain K sets of feature data for each operating condition, where K is an integer greater than or equal to 2; wherein, each set of feature data includes the feature data of each cell in the battery under test; the optimal bandwidth of the K sets of feature data corresponding to each operating condition is determined, and a kernel density estimation model is constructed based on the optimal bandwidth and the K sets of feature data; a grid is divided within the space formed by the K sets of feature data corresponding to each operating condition, and the kernel density value of each grid point is calculated based on the kernel density estimation model; abnormal cells in the battery under test are identified according to the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

2. The abnormal cell identification method according to claim 1, characterized in that, The step of extracting features from the operating data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition includes: converting the operating data corresponding to each operating condition into a first matrix; the first matrix having time as columns and cell operation data as rows; performing decentralization processing on the first matrix to obtain a decentralized first matrix; and performing feature extraction based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition.

3. The abnormal cell identification method according to claim 2, characterized in that, The process of decentralizing the first matrix includes: determining the average operating data of each cell under the corresponding operating condition based on the first matrix; and subtracting the average operating data of the corresponding cell from the operating data of each cell at different times in the first matrix to obtain the decentralized first matrix.

4. The abnormal cell identification method according to claim 2, characterized in that, The step of extracting features based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition includes: determining the corresponding covariance matrix based on the decentralized first matrix; determining the correlation coefficient between the operating data of each cell at different times based on the covariance matrix; and taking the first K operating data of each cell in the decentralized first matrix as the K sets of feature data in descending order of the correlation coefficient.

5. The abnormal cell identification method according to claim 2, characterized in that, The step of extracting features based on the decentralized first matrix to obtain K sets of feature data corresponding to each operating condition includes: determining the time-frequency domain features corresponding to each operating data in the decentralized first matrix based on the decentralized first matrix; and taking the first K operating data of each cell in the decentralized first matrix as the K sets of feature data in descending order of the time-frequency domain features.

6. The abnormal cell identification method according to claim 1, characterized in that, Determining the optimal bandwidth for the K sets of feature data corresponding to each operating condition includes: determining the standard deviation of all operating data in the battery to be tested; determining the adjustment coefficient based on the Gaussian kernel function; and multiplying the standard deviation by the adjustment coefficient to obtain the optimal bandwidth.

7. The abnormal cell identification method according to claim 1, characterized in that, The step of constructing a kernel density estimation model based on the optimal bandwidth and the K sets of feature data includes: constructing the kernel density estimation model according to the following expression: Where n is the number of cells in the battery to be tested. Indicates the first Operating data of each battery cell Let represent the average operating data of n battery cells, and h be the optimal bandwidth. This represents the kernel density value.

8. The abnormal cell identification method according to claim 1, characterized in that, Identifying abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point includes: determining outlier regions based on the kernel density value of each grid point, and determining candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the outlier regions; and / or determining the point corresponding to the mean of the kernel density values ​​of all grid points based on the kernel density value of each grid point, and determining candidate abnormal cells based on the positional relationship between the feature data of each cell in the K sets of feature data and the point corresponding to the mean of the kernel density values; determining whether the feature data of the candidate abnormal cells meets preset conditions; if the feature data of the candidate abnormal cells meets the preset conditions, then determining that the candidate abnormal cells are abnormal cells under the current operating conditions.

9. The abnormal cell identification method according to claim 8, characterized in that, The step of determining outlier regions based on the kernel density value of each grid point includes: determining the continuous region formed by grid points whose kernel density values ​​are less than a preset kernel density threshold as outlier regions, wherein the preset kernel density threshold is the kernel density value of all grid points that is located at a predetermined quantile after being arranged in ascending order.

10. The abnormal cell identification method according to claim 8, characterized in that, Also includes: Based on the kernel density estimation model, the kernel density estimate corresponding to the feature data of each cell in the K sets of feature data is determined; based on the relationship between the kernel density estimate corresponding to the feature data of each cell in the K sets of feature data and a preset threshold, candidate abnormal cells are determined; it is determined whether the feature data of the candidate abnormal cells meets the preset conditions; if the feature data of the candidate abnormal cells meets the preset conditions, the candidate abnormal cells are determined to be abnormal cells under the current operating conditions.

11. The abnormal cell identification method according to claim 8 or 10, characterized in that, Also includes: If the battery cell is identified as an abnormal battery cell under operating conditions exceeding a preset number, then the battery cell is determined to be an abnormal battery cell.

12. The abnormal cell identification method according to claim 1, characterized in that, The step of obtaining the operating condition data of each cell in the battery under test under different operating conditions includes: dividing the operating data of each cell in the battery under test into multiple operating condition data of each cell according to different operating conditions.

13. An abnormal battery cell identification device, characterized in that, include: The operating condition data acquisition module is used to acquire the operating condition data of each cell in the battery under test under different operating conditions, wherein the operating conditions include at least the discharge operating condition and the charging operating condition. The feature extraction module is used to extract features from the operating condition data corresponding to each operating condition to obtain K sets of feature data corresponding to each operating condition, where K is an integer greater than or equal to 2; wherein, each set of feature data includes the feature data of each cell in the battery under test; the kernel density estimation model construction module is used to determine the optimal bandwidth of the K sets of feature data corresponding to each operating condition, and construct a kernel density estimation model based on the optimal bandwidth and the K sets of feature data; the kernel density value calculation module is used to perform grid division within the space formed by the K sets of feature data corresponding to each operating condition, and calculate the kernel density value of each grid point based on the kernel density estimation model; the abnormal cell identification module is used to identify abnormal cells in the battery under test based on the feature data of each cell in the K sets of feature data and the kernel density value of each grid point.

Images