A battery pack fault detection method, device and vehicle

Abstract

The application provides a battery pack fault detection method and device and a vehicle, and the method comprises the following steps: obtaining characteristic data combinations of each battery cell in the battery pack, wherein the characteristic data combinations comprise a plurality of data characteristics and a characteristic value time sequence corresponding to each data characteristic; for each data characteristic, comparing the characteristic value time sequence corresponding to the data characteristic in each characteristic data combination, and extracting difference characteristic data of each battery cell in the data characteristic and other battery cells; for each battery cell, combining all difference characteristic data corresponding to the battery cell to obtain difference characteristic combination data of the battery cell; and based on an anomaly detection algorithm, determining whether the corresponding battery cell has a fault according to the difference characteristic combination data. The technical scheme of the application improves the accuracy of battery pack fault detection and can accurately detect battery cells with a fault risk.

Description

Technical Field

[0001] This invention relates to the field of battery fault detection technology, and more specifically, to a battery pack fault detection method, device, and vehicle. Background Technology

[0002] Battery packs consist of multiple individual battery cells connected in series and / or parallel, and are widely used in fields such as new energy vehicles and consumer electronics. For example, new energy vehicles use power battery packs to provide power to the vehicle. During use, changes in the operating environment and load can affect the performance of the battery pack, and may even lead to battery pack failure. Therefore, fault detection of battery packs is necessary to ensure their safety.

[0003] Currently, battery pack fault detection methods mainly include: (1) detection methods based on equivalent circuit models; (2) detection methods based on knowledge and statistics; and (3) data-driven artificial intelligence algorithms. Among them, methods (1) and (2) have poor real-time performance and can only use single data features for analysis, resulting in poor accuracy. Method (3) can use multi-dimensional data features for fault diagnosis, but it cannot effectively identify battery cells with fault risks, which is not conducive to subsequent accurate early warning and maintenance. Summary of the Invention

[0004] The problem addressed by this invention is how to improve the accuracy of battery pack fault detection and achieve fault detection at the individual battery cell level.

[0005] To address the above problems, the present invention provides a battery pack fault detection method, device, and vehicle.

[0006] In a first aspect, the present invention provides a battery pack fault detection method, comprising:

[0007] Obtain a combination of feature data for each individual battery cell in the battery pack. The combination of feature data includes multiple data features and a time series of feature values ​​corresponding to each data feature.

[0008] For each data feature, the time series of the feature values ​​corresponding to that data feature in each combination of feature data is compared and processed to extract the difference feature data between each battery cell and other battery cells on that data feature.

[0009] For each battery cell, all the differential feature data corresponding to that battery cell are combined to obtain the differential feature combination data of that battery cell;

[0010] Based on the anomaly detection algorithm, the system determines whether the corresponding battery cell has a fault based on the combination of the differential features.

[0011] Optionally, the feature value time series includes the feature values ​​of the data feature at multiple different acquisition times;

[0012] For each data feature, the time series of the feature values ​​corresponding to that data feature in each combination of feature data is compared and processed to extract the difference feature data between each battery cell and other battery cells on that data feature, including:

[0013] For each of the data features, determine the first mean and the first standard deviation of the feature values ​​at the same collection time in the corresponding feature value time series;

[0014] Based on the corresponding first mean, the feature values ​​in each feature value time series are sequentially centered and absolute value taken to obtain the preprocessed feature values;

[0015] A threshold is determined based on the first standard deviation and a preset hyperparameter. The feature values ​​in each feature value time series are then binarized according to the corresponding threshold to obtain the binarized feature values.

[0016] For each battery cell, the preprocessed feature values ​​corresponding to that battery cell are added together to obtain the first feature data of that battery cell; the binarized feature values ​​corresponding to that battery cell are added together to obtain the second feature data of that battery cell.

[0017] The first feature data is combined with the corresponding second feature data to obtain the difference feature data of each battery cell.

[0018] Optionally, the step of sequentially centering and taking the absolute value of the feature values ​​in each feature value time series according to the corresponding first mean to obtain the preprocessed feature values ​​includes:

[0019] For each acquisition time, the first mean value corresponding to the acquisition time is subtracted from the feature value in each feature value time series to obtain the centered feature value.

