Battery pack anomaly detection method, device and equipment and storage medium

Abstract

The invention provides a battery pack anomaly detection method, device and equipment and a storage medium, and the method comprises the steps: obtaining multi-frame charging data of each battery unit in a battery pack in the same preset time period, the charging data being time series data, and the charging data at least comprising a timestamp, a voltage value and a resistance value; for each frame of charging data of each battery unit, calculating a voltage characteristic value and a resistance characteristic value according to the charging data; for each battery unit, forming a resistance characteristic matrix by the corresponding resistance characteristic values, and forming a voltage characteristic matrix according to the corresponding voltage characteristic values; and judging whether the battery unit is abnormal or not according to the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix and the resistance characteristic matrix. The voltage and resistance characteristic matrix is constructed by comprehensively utilizing the voltage and resistance data of the battery unit under the multi-frame time sequence, so that the dynamic characteristic change of the battery in the continuous charging and discharging process can be more comprehensively captured, and the accurate identification of the battery abnormality is realized.

Description

Technical Field

[0001] This invention belongs to the field of battery management technology, and in particular relates to a method, device, equipment and storage medium for detecting abnormalities in battery packs. Background Technology

[0002] With the rapid development of electric vehicles and energy storage systems, the safety and reliability of battery packs are receiving increasing attention. In actual operation, individual battery cells within a battery pack may exhibit abnormalities due to manufacturing differences or aging. Failure to detect these abnormalities in a timely manner can affect overall performance and even lead to safety accidents. Currently, common detection methods are mostly based on the analysis of single frames or single types of data, which cannot comprehensively reflect the dynamic characteristics of battery cells during continuous charging and discharging, resulting in limited accuracy in anomaly identification. Furthermore, the joint features in multi-frame voltage and resistance time-series data have not been fully explored. Therefore, there is an urgent need for a method that can comprehensively utilize multi-dimensional time-series features to achieve more accurate battery cell anomaly detection. Summary of the Invention

[0003] In view of this, the present invention provides a method, apparatus, device and storage medium for detecting abnormalities in battery packs, so as to improve the accuracy of detecting abnormalities in battery cells.

[0004] A first aspect of the present invention provides a battery pack anomaly detection method, comprising:

[0005] Acquire multiple frames of charging data for each battery cell in the battery pack within the same preset time period. The charging data is time-series data and includes at least timestamps, voltage values, and resistance values.

[0006] For each frame of charging data of each battery cell, voltage characteristic value and resistance characteristic value are calculated based on the charging data;

[0007] For each battery cell, a resistance feature matrix is ​​formed by combining the corresponding resistance feature values, and a voltage feature matrix is ​​formed by combining the corresponding voltage feature values.

[0008] The battery cell is determined to be abnormal based on the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix, and the resistance characteristic matrix.

[0009] A second aspect of the present invention provides a battery pack anomaly detection device, comprising:

[0010] The data acquisition module is used to acquire multiple frames of charging data of each battery cell in the battery pack within the same preset time period. The charging data is time-series data and includes at least timestamps, voltage values, and resistance values.

[0011] The feature value calculation module is used to calculate voltage feature value and resistance feature value based on each frame of charging data for each of the battery cells;

[0012] A matrix construction module is used to form a resistance feature matrix based on the corresponding resistance feature values ​​for each battery cell, and to form a voltage feature matrix based on the corresponding voltage feature values.

[0013] An anomaly detection module is used to determine whether the battery cell is abnormal based on the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix, and the resistance characteristic matrix.

[0014] A third aspect of the present invention 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, when executing the computer program, implements the battery pack anomaly detection method as described in the first aspect above.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the battery pack anomaly detection method as described in the first aspect above.

[0016] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0017] By comprehensively utilizing voltage and resistance data of battery cells across multiple time frames, a voltage and resistance feature matrix is ​​constructed, enabling accurate identification of battery anomalies. Compared to traditional single-frame or single-parameter detection methods, this approach can more comprehensively capture the dynamic characteristic changes of the battery during continuous charging and discharging, effectively identifying early anomalies and potential faults. Through multi-dimensional feature joint analysis, the accuracy and reliability of detection are significantly improved, providing strong support for the safe operation of battery packs. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of a battery pack anomaly detection method provided in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of a battery cell anomaly detection process provided in an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of another battery cell anomaly detection process provided in an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of a battery pack anomaly detection method provided in an embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of a battery pack anomaly detection device provided in an embodiment of the present invention;

[0024] Figure 6 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the present invention. However, those skilled in the art will recognize that the present application may be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted to avoid unnecessary detail that could obscure the description of the present application.

[0026] The technical solution of the present invention will be illustrated below through specific embodiments.

[0027] Reference Figure 1 The diagram illustrates a battery pack anomaly detection method provided by an embodiment of the present invention, which may specifically include the following steps:

[0028] S101. Obtain multiple frames of charging data for each battery cell in the battery pack within the same preset time period. The charging data is time-series data and includes at least timestamps, voltage values, and resistance values.

[0029] The charging data of all battery cells were collected under the same operating conditions (such as the same ambient temperature and the same charging and discharging current), which eliminated the interference of external variables on the analysis results and made the data between different battery cells comparable.