[0020] The absolute value of each of the centered feature values ​​is taken to obtain the preprocessed feature values.

[0021] Optionally, the step of binarizing the feature values ​​in each feature value time series according to the corresponding threshold to obtain the binarized feature values ​​includes:

[0022] For each acquisition time, the feature value corresponding to that acquisition time in each feature value time series is compared with the threshold corresponding to that acquisition time;

[0023] Based on the comparison results, feature values ​​that are greater than or equal to the threshold are set to 1, and feature values ​​that are less than the threshold are set to 0.

[0024] Optionally, after combining all the differential feature data corresponding to each battery cell to obtain the differential feature combination data of that battery cell, the method further includes:

[0025] For each battery cell, the differential feature combination data of that battery cell is standardized to obtain standardized differential feature combination data.

[0026] Optionally, for each of the battery cells, the standardization process for the combination of differential features of that battery cell to obtain standardized combination of differential features includes:

[0027] For each of the battery cells, determine the second mean and the second standard deviation of all the differential feature data in the differential feature combination data of that battery cell;

[0028] Subtract the corresponding second mean from each of the differential feature data to obtain the processed differential feature data;

[0029] Divide each of the processed differential feature data by the corresponding second standard deviation to obtain the standardized differential feature combination data.

[0030] Optionally, the data features include the voltage and temperature of the battery cell, and the anomaly detection algorithm includes at least one of the OPTICS clustering algorithm, the isolated forest algorithm, and the DBSCAN algorithm.

[0031] In a second aspect, the present invention provides a battery pack fault detection device, comprising:

[0032] The acquisition module is used to acquire the feature data combination of each battery cell in the battery pack. The feature data combination includes multiple data features and the feature value time series corresponding to each data feature.

[0033] The extraction module is used to compare the time series of the feature values ​​corresponding to the data feature in each combination of feature data for each data feature, and extract the difference feature data between each battery cell and other battery cells on the data feature; for each battery cell, the difference feature data corresponding to the battery cell is combined to obtain the difference feature combination data of the battery cell.

[0034] The detection module is used to determine whether the corresponding battery cell has a fault based on the combination of differential feature data using an anomaly detection algorithm.

[0035] Thirdly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the battery pack fault detection method as described in any of the first aspects.

[0036] Fourthly, the present invention provides a vehicle including a memory and a processor;

[0037] The memory is used to store computer programs;

[0038] The processor is configured to, when executing the computer program, implement the battery pack fault detection method as described in any of the first aspects.

[0039] The beneficial effects of the battery pack fault detection method, device, and vehicle of the present invention are as follows: For each battery cell, the feature value time series of the battery cell under different data features can be obtained, and the feature value time series may include feature values ​​at multiple different acquisition times. For a battery cell, its feature value time series is compared with the feature time series of other battery cells under the same data feature. For example, for each data feature, the feature time series of a battery cell is compared with the feature time series of other battery cells. By comparing, the difference feature data between the battery cell and other battery cells is extracted. The difference feature data reflects the differences between the battery cell and other battery cells, which is beneficial for realizing fault detection at the battery cell level. All the difference feature data corresponding to the battery cell are combined, that is, the difference feature data under each data feature corresponding to the battery cell are combined to obtain the difference feature combination data of the battery cell. The difference feature combination data reflects the differences between the battery cell and other battery cells from multiple data feature dimensions, improving the comprehensiveness of the extracted difference feature data. An anomaly detection algorithm is used to process the differential feature combination data to determine whether the corresponding battery cell has a fault. This achieves fault detection at the battery cell level. The differential feature combination data consists of differential feature data from multiple dimensions, which improves the accuracy of fault detection and can accurately detect battery cells in the battery pack that have a fault risk. Attached Figure Description

[0040] Figure 1 This is a schematic flowchart of a battery pack fault detection method according to an embodiment of the present invention;

[0041] Figure 2 This is a detailed flowchart of step S200 of the battery pack fault detection method according to an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of a battery pack fault detection device according to an embodiment of the present invention. Detailed Implementation

[0043] 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.

[0044] 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.

[0045] 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"; 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.

[0046] 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".

[0047] 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.

[0048] like Figure 1 As shown, the present invention provides a battery pack fault detection method, comprising:

[0049] Step S100: Obtain the feature data combination of each battery cell in the battery pack. The feature data combination includes multiple data features and the feature value time series corresponding to each data feature.

[0050] Specifically, the data features include the temperature and voltage of individual battery cells, and the feature value time series includes the feature values ​​of the data features at multiple different acquisition times. The values ​​of data features such as temperature and voltage of each individual battery cell can be acquired at fixed time intervals.

[0051] For example, for any single battery cell, arranging all the voltage values ​​collected for that battery cell in chronological order according to the time of collection can form a time series of voltage values ​​for that battery cell; arranging all the temperature values ​​collected for that battery cell in chronological order according to the time of collection can form a time series of temperature values ​​for that battery cell.

[0052] Step S200: For each data feature, compare the time series of the feature values ​​corresponding to the data feature in each feature data combination, and extract the difference feature data between each battery cell and other battery cells on the data feature.

[0053] For example, by comparing the voltage time series of any given battery cell with the voltage time series of other battery cells, the voltage-related differences between the given battery cell and other battery cells can be extracted. Similarly, the temperature and other data-related differences between the given battery cell and other battery cells can also be extracted.

[0054] Step S300: For each battery cell, combine all the differential feature data corresponding to that battery cell to obtain the differential feature combination data of that battery cell.

[0055] Specifically, for any given battery cell, the difference characteristic data between the battery cell and other battery cells in terms of voltage and temperature are combined to obtain the difference characteristic combination data of the battery cell. The difference characteristic combination data reflects the differences between the battery cell and other battery cells in multiple dimensions.

[0056] Step S400: Based on the anomaly detection algorithm, determine whether the corresponding battery cell has a fault according to the combination of the difference features.

[0057] Optionally, the anomaly detection algorithm includes at least one of the OPTICS clustering algorithm, the Isolation Forest algorithm, and the DBSCAN algorithm.

[0058] In this embodiment, for each battery cell, the feature value time series under different data features can be obtained. The feature value time series may include feature values ​​from multiple different acquisition times. For a battery cell, its feature value time series is compared with the feature time series of other battery cells under the same data feature. For example, for each data feature, the feature time series of a battery cell is compared with the feature time series of other battery cells. By comparing, the difference feature data between the battery cell and other battery cells is extracted. The difference feature data reflects the differences between the battery cell and other battery cells, which is beneficial for realizing fault detection at the battery cell level. All the difference feature data corresponding to the battery cell are combined, that is, the difference feature data under each data feature corresponding to the battery cell are combined to obtain the difference feature combination data of the battery cell. The difference feature combination data reflects the differences between the battery cell and other battery cells from multiple data feature dimensions, improving the comprehensiveness of the extracted difference feature data. An anomaly detection algorithm is used to process the differential feature combination data to determine whether the corresponding battery cell has a fault. This achieves fault detection at the battery cell level. The differential feature combination data consists of differential feature data from multiple dimensions, which improves the accuracy of fault detection and can accurately detect battery cells in the battery pack that have a fault risk.

[0059] Optionally, such as Figure 2 As shown, for each data feature, the process of comparing the time series of the feature values ​​corresponding to that data feature in each combination of feature data to extract the difference feature data between each battery cell and other battery cells on that data feature includes:

[0060] Step S210: For each data feature, determine the first mean and first standard deviation of the feature values ​​at the same acquisition time in the corresponding feature value time series.

[0061] Specifically, taking voltage as an example of data feature, the first mean and first standard deviation of the voltage at the same acquisition time in the feature value time series corresponding to each battery cell are determined.

[0062] In an optional embodiment, for each data feature, a data feature matrix can be constructed based on the time series of feature values ​​corresponding to the data feature in each feature data combination. The rows of the data feature matrix represent the feature values ​​of each battery cell at the same acquisition time, and the columns represent the feature values ​​of each battery cell at different acquisition times. Alternatively, the columns of the data feature matrix represent the feature values ​​of each battery cell at the same acquisition time, and the rows represent the feature values ​​of each battery cell at different acquisition times.