[0030] Charging data includes at least a timestamp, voltage value, and resistance value. A single sample (single frame of charging data) can only reflect the state at a certain instant and cannot capture the behavioral trend of the battery during the charging and discharging process. Multiple frames of charging data constitute a time series, which can reveal the changing patterns of voltage and resistance over time.

[0031] Voltage and resistance are core parameters reflecting the health of a battery. Abnormal voltage may directly manifest as overcharging or undercharging, while changes in resistance (usually referring to internal resistance) are sensitive indicators of battery aging, internal short circuits, or loose connections. Therefore, combining the two can provide more comprehensive diagnostic information.

[0032] S102. For each frame of charging data of each battery cell, calculate the voltage characteristic value and resistance characteristic value based on the charging data.

[0033] This step involves preliminary information processing of the raw data. On the one hand, the raw charging data is usually unstructured, making it difficult to directly compare data between battery cells using mathematical operations. On the other hand, the raw charging data may contain noise, resulting in poor performance when used directly.

[0034] Each frame of charging data may contain voltage and resistance characteristic values ​​within a time window. Possible forms of these characteristic values ​​include:

[0035] Statistical characteristics: For example, this could be calculating the average, variance, standard deviation, etc. of voltage / resistance within a time window (which can be considered as one or more frames).

[0036] Rate of change characteristics: For example, calculating the slope of voltage or resistance change between adjacent frames.

[0037] Transformation features: For example, frequency domain features are obtained through Fourier transform.

[0038] By calculating eigenvalues, the behavior pattern of the battery during the time period can be characterized more stably and effectively, preparing for subsequent matrix construction and comprehensive judgment.

[0039] S103. For each battery cell, the corresponding resistance characteristic values ​​are used to form a resistance characteristic matrix, and the corresponding voltage characteristic values ​​are used to form a voltage characteristic matrix.

[0040] Suppose N frames of data were collected from a battery cell, and a voltage characteristic value and a resistance characteristic value were calculated for each frame. Then:

[0041] The voltage characteristic matrix can be an N×1 column vector [V_f1,V_f2,...,V_fn]. T .

[0042] The resistance characteristic matrix is ​​also an N×1 column vector [R_f1,R_f2,...,R_fn]. T .

[0043] This step formally introduces the time dimension into the model, transforming isolated numerical values ​​into curves or trajectories showing how features change over time. This clearly demonstrates the evolution trend of the battery cell's voltage and resistance characteristics throughout the entire charging process. For example, the voltage characteristic matrix of a healthy battery should show a steady upward trend, while the matrix of an abnormal battery may exhibit drastic fluctuations or abnormal drops.

[0044] S104. Determine whether the battery cell is abnormal based on the voltage characteristic value, resistance characteristic value, voltage characteristic matrix, and resistance characteristic matrix.

[0045] In Example A, the voltage / resistance characteristics of an individual battery cell are compared to a preset safety threshold. For example, at some point, the resistance characteristic suddenly exceeds the normal range.

[0046] In Example B, at the same time, the characteristic values ​​of all battery cells are compared. If the characteristic value of a battery cell deviates significantly from that of most other cells in the group, it may be determined as an anomaly.

[0047] In Example C, we analyze whether the trend of the voltage / resistance characteristic matrix of the battery cell itself is normal. For example, during the constant current charging phase, the voltage should rise steadily; if its voltage characteristic matrix shows a plateau or a drop, it is abnormal.

[0048] Specifically, traditional methods rely on single frames or single types of data, making it difficult to capture the characteristic changes of batteries during dynamic charging and discharging, resulting in insensitivity to early anomalies and potential faults. This solution acquires multiple frames of time-series charging data and constructs feature matrices for voltage and resistance respectively, achieving a shift from static single-point judgment to dynamic trend analysis.

[0049] This invention provides a battery pack anomaly detection method. It acquires multiple frames of charging data from each battery cell within the same preset time period. The charging data is time-series data, including at least a timestamp, voltage value, and resistance value. For each frame of charging data for each battery cell, voltage and resistance feature values ​​are calculated. For each battery cell, a resistance feature matrix is ​​formed from the corresponding resistance feature values, and a voltage feature matrix is ​​formed from the corresponding voltage feature values. The method determines whether a battery cell is abnormal based on the voltage, resistance, voltage, and resistance feature matrices. By comprehensively utilizing the voltage and resistance data of the battery cells across multiple time frames, a voltage and resistance feature matrix is ​​constructed, enabling accurate identification of battery anomalies. Compared to traditional single-frame or single-parameter detection methods, this method can more comprehensively capture the dynamic characteristic changes of the battery during continuous charging and discharging, effectively identifying early anomalies and potential faults. Through multi-dimensional feature joint analysis, the accuracy and reliability of detection are significantly improved, providing strong protection for the safe operation of the battery pack.

[0050] In an optional embodiment, for each frame of charging data for each battery cell, a voltage characteristic value is calculated based on the charging data, including:

[0051] The average voltage and standard deviation of voltage are calculated based on the voltage values ​​in the charging data of all battery cells to obtain the comprehensive average voltage and comprehensive standard deviation of voltage.

[0052] For each frame of charging data for each battery cell, determine the voltage value in the charging data;

[0053] By combining the comprehensive voltage mean and the comprehensive voltage standard deviation, the degree of deviation of the current voltage value from the comprehensive voltage mean is determined, and the voltage characteristic value corresponding to the current voltage value is obtained.