[0063] For example, assuming the data feature is voltage and the data feature matrix is ​​voltage matrix V, where the rows of voltage matrix V represent the voltage values ​​of each battery cell at the same acquisition time and the columns represent the voltage values ​​of each battery cell at different acquisition times, the first mean and the first standard deviation of the voltage values ​​in each row of voltage matrix V can be directly calculated.

[0064] Step S220: Based on the corresponding first mean, the feature values ​​in each feature value time series are sequentially centered and absolute valued to obtain preprocessed feature values.

[0065] Specifically, each feature value is centered according to the first mean at the corresponding acquisition time to obtain the centered feature value. Then, the absolute value of the centered feature value is taken to obtain the preprocessed feature value. Through centering and absolute value taking, the deviation between each battery cell and the average value of all battery cells at the same acquisition time can be extracted. The larger the preprocessed feature value, the more the feature value of the battery cell deviates from the mean for that data feature.

[0066] Continuing from the previous example, each voltage value in the voltage matrix V is centered based on the first mean of its row to obtain the centered voltage matrix V_mean. Then, the absolute values ​​of each value in the centered voltage matrix V_mean are taken to obtain the voltage absolute deviation matrix V_abs, where each value in the voltage absolute deviation matrix V_abs is the preprocessed eigenvalue.

[0067] Step S230: Determine a threshold based on the first standard deviation and a preset hyperparameter, and perform binarization processing on the feature values ​​in each feature value time series according to the corresponding threshold to obtain the binarized feature values.

[0068] Specifically, the product of the first standard deviation and the preset hyperparameter can be used as the threshold. Each feature value is then binarized according to the threshold calculated based on the first standard deviation at the corresponding acquisition time to obtain the binarized feature value.

[0069] Continuing from the previous example, the product of the first standard deviation (std) of the voltage values ​​in each row of the voltage matrix V and the preset hyperparameter Z_score can be used as the threshold M for the voltage values ​​in that row. The preset hyperparameter Z_score can be set to 3. By binarizing each voltage value in the voltage matrix V according to the threshold M of its row, the voltage over-threshold matrix V_margin can be obtained. The values ​​in the voltage over-threshold matrix V_margin are the eigenvalues ​​after binarization.

[0070] Step S240: For each battery cell, add all the preprocessed feature values ​​corresponding to that battery cell to obtain the first feature data of that battery cell; add all the binarized feature values ​​corresponding to that battery cell to obtain the second feature data of that battery cell.

[0071] Specifically, for any single battery cell, the first feature data of the battery cell can be obtained by summing the preprocessed feature values ​​of the battery cell at each acquisition time. The first feature data is the cumulative value of all deviations of the battery cell that exceed the mean of the corresponding data at each acquisition time. The second feature data of the battery cell can be obtained by summing all the binarized feature values ​​of the battery cell at each acquisition time. The second feature data is the cumulative value of the number of times the battery cell exceeds the tolerance at each acquisition time.

[0072] Continuing the previous example, the data in the absolute voltage deviation matrix V_abs are summed column-wise to obtain the cumulative voltage deviation matrix V_acc. Each data point in the cumulative voltage deviation matrix V_acc represents the first feature data. Similarly, the data in the voltage exceedance matrix V_margin are summed column-wise to obtain the cumulative voltage exceedance count matrix V_num. Each data point in the cumulative voltage exceedance count matrix V_num represents the second feature data. Assuming the voltage matrix V is an m×n matrix, then the cumulative voltage deviation matrix V_acc and the cumulative voltage exceedance count matrix V_num are both 1×n matrices.

[0073] Step S250: Combine the first feature data with the corresponding second feature data to obtain the difference feature data of each battery cell.

[0074] Specifically, for any given battery cell, combining its corresponding first feature data and second feature data yields the differential feature matrix of that battery cell.

[0075] Continuing from the previous example, the cumulative voltage deviation matrix V_acc and the cumulative voltage deviation count matrix V_num are merged column by column to obtain the voltage feature matrix V_fea. The voltage feature matrix V_fea is a 2×n matrix, and one column of data in the voltage feature matrix V_fea is the voltage difference feature data of a single battery cell.

[0076] Calculate the difference feature matrix of each battery cell when the data features are equal in temperature using the same method as in the example above. For example, by performing the same processing on the temperature matrix T, we can obtain the temperature feature matrix T_fea. One column of the temperature feature matrix T_fea represents the temperature difference feature data of a battery cell.

[0077] In this optional embodiment, firstly, the first mean and first standard deviation of all feature values ​​of the same data feature collected from each battery cell at the same time are calculated. Based on the first mean at the corresponding collection time, each feature value is centered and its absolute value is taken. This determines the deviation between each feature value and the first mean at the corresponding collection time, extracting the difference features between each battery cell and the mean of all battery cells at different times. The first standard deviation can be multiplied by a preset hyperparameter to determine a threshold. Based on the threshold at the corresponding collection time, each feature value is binarized. The binarized feature values ​​can determine whether each feature value exceeds the tolerance, which can be used to determine whether the corresponding cell has an anomaly. Adding the preprocessed feature values ​​at each collection time, and adding the binarized feature values ​​at each collection time, can reduce the impact of data errors at a single collection time on the accuracy of the overall anomaly feature. Combining the first feature data with the second feature data, that is, combining the difference features extracted from multiple data features, yields difference feature data reflecting the differences between this battery cell and other battery cells. By combining anomaly detection algorithms with the differential feature data of each battery cell, fault detection at the battery cell level can be achieved. Furthermore, by extracting differential features between battery cells from multiple data feature dimensions, the accuracy of the extracted differential features is improved, thereby enhancing the accuracy of detecting battery cells with potential fault risks.

[0078] Optionally, the step of sequentially centering and taking the absolute value of the feature values ​​in each feature value time series according to the corresponding first mean to obtain the preprocessed feature values ​​includes:

[0079] For each acquisition time, the first mean value corresponding to that acquisition time is subtracted from the feature value in each feature value time series to obtain the centered feature value.

[0080] The absolute value of each of the centered feature values ​​is taken to obtain the preprocessed feature values.

[0081] Continuing the previous example, subtracting the mean of the first row from each voltage value in the voltage matrix V yields a centered matrix. The data in this centered matrix are the centered eigenvalues. Taking the absolute value of each centered eigenvalue yields the voltage absolute deviation matrix V_abs. The data in this voltage absolute deviation matrix V_abs are the preprocessed eigenvalues.

[0082] In this optional embodiment, the preprocessed feature values ​​reflect the deviation of each battery cell from the average value of all corresponding battery cells at different acquisition times. A larger preprocessed feature value indicates a greater deviation from the average value, suggesting a higher likelihood of anomalies; conversely, a smaller preprocessed feature value indicates a smaller deviation from the average value, suggesting a lower probability of anomalies. By using centering and absolute value processing to extract preprocessed feature values ​​reflecting the deviation from the average value, it is beneficial for achieving fault detection at the battery cell level.

[0083] Subsequently, the preprocessed feature values ​​from each acquisition time are summed to obtain the first feature data, which reflects the deviation of the battery cell from the corresponding average value across multiple acquisition times. A larger first feature data indicates that the battery cell may have deviated from the average value over a long period, increasing the likelihood of an anomaly. Therefore, summing the preprocessed feature values ​​from each acquisition time reduces the impact of potential errors in the preprocessed feature values ​​from a single acquisition time, improving the accuracy of the extracted features.

[0084] Optionally, the step of binarizing the feature values ​​in each feature value time series according to the corresponding threshold to obtain the binarized feature values ​​includes:

[0085] For each acquisition time, the feature value corresponding to that acquisition time in each feature value time series is compared with the threshold corresponding to that acquisition time;

[0086] Based on the comparison results, feature values ​​that are greater than or equal to the threshold are set to 1, and feature values ​​that are less than the threshold are set to 0.

[0087] Continuing from the previous example, each voltage value in the voltage matrix V is compared with the threshold M of its row. Voltage values ​​greater than or equal to the threshold M are set to 1, and voltage values ​​less than the threshold M are set to 0, thus forming the voltage threshold matrix V_margin.