[0054] The formula for calculating the voltage characteristic value is: feature1 = (V - Vmean) / (3 × Vstd);

[0055] Where feature1 represents the voltage feature value corresponding to the current frame charging data of the battery cell, V represents the voltage value of the current frame charging data, Vstd is the comprehensive voltage standard deviation, and Vmean is the comprehensive voltage mean.

[0056] The advantage of calculating voltage characteristic values ​​in this way is that it eliminates the need to set a fixed voltage anomaly threshold. This is because the normal voltage range of a battery varies with its state of charge (SOC), temperature, and aging. The benchmarks (mean and standard deviation) used in this method are calculated in real time, automatically adapting to the current operating conditions of the battery pack, greatly improving the adaptability of the detection. It can immediately identify abnormal battery cells with significantly high or low voltage. For example, a battery cell with an internal short circuit will have a significantly lower voltage than its peers, and its voltage characteristic value (Z-Score) will be a large negative number, easily detected by the algorithm.

[0057] Furthermore, the obtained voltage characteristic value is a standardized, dimensionless numerical value that can be used to construct the voltage characteristic matrix in S103. This matrix will depict how the "relative position" of each battery cell relative to the entire group of cells changes over time, which is crucial for detecting faulty battery cells.

[0058] In an optional embodiment, before calculating the voltage mean and voltage standard deviation based on the voltage values ​​in the charging data of all battery cells to obtain the comprehensive voltage mean and comprehensive voltage standard deviation, the method further includes:

[0059] The charging data of each battery cell with the same timestamp is taken as a frame of raw data. For each frame of raw data, the voltage value is used as the filtering object, and the raw data is cyclically filtered according to the preset confidence interval and the preset number of cycles to remove charging data in the raw data whose voltage value deviates from the confidence interval.

[0060] Optionally, the preset number of cycles is 3, the purpose of which is to exclude as many abnormal cells as possible, leaving only the data of normal cells for calculating the mean and standard deviation.

[0061] In an optional embodiment, the resistance parameters, i.e. resistance characteristic values, of the second-order equivalent circuit model are obtained by using the forgetting factor recursive least squares method (FFRLS).

[0062] In an optional embodiment, refer to Figure 2 This diagram illustrates a battery cell anomaly detection process according to an embodiment of the present invention. The process determines whether a battery cell is abnormal based on voltage characteristic values, resistance characteristic values, a voltage characteristic matrix, and a resistance characteristic matrix, including:

[0063] S201. Perform time-frequency domain transformation on the voltage characteristic matrix to obtain the time-frequency characteristic matrix.

[0064] Time-frequency domain transformation methods include Short-Time Fourier Transform (STFT) and Wavelet Transform. Time-frequency domain transformation converts a time-series signal (i.e., the voltage characteristic matrix) into a two-dimensional matrix containing both time and frequency information. The horizontal axis of this matrix represents time, and the vertical axis represents frequency. The values ​​of the matrix elements represent the intensity of a specific frequency component at a specific time. This is because many early battery faults (such as micro-short circuits, lithium plating, and loose connections) may not be obvious on the time-domain voltage curve, but they will produce abnormal fluctuations or harmonics at specific frequency components. Therefore, time-frequency analysis can capture these subtle characteristics.

[0065] It should be noted that time-frequency domain transformation is performed only on the voltage characteristic matrix, but not on the resistance characteristic matrix. The core reason is that voltage is a dynamically changing process signal, while resistance is a relatively stable state parameter.

[0066] Specifically, when a battery is operating, its terminal voltage is a dynamic signal that changes dramatically over time. Especially during charging, discharging, and load changes, the voltage exhibits complex transient responses (such as sudden drops, recovery, and ripple). These changes contain a wealth of information about the battery's health status. Performing time-frequency domain transformation on the voltage (such as Short-Time Fourier Transform (STFT) or wavelet transform) to convert the time-domain signal to the frequency domain allows us to clearly see which frequency components are at play. Therefore, performing time-frequency transformation on the voltage is to focus on studying the subtle degradation-related features in its dynamic behavior that are imperceptible to the naked eye.

[0067] The internal resistance of a battery is a relatively stable parameter that changes slowly. It is typically related to the state of charge, state of health, and temperature, but has little to do with dynamic operating processes on the order of seconds or even milliseconds. Even with high-frequency continuous measurement of internal resistance, the resulting internal resistance-time series changes very slowly and smoothly, containing almost no valuable high-frequency components. Time-frequency analysis of a smooth, slowly changing signal yields extremely low information gain in the frequency domain. Therefore, resistance itself is an effective state indicator; its numerical value or slow trend is sufficient, and time-frequency analysis is unnecessary.

[0068] S202. The time-frequency feature matrix and the resistance feature matrix are fused, dimensionally reduced, and normalized to obtain a fused feature matrix with the same size as the time-frequency feature matrix and the resistance feature matrix.

[0069] Fusion refers to merging two information sources representing different physical meanings (frequency domain characteristics of voltage and time domain characteristics of resistance). This involves fusing the time-frequency feature matrix and the resistance feature matrix. Specifically, this can be a simple matrix concatenation or a more complex tensor-based fusion, with the aim of creating a feature set containing all the original information.