[0088] In this optional embodiment, comparing the feature value with a multiple of its corresponding standard deviation determines the distance of the feature value from the mean. The larger the difference between the feature value and the threshold, the farther the feature value is from the mean, indicating a lower probability of the feature value appearing, and thus a higher probability that the corresponding battery cell is abnormal. Through binarization, feature values ​​greater than or equal to the threshold are set to 1, indicating a higher probability of abnormality in the corresponding battery cell; feature values ​​less than the threshold are set to 0, indicating a lower probability of abnormality in the corresponding battery cell. The extracted binarized feature values ​​can intuitively reflect the probability of abnormality in a battery cell, which is beneficial for subsequent battery cell-level fault detection.

[0089] Subsequently, the binarized feature values ​​from each acquisition time point are summed to obtain the second feature data. This second feature data reflects the number of times the feature value is 1, which in turn reflects the number of times the battery cell is judged to have a high probability of being abnormal. Therefore, the larger the second feature data, the more times the battery cell is judged to have a high probability of being abnormal, and the greater the probability that the battery cell is abnormal. Summing the binarized feature values ​​from each acquisition time point can reduce the impact of potential errors in the binarized feature values ​​from a single acquisition time point, thereby improving the accuracy of the extracted features.

[0090] Optionally, after combining all the differential feature data corresponding to each battery cell to obtain the differential feature combination data of that battery cell, the method further includes:

[0091] For each battery cell, the differential feature combination data of that battery cell is standardized to obtain standardized differential feature combination data.

[0092] Continuing from the previous example, the voltage feature matrix V_fea and the temperature feature matrix T_fea are combined column by column to form a 4×n matrix, which yields the feature matrix F. Each data point in the feature matrix F represents the differential feature data, and each column of data represents the combination of differential features of a single battery cell.

[0093] Specifically, standardization involves subtracting the mean from each differential feature in the differential feature combination data and then dividing by its standard deviation.

[0094] Optionally, for each of the battery cells, the standardization process for the combination of differential features of that battery cell to obtain standardized combination of differential features includes:

[0095] For each of the battery cells, determine the second mean and the second standard deviation of all the differential feature data in the differential feature combination data of that battery cell.

[0096] Continuing from the previous example, determine the second mean and second standard deviation of each column of data in the feature matrix F.

[0097] Subtract the corresponding second mean from each of the differential feature data to obtain the processed differential feature data.

[0098] Continuing from the previous example, subtract the second mean of the column from each data point in the feature matrix F to obtain the processed feature matrix. The data points in the processed feature matrix are the processed difference feature matrix.

[0099] Divide each of the processed differential feature data by the corresponding second standard deviation to obtain the standardized differential feature combination data.

[0100] Continuing from the previous example, dividing each data point in the processed feature matrix by the second standard deviation of its column yields the final feature matrix F_final. Each column in the final feature matrix F_final represents the standardized combination of differential features of a single battery cell.

[0101] Based on the OPTICS clustering algorithm, cluster analysis is performed according to the final feature matrix F_final. If a point in the final feature matrix F_final is identified as a noise point, the corresponding battery cell is determined to be faulty.

[0102] The OPTICS clustering algorithm is a density-based clustering algorithm, an improved version of DBSCAN, with better generalization ability. The algorithm builds a reachability graph, assigning each sample a reachability distance and a cluster ordering attribute; these two attributes are used to determine cluster membership relationships when fitting the model.

[0103] In this optional embodiment, standardization processing can unify multiple data features into the same evaluation dimension, which facilitates subsequent data processing to achieve fault detection.

[0104] It is understandable that the above example uses the data feature matrix to represent the feature values ​​of each battery cell at the same acquisition time in the rows and the feature values ​​of each battery cell at different acquisition times in the columns. The case where the data feature matrix represents the feature values ​​of each battery cell at the same acquisition time in the columns and the feature values ​​of each battery cell at different acquisition times in the rows is similar to the above example. The processing can be adjusted accordingly for the rows and columns, and will not be elaborated here.

[0105] Secondly, such as Figure 3 As shown, the present invention provides a battery pack fault detection device, comprising:

[0106] The acquisition module is used to acquire the feature data combination of each battery cell in the battery pack. The feature data combination includes multiple data features and the feature value time series corresponding to each data feature.

[0107] The extraction module is used to compare the time series of the feature values ​​corresponding to the data feature in each combination of feature data for each data feature, and extract the difference feature data between each battery cell and other battery cells on the data feature; for each battery cell, the difference feature data corresponding to the battery cell is combined to obtain the difference feature combination data of the battery cell.