[0070] Dimensionality reduction: The dimensionality of the fused features may be very high, containing a lot of redundancy and noise. Dimensionality reduction aims to retain the most critical information while reducing the amount of data, thereby improving subsequent computational efficiency and model performance. Common dimensionality reduction methods include Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), or Autoencoders.

[0071] Normalization aims to unify features of different dimensions and magnitudes (such as voltage frequency domain energy and resistance value) to the same scale (such as the [0,1] interval) to avoid some features from excessively affecting the entire model due to excessively large values.

[0072] After the above three steps, a standardized fusion feature matrix is ​​obtained, which includes both the deep frequency domain characteristics of voltage and the time domain characteristics of resistance, and removes redundancy and noise.

[0073] S203. Configure each element in the fusion feature matrix as the fusion feature value of each battery cell.

[0074] This step flattens the matrix obtained in the previous step, so that each value (or each group of values) in the matrix can clearly correspond to a specific battery cell. For example, assuming the fusion feature matrix is ​​K×L dimensional, then each battery cell will eventually obtain K×L fusion feature values, which together constitute a high-dimensional, digital description of the overall state of the battery cell.

[0075] S204. Determine whether the battery cell is abnormal based on the voltage characteristic value, resistance characteristic value, and fusion characteristic value.

[0076] The judgment logic in this step is hierarchical, weighted, and involves the fusion of multiple pieces of evidence:

[0077] From the perspective of eigenvalues, voltage eigenvalues ​​reflect the instantaneous deviation of real-time voltage from the group average level (such as the previously calculated Z-score), while resistance eigenvalues ​​reflect the absolute magnitude or relative deviation of real-time resistance, and can quickly capture obvious and severe anomalies. For example, a sudden drop in voltage or a sudden surge in resistance.

[0078] From the perspective of fused eigenvalues, this reflects deep and complex patterns of voltage and resistance co-variation within the battery cell over the entire time period. It can be used to identify potential, progressive, and complex anomalies. For example:

[0079] Early aging: Energy changes slowly in specific frequency bands of the voltage spectrum, while resistance shows a slight upward trend. Individual features are not obvious, but fused features can capture this coordinated change.

[0080] Internal micro-short circuits: They may not show significant changes in voltage in the time domain, but they will exhibit specific noise patterns in the frequency domain, which are captured by the fusion feature along with the minute changes in resistance.

[0081] In an optional embodiment, refer to Figure 3 This diagram illustrates another battery cell anomaly detection process provided by an embodiment of the present invention. It determines whether a battery cell is abnormal based on voltage characteristic values, resistance characteristic values, and fusion characteristic values, including:

[0082] S301. For the current battery cell, determine the first anomaly score of the battery cell based on the dispersion of the corresponding voltage characteristic value.

[0083] Voltage characteristic values ​​refer to the characteristic values ​​(i.e., those values ​​that constitute its voltage characteristic matrix) calculated at multiple time points within a preset time period for the battery cell; it is a time series.

[0084] The degree of dispersion is usually calculated using standard deviation or coefficient of variation (standard deviation / mean). The greater the degree of dispersion, the more drastic the voltage fluctuation of the battery cell, and the more unstable it is.

[0085] The first anomaly score primarily captures the instability and consistency deviations of battery cells. A healthy battery cell should exhibit relatively stable voltage changes. High dispersion in its voltage characteristic value sequence indicates a potentially erratic state, a significant sign of anomaly. Corresponding faults include intermittent poor contact and unstable internal chemical reactions.

[0086] In an optional embodiment, determining a first anomaly score for the battery cell based on the dispersion of the corresponding voltage characteristic value includes:

[0087] Based on a preset clustering range and density, the voltage characteristic values ​​corresponding to the battery cells are clustered to obtain multiple clusters and discrete points; the ratio of the number of discrete points to the total number of corresponding voltage characteristic values ​​is calculated; when the ratio is greater than a preset ratio, the first anomaly score is determined as the first score; when the ratio is not greater than the preset ratio, the first anomaly score is determined as the second score; the first score has an additive effect on the comprehensive anomaly score.

[0088] When performing clustering, clustering algorithms can be used, such as DBSCAN. The core parameters of DBSCAN are the neighborhood range and the minimum number of neighbors required for the core point, which correspond exactly to the preset clustering range and density.

[0089] Clusters represent stable, normal states that repeatedly occur during battery cell operation. For example, one cluster might correspond to a constant current charging state, while another cluster might correspond to a constant voltage charging state. Discrete points, on the other hand, represent abnormal moments that cannot be classified into any stable state. These points are caused by sudden and drastic voltage fluctuations due to internal short circuits, external interference, unstable connections, or other reasons.

[0090] Discrete point ratio = Number of discrete points / Total number of voltage characteristic values;

[0091] The discrete point ratio represents the proportion of time within a given time period that the battery cell is in an abnormal transient state. Example: If 100 frames of data are collected and clustering reveals 8 discrete points, then the ratio is 0.08. This means that the battery cell behaves abnormally 8% of the time.

[0092] If the corresponding ratio is greater than the preset ratio, it indicates that the unit has poor stability and is abnormally frequent. In this case, the score is the first score.

[0093] If the corresponding ratio is less than or equal to the preset ratio, it indicates that the unit has good stability and a low or no abnormality rate. In this case, the score is the second score.

[0094] The preset ratio can be set to 5%.