[0108] The detection module is used to determine whether the corresponding battery cell has a fault based on the combination of differential feature data using an anomaly detection algorithm.

[0109] This battery pack fault detection device can be deployed on vehicles or in the cloud, and is used to perform online detection and early warning based on the combination of characteristic data of individual battery cells acquired in real time.

[0110] The battery pack fault detection device of the present invention is used to implement the battery pack fault detection method described above. Its advantages over the prior art are the same as the advantages of the battery pack fault detection method described above over the prior art, and will not be repeated here.

[0111] Optionally, the feature value time series includes the feature values ​​of the data features at multiple different acquisition times; the data features include the voltage and temperature of the battery cells; and the anomaly detection algorithm includes at least one of the OPTICS clustering algorithm, the isolated forest algorithm, and the DBSCAN algorithm.

[0112] Optionally, the extraction module is specifically used for:

[0113] For each of the data features, determine the first mean and the first standard deviation of the feature values ​​at the same collection time in the corresponding feature value time series;

[0114] Based on the corresponding first mean, the feature values ​​in each feature value time series are sequentially centered and absolute value taken to obtain the preprocessed feature values;

[0115] A threshold is determined based on the first standard deviation and a preset hyperparameter. The feature values ​​in each feature value time series are then binarized according to the corresponding threshold to obtain the binarized feature values.

[0116] For each battery cell, the preprocessed feature values ​​corresponding to that battery cell are added together to obtain the first feature data of that battery cell; the binarized feature values ​​corresponding to that battery cell are added together to obtain the second feature data of that battery cell.

[0117] The first feature data is combined with the corresponding second feature data to obtain the difference feature data of each battery cell.

[0118] Optionally, the extraction module is specifically used to: for each acquisition time, subtract the first mean value corresponding to the acquisition time from the feature value in each feature value time series to obtain the centered feature value; and take the absolute value of each centered feature value to obtain the preprocessed feature value.

[0119] Optionally, the extraction module is specifically used to: for each acquisition time, compare the feature value corresponding to the acquisition time in each feature value time series with the threshold corresponding to the acquisition time; and set the feature value that is greater than or equal to the threshold to 1 and the feature value that is less than the threshold to 0 according to the comparison result.

[0120] Optionally, the battery pack fault detection device further includes a standardization processing module, which is used to: for each battery cell, perform standardization processing on the combination data of the differential features of the battery cell to obtain standardized combination data of the differential features.

[0121] Optionally, the standardization processing module is specifically used to: for each battery cell, determine the second mean and second standard deviation of all the differential feature data in the differential feature combination data of the battery cell; subtract the corresponding second mean from each differential feature data to obtain the processed differential feature data; and divide each processed differential feature data by the corresponding second standard deviation to obtain the standardized differential feature combination data.

[0122] Thirdly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the battery pack fault detection method as described in any of the first aspects.

[0123] Fourthly, the present invention provides a vehicle including a memory and a processor;

[0124] The memory is used to store computer programs;

[0125] The processor is configured to, when executing the computer program, implement the battery pack fault detection method as described in any of the first aspects.

[0126] 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.

[0127] 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 battery pack fault detection method, characterized in that, include: The feature data combination of each battery cell in the battery pack is obtained. The feature data combination includes multiple data features and a time series of feature values ​​corresponding to each data feature. The time series of feature values ​​includes the feature values ​​of the data features at multiple different acquisition times. For each data feature, the time series of the feature values ​​corresponding to that data feature in each feature data combination is compared and processed to extract the difference feature data between each battery cell and other battery cells on that data feature, including: for each data feature, determining the first mean and first standard deviation of the feature values ​​at the same acquisition time in each of the corresponding feature value time series; performing centering and absolute value processing on the feature values ​​in each feature value time series according to the corresponding first mean to obtain preprocessed feature values; determining a threshold according to the first standard deviation and a preset hyperparameter, and then... The threshold is used to binarize the feature values ​​in the time series of each feature value to obtain binarized feature values; for each battery cell, all the preprocessed feature values ​​corresponding to that battery cell are added together to obtain the first feature data of that battery cell; all the binarized feature values ​​corresponding to that battery cell are added together to obtain the second feature data of that battery cell; the first feature data and the corresponding second feature data are combined to obtain the difference feature data of each battery cell; the difference feature data is used to reflect the difference between that battery cell and other battery cells; For each battery cell, all the differential feature data corresponding to that battery cell are combined to obtain the differential feature combination data of that battery cell; Based on the anomaly detection algorithm, the system determines whether the corresponding battery cell has a fault based on the combination of the differential features.