[0095] The first score has a positive effect on the overall anomaly score. Optionally, the first score is 1 and the second score is 0. When the first anomaly score is 1, it has a positive effect on the overall anomaly score. When the first anomaly score is 0, it has no effect on the overall anomaly score.

[0096] S302. Calculate the second anomaly score of the battery cell based on the voltage characteristic value, resistance characteristic value and fusion characteristic value.

[0097] The second anomaly score is the most informative and most likely to be applied using a machine learning model. Possible calculation methods include:

[0098] Weighted summation: Multiply the three feature values ​​by their respective weights and then sum them. The weights may come from experience or model training.

[0099] Machine learning model: Input these three feature values ​​into a simple classifier (such as a single-class SVM, Isolation Forest) or regression model, and output an anomaly probability or score.

[0100] The second anomaly score employs a multi-evidence joint analysis, comprehensively considering both instantaneous voltage / resistance deviations and deep dynamic patterns represented by fused characteristics. Anomalies in any dimension will increase this score. Corresponding faults: comprehensive faults, especially potential faults requiring coordinated judgment of multiple parameters, such as early aging.

[0101] In an optional embodiment, calculating the second anomaly score of the battery cell based on the voltage characteristic value, the resistance characteristic value, and the fused characteristic value includes:

[0102] Step a: Calculate the mean of voltage characteristic values, the mean of resistance characteristic values, and the mean of fused characteristic values ​​for all battery cells to obtain the mean of reference voltage characteristic, the mean of reference resistance characteristic, and the mean of reference fused characteristic.

[0103] Step b: Calculate the mean of the voltage characteristic value, the mean of the resistance characteristic value, and the mean of the fused characteristic value corresponding to the current battery cell to obtain the mean of the cell voltage characteristic, the mean of the cell resistance characteristic, and the mean of the cell fused characteristic.

[0104] Step c: Using the reference voltage characteristic mean, reference resistance characteristic mean, and reference fusion characteristic mean as reference points, calculate the deviation of the cell voltage characteristic mean, cell resistance characteristic mean, and fusion characteristic mean from the reference points to obtain the second anomaly score of the current battery cell.

[0105] In step a, a group baseline is established. The mean of the reference voltage characteristics represents the average normal level of the voltage characteristics of all battery cells in the battery pack. The mean of the reference resistance characteristics represents the average normal level of the resistance characteristics of all battery cells in the battery pack. The mean of the reference fusion characteristics represents the average normal mode of the deep dynamic characteristics of the battery pack.

[0106] This calculation covers all monitored battery cells and depicts the statistically normal center point of the entire battery pack. Mapping this normal center point to the feature vector space of the battery cells, these three benchmark means together constitute a multi-dimensional reference benchmark, providing an objective standard for judging whether any individual cell is abnormal.

[0107] Specifically, the reference voltage characteristic F1_mean, the reference resistance characteristic mean F2_mean, and the reference fusion characteristic mean F3_mean are set.

[0108] In step b, the average value of the cell voltage characteristic F1, the average value of the cell resistance characteristic F2, and the average value of the cell fusion characteristic F3 are set. Mapping these three features to the feature vector space of the battery cell yields a specific point.

[0109] In step c, the differences between individual battery cells and the group are quantified. Specifically, the deviation of each dimension can be calculated separately, or the combined deviation of multiple dimensions can be calculated.

[0110] In an optional example, the second anomaly score is calculated using the following formula:

[0111] S2 = sqrt{(F1 - F1_mean)} 2 +(F2-F2_mean) 2 +(F3-F3_mean) 2}

[0112] Treating the reference point and the cell point as two points in three-dimensional space, the Euclidean distance between them is calculated as an anomaly score; the greater the distance, the higher the degree of anomaly. This score quantifies the deviation of the battery cell from the center point of the normal group in terms of overall performance. A battery cell with a very high second anomaly score means that it differs significantly from other battery cells in terms of the overall average level of voltage, resistance, and dynamic characteristics.

[0113] Because it uses the mean for calculation, the second anomaly score is insensitive to instantaneous noise and random fluctuations in the data, making it more effective in revealing persistent and trending anomalies. It also considers voltage, resistance, and fusion characteristics, avoiding the limitations of judging based on a single parameter. Furthermore, the distance- or deviation-based calculation method is simple in principle, highly efficient, and the results are easy to interpret.

[0114] S303. Calculate the mean value of the corresponding voltage characteristic value to obtain the third anomaly score of the battery cell.

[0115] The third anomaly score is the average of the voltage characteristic value sequence. This score reflects the average voltage level of the battery cell over the entire time period. If the average voltage characteristic value of a battery cell is consistently low or consistently high, it may be an abnormal cell. Corresponding faults: capacity degradation, inconsistency deterioration.

[0116] S304. Calculate a comprehensive anomaly score based on the first anomaly score, the second anomaly score, and the third anomaly score. The comprehensive anomaly score is positively correlated with the degree of anomaly of the battery cell.

[0117] In an optional example, the comprehensive anomaly score is the weighted sum of the first, second, and third anomaly scores, as shown in the formula:

[0118] The comprehensive abnormality score S = w1×S1 + w2×S2 + w3×S3;

[0119] Here, S1, S2, and S3 represent the first, second, and third anomaly scores, respectively, while w1, w2, and w3 are weighting coefficients. Their sum is usually 1. The weighting reflects the degree of importance attached to different anomaly dimensions and can be optimized using historical data. The higher each sub-score, the more severe the anomaly in that dimension, and the higher the final comprehensive score.