2. The battery pack fault detection method according to claim 1, characterized in that, The step of performing centering and absolute value processing on the feature values ​​in each feature value time series according to the corresponding first mean to obtain preprocessed feature values ​​includes: For each acquisition time, the first mean value corresponding to the acquisition time is subtracted from the feature value in each feature value time series to obtain the centered feature value. The absolute value of each of the centered feature values ​​is taken to obtain the preprocessed feature values.

3. The battery pack fault detection method according to claim 1, characterized in that, The step of binarizing the feature values ​​in each feature value time series according to the corresponding threshold to obtain the binarized feature values ​​includes: For each acquisition time, the feature value corresponding to that acquisition time in each feature value time series is compared with the threshold corresponding to that acquisition time; Based on the comparison results, feature values ​​that are greater than or equal to the threshold are set to 1, and feature values ​​that are less than the threshold are set to 0.

4. The battery pack fault detection method according to any one of claims 1 to 3, characterized in that, After combining all the differential feature data corresponding to each battery cell to obtain the differential feature combination data of that battery cell, the method further includes: For each battery cell, the differential feature combination data of that battery cell is standardized to obtain standardized differential feature combination data.

5. The battery pack fault detection method according to claim 4, characterized in that, For each of the battery cells, the standardized data of the differential feature combinations for that battery cell are processed to obtain standardized differential feature combinations, including: For each of the battery cells, determine the second mean and the second standard deviation of all the differential feature data in the differential feature combination data of that battery cell; Subtract the corresponding second mean from each of the differential feature data to obtain the processed differential feature data; Divide each of the processed differential feature data by the corresponding second standard deviation to obtain the standardized differential feature combination data.

6. The battery pack fault detection method according to any one of claims 1 to 3, characterized in that, The data features include the voltage and temperature of the individual battery cells, and the anomaly detection algorithm includes at least one of the OPTICS clustering algorithm, the isolated forest algorithm, and the DBSCAN algorithm.

7. A battery pack fault detection device, characterized in that, include: The acquisition module is used to acquire the feature data combination of each battery cell in the battery pack. The feature data combination includes multiple data features and a feature value time series corresponding to each data feature. The feature value time series includes the feature values ​​of the data features at multiple different acquisition times. The extraction module is used to compare the time series of the feature values ​​corresponding to the data feature in each combination of feature data for each data feature, and extract the difference feature data between each battery cell and other battery cells on the data feature. The difference feature data is used to reflect the difference between the battery cell and other battery cells. For each battery cell, all the difference feature data corresponding to the battery cell are combined to obtain the difference feature combination data of the battery cell. The detection module is used to determine whether the corresponding battery cell has a fault based on the combination of differential feature data, using an anomaly detection algorithm. The extraction module is specifically used to: for each data feature, determine the first mean and the first standard deviation of the feature values ​​at the same collection time in each of the corresponding feature value time series; and perform centering and absolute value processing on the feature values ​​in each of the feature value time series according to the corresponding first mean to obtain the preprocessed feature values. A threshold is determined based on the first standard deviation and a preset hyperparameter. The feature values ​​in each feature value time series are then binarized according to the corresponding threshold to obtain the binarized feature values. For each of the battery cells, the first feature data of the battery cell is obtained by adding all the preprocessed feature values ​​corresponding to that battery cell. The first feature data is obtained by summing all the binarized feature values ​​corresponding to the battery cell; the second feature data is obtained by combining the first feature data with the corresponding second feature data.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the battery pack fault detection method as described in any one of claims 1 to 6.

9. A vehicle, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to implement the battery pack fault detection method as described in any one of claims 1 to 6 when executing the computer program.

Images