[0120] The first anomaly score focuses on forming a comprehensive, multi-level battery anomaly diagnostic system that considers both instantaneous risks and overall health.

[0121] In another optional example, a comprehensive anomaly score is calculated based on the first anomaly score, the second anomaly score, and the third anomaly score, including:

[0122] Obtain the preset weights corresponding to the first, second, and third anomaly scores; calculate the comprehensive anomaly score based on the first, second, and third anomaly scores and their corresponding preset weights.

[0123] That is, different weights are used based on the accuracy of the calculation of each type of anomaly score; higher accuracy is given a larger weight, and vice versa.

[0124] S305. Determine whether the battery cell is abnormal based on the comprehensive anomaly score and the preset score threshold.

[0125] The preset scoring threshold can be obtained based on historical experience. If the overall abnormal score is higher than the scoring threshold, it indicates that the battery cell is abnormal.

[0126] The battery cell anomaly detection method of this embodiment has the following advantages:

[0127] First, multi-dimensional and three-dimensional diagnosis can more accurately and stably detect abnormal battery cells. The first abnormality score tends to focus on stability, the second abnormality score tends to focus on comprehensiveness, and the third abnormality score tends to focus on persistence. The evaluation of these three different and complementary dimensions can cover various manifestations of abnormalities.

[0128] Second, it transforms ambiguous anomalies into precise scores, allowing administrators to flexibly adjust thresholds and weights according to actual needs (such as in high-security scenarios) to achieve configurable detection sensitivity; at the same time, it records, analyzes, classifies alarms, and visualizes them in the upstream system.

[0129] Third, avoid false alarms caused by occasional fluctuations in a single indicator. For example, even if the instantaneous voltage of a battery cell is somewhat discrete (the first score is slightly higher), as long as its overall condition (second score) and average level (third score) are normal, it may still be judged as normal in the end.

[0130] To clearly illustrate the battery pack anomaly detection method of this solution, the following examples and... Figure 4 To explain, Figure 4 This is a schematic diagram of a battery pack anomaly detection method, as shown below. Figure 4 As shown, the battery pack anomaly detection methods include:

[0131] First, the resistance and voltage values ​​of the battery cells are obtained; then, they are converted to obtain resistance and voltage characteristic values, which are then used to construct resistance and voltage characteristic matrices; finally, the voltage characteristic matrix is ​​transformed in the time-frequency domain to obtain the time-frequency characteristic matrix.

[0132] Calculate the voltage characteristic value F1 and the unit resistance characteristic value F2 based on the resistance characteristic value and the voltage characteristic value, respectively.

[0133] Then, the resistance feature matrix and time-frequency feature matrix of each battery cell are fused, dimensionally reduced and normalized to obtain the fused feature moments, which include the cell fused feature value F3 of each battery cell;

[0134] Calculate the mean of the voltage characteristic value, the mean of the resistance characteristic value, and the mean of the fused characteristic value corresponding to all the battery cells to obtain the reference voltage characteristic mean F1_mean, the reference resistance characteristic mean F2_mean, and the reference fused characteristic mean F3_mean;

[0135] Calculate the anomaly score 2 based on F1, F2, F3, F1_mean, F2_mean, and F3_mean;

[0136] The voltage characteristic values ​​were calculated using statistical algorithms and DBSCAN respectively, resulting in anomaly scores 1 and 3.

[0137] Finally, a comprehensive anomaly score is calculated based on anomaly score 2, anomaly score 1, and anomaly score 3.

[0138] By calculating the comprehensive anomaly score using the above method, early warning of battery pack failures is achieved, improving the generalization ability and robustness of the method.

[0139] It should be noted that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0140] Reference Figure 5 The diagram shows a schematic of a battery pack anomaly detection device provided by an embodiment of the present invention, which may specifically include the following modules:

[0141] The data acquisition module 501 is used to acquire multiple frames of charging data of each battery cell in the battery pack within the same preset time period. The charging data is time-series data and includes at least timestamps, voltage values ​​and resistance values.

[0142] The feature value calculation module 502 is used to calculate voltage feature value and resistance feature value based on each frame of charging data for each of the battery cells;

[0143] The matrix construction module 503 is used to form a resistance feature matrix based on the corresponding resistance feature values ​​for each battery cell, and to form a voltage feature matrix based on the corresponding voltage feature values.

[0144] The anomaly detection module 504 is used to determine whether the battery cell is abnormal based on the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix, and the resistance characteristic matrix.

[0145] Optionally, the feature value calculation module 502 is used for:

[0146] The average voltage and standard deviation of voltage are calculated based on the voltage values ​​in the charging data of all the battery cells to obtain the comprehensive average voltage and comprehensive standard deviation of voltage.

[0147] For each frame of charging data of each of the battery cells, determine the voltage value in the charging data;

[0148] By combining the comprehensive voltage mean and the comprehensive voltage standard deviation, the degree of deviation of the current voltage value from the comprehensive voltage mean is determined, and the voltage characteristic value corresponding to the current voltage value is obtained.

[0149] Optionally, the feature value calculation module 502 is further configured to:

[0150] The charging data of each battery cell with the same timestamp is used as a frame of raw data;

[0151] For each frame of the original data, the original data is filtered by voltage value, according to a preset confidence interval and a preset number of cycles, in order to remove the charging data in the original data whose voltage value deviates from the confidence interval.

[0152] Optionally, the anomaly detection module 504 includes:

[0153] The time-frequency feature matrix construction submodule is used to perform time-frequency domain transformation on the voltage feature matrix to obtain the time-frequency feature matrix;

[0154] The feature matrix fusion construction submodule is used to fuse, reduce the dimension and normalize the time-frequency feature matrix and the resistance feature matrix to obtain a fused feature matrix with the same size as the time-frequency feature matrix and the resistance feature matrix;

[0155] The fusion feature value calculation submodule is used to configure each element in the fusion feature matrix as the fusion feature value of each battery cell.

[0156] The anomaly detection submodule is used to determine whether the battery cell is abnormal based on the voltage characteristic value, the resistance characteristic value, and the fused characteristic value.

[0157] Optionally, the anomaly detection submodule includes:

[0158] The first unit is used to determine a first anomaly score of the battery cell based on the degree of dispersion of the corresponding voltage characteristic value for the current battery cell.

[0159] The second unit is used to calculate a second anomaly score of the battery cell based on the voltage characteristic value, the resistance characteristic value and the fusion characteristic value;

[0160] The third unit is used to calculate the mean value of the corresponding voltage characteristic values ​​to obtain the third anomaly score of the battery cell;

[0161] The fourth unit is used to calculate a comprehensive anomaly score based on the first anomaly score, the second anomaly score, and the third anomaly score, wherein the comprehensive anomaly score is positively correlated with the degree of anomaly of the battery unit;

[0162] The fifth unit is used to determine whether the battery cell is abnormal based on the comprehensive anomaly score and a preset score threshold.

[0163] Optionally, the first unit is used for:

[0164] Based on a preset clustering range and density, the voltage characteristic values ​​corresponding to the battery cells are clustered to obtain multiple clusters and discrete points;

[0165] Calculate the ratio of the number of discrete points to the total number of corresponding voltage characteristic values;

[0166] When the ratio is greater than a preset ratio, the first abnormal score is determined to be the first score;

[0167] When the ratio is not greater than the preset ratio, the first abnormal score is determined to be the second score;

[0168] The first score has an additive effect on the comprehensive anomaly score.

[0169] Optionally, the second unit is used for:

[0170] Calculate the mean of the voltage characteristic value, the mean of the resistance characteristic value, and the mean of the fused characteristic value corresponding to all the battery cells to obtain the reference voltage characteristic mean, the reference resistance characteristic mean, and the reference fused characteristic mean;

[0171] Calculate the mean of the voltage characteristic value, the mean of the resistance characteristic value, and the mean of the fusion characteristic value corresponding to the current battery cell to obtain the mean of cell voltage characteristic, the mean of cell resistance characteristic, and the mean of cell fusion characteristic;

[0172] Using the average reference voltage characteristic, the average reference resistance characteristic, and the average reference fusion characteristic as reference points, the deviation of the average cell voltage characteristic, the average cell resistance characteristic, and the average fusion characteristic relative to the reference points is calculated to obtain the second anomaly score of the current battery cell.

[0173] The present invention provides a battery pack anomaly detection device, which can be used to implement the steps in the aforementioned battery pack anomaly detection method embodiments.

[0174] It should be noted that the module division in the various battery pack anomaly detection devices provided in the above embodiments is illustrative and only represents a logical functional division. In actual implementation, other division methods may also be used. Furthermore, the functional modules in the various embodiments of this invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0175] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the technical solution of the embodiments of the present invention can be embodied in the form of a computer program product, which is stored in a computer storage medium and includes several instructions to cause an electronic device or processor to execute all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned computer 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.

[0176] Furthermore, the battery pack anomaly detection device and the battery pack anomaly detection method provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0177] Reference Figure 6 The diagram illustrates an electronic device according to an embodiment of the present invention. Figure 6 As shown, the electronic device in this embodiment of the invention includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the above-described battery pack anomaly detection method embodiment. Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-described battery pack anomaly detection device embodiment.

[0178] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which can be used to describe the execution process of the computer program in the electronic device.

[0179] The electronic device may be a desktop computer, a cloud server, or other computing device. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 6 This is merely one example of an electronic device and does not constitute a limitation on the electronic device. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0180] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0181] The memory can be an internal storage unit of the electronic device, such as a hard drive or RAM. Alternatively, it can be an external storage device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory can include both internal and external storage units. The memory is used to store the computer program and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or will be output.

[0182] This invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the battery pack anomaly detection method as described in the foregoing embodiments.

[0183] This invention also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the battery pack anomaly detection method as described in the foregoing embodiments.

[0184] This invention also discloses a computer program product that, when run on a computer, causes the computer to execute the battery pack anomaly detection method described in the foregoing embodiments.

[0185] The embodiments described above are only used to illustrate the technical solutions of this application, and are not intended to limit it. 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; and these 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, and should all be included within the protection scope of this application.

Claims

1. A method for detecting abnormalities in a battery pack, characterized in that, include: Acquire multiple frames of charging data for each battery cell in the battery pack within the same preset time period. The charging data is time-series data and includes at least timestamps, voltage values, and resistance values. For each frame of charging data of each battery cell, voltage characteristic value and resistance characteristic value are calculated based on the charging data; For each battery cell, a resistance feature matrix is ​​formed by combining the corresponding resistance feature values, and a voltage feature matrix is ​​formed by combining the corresponding voltage feature values. The battery cell is determined to be abnormal based on the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix, and the resistance characteristic matrix.

2. The method as described in claim 1, characterized in that, The calculation of voltage characteristic values ​​based on the charging data for each frame of each battery cell includes: The average voltage and standard deviation of voltage are calculated based on the voltage values ​​in the charging data of all the battery cells to obtain the comprehensive average voltage and comprehensive standard deviation of voltage. For each frame of charging data of each of the battery cells, determine the voltage value in the charging data; By combining the comprehensive voltage mean and the comprehensive voltage standard deviation, the degree of deviation of the current voltage value from the comprehensive voltage mean is determined, and the voltage characteristic value corresponding to the current voltage value is obtained.

3. The method as described in claim 2, characterized in that, Before calculating the voltage mean and voltage standard deviation based on the voltage values ​​in the charging data of all the battery cells to obtain the comprehensive voltage mean and comprehensive voltage standard deviation, the method further includes: The charging data of each battery cell with the same timestamp is used as a frame of raw data; For each frame of the original data, the original data is filtered by voltage value, according to a preset confidence interval and a preset number of cycles, in order to remove the charging data in the original data whose voltage value deviates from the confidence interval.

4. The method according to any one of claims 1-3, characterized in that, The step of determining whether the battery cell is abnormal based on the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix, and the resistance characteristic matrix includes: The voltage feature matrix is ​​transformed in the time-frequency domain to obtain the time-frequency feature matrix; The time-frequency feature matrix and the resistance feature matrix are fused, dimensionally reduced, and normalized to obtain a fused feature matrix with the same size as the time-frequency feature matrix and the resistance feature matrix; Each element in the fusion feature matrix is ​​configured as a fusion feature value for each of the battery cells; The battery cell is determined to be abnormal based on the voltage characteristic value, the resistance characteristic value, and the fusion characteristic value.

5. The method as described in claim 4, characterized in that, The step of determining whether the battery cell is abnormal based on the voltage characteristic value, the resistance characteristic value, and the fused characteristic value includes: For the current battery cell, a first anomaly score of the battery cell is determined based on the degree of dispersion of the corresponding voltage characteristic value; The second anomaly score of the battery cell is calculated based on the voltage characteristic value, the resistance characteristic value, and the fusion characteristic value; The average value of the corresponding voltage characteristic values ​​is used to obtain the third anomaly score of the battery cell; A comprehensive anomaly score is calculated based on the first anomaly score, the second anomaly score, and the third anomaly score, and the comprehensive anomaly score is positively correlated with the degree of anomaly of the battery cell; The battery cell is determined to be abnormal based on the comprehensive anomaly score and the preset score threshold.

6. The method as described in claim 5, characterized in that, Determining the first anomaly score of the battery cell based on the dispersion of the corresponding voltage characteristic value includes: Based on a preset clustering range and density, the voltage characteristic values ​​corresponding to the battery cells are clustered to obtain multiple clusters and discrete points; Calculate the ratio of the number of discrete points to the total number of corresponding voltage characteristic values; When the ratio is greater than a preset ratio, the first abnormal score is determined to be the first score; When the ratio is not greater than the preset ratio, the first abnormal score is determined to be the second score; The first score has an additive effect on the comprehensive anomaly score.

7. The method as described in claim 5, characterized in that, The calculation of the second anomaly score of the battery cell based on the voltage characteristic value, the resistance characteristic value, and the fused characteristic value includes: Calculate the mean of the voltage characteristic value, the mean of the resistance characteristic value, and the mean of the fused characteristic value corresponding to all the battery cells to obtain the reference voltage characteristic mean, the reference resistance characteristic mean, and the reference fused characteristic mean; Calculate the mean of the voltage characteristic value, the mean of the resistance characteristic value, and the mean of the fusion characteristic value corresponding to the current battery cell to obtain the mean of cell voltage characteristic, the mean of cell resistance characteristic, and the mean of cell fusion characteristic; Using the average reference voltage characteristic, the average reference resistance characteristic, and the average reference fusion characteristic as reference points, the deviation of the average cell voltage characteristic, the average cell resistance characteristic, and the average fusion characteristic relative to the reference points is calculated to obtain the second anomaly score of the current battery cell.

8. A battery pack anomaly detection device, characterized in that, include: The data acquisition module is used to acquire multiple frames of charging data of each battery cell in the battery pack within the same preset time period. The charging data is time-series data and includes at least timestamps, voltage values, and resistance values. The feature value calculation module is used to calculate voltage feature value and resistance feature value based on each frame of charging data for each of the battery cells; A matrix construction module is used to form a resistance feature matrix based on the corresponding resistance feature values ​​for each battery cell, and to form a voltage feature matrix based on the corresponding voltage feature values. An anomaly detection module is used to determine whether the battery cell is abnormal based on the voltage characteristic value, the resistance characteristic value, the voltage characteristic matrix, and the resistance characteristic matrix.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the battery pack anomaly detection method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the battery pack anomaly detection method as described in any one of claims 1-7.

Images