An abnormality detection method, system and device for a battery cell

Abstract

The application discloses a kind of abnormal detection methods of battery cell, comprising the following steps: collecting the operating parameter of battery pack, and filtering cleaning;The operating parameter after filtering cleaning is segmented according to multiple operating intervals of battery pack;Multiple operating intervals include charging interval, discharging interval, standing interval after discharging, standing interval after charging;According to multiple operating intervals, calculate cell voltage consistency coefficient and adaptive voltage difference, and carry out real-time abnormal alarm identification;Extract the voltage characteristics of charging interval and periodic signal statistical feature vector, and carry out periodic abnormal feature identification;According to real-time abnormal alarm and periodic abnormal feature, locate abnormal cell.The detection method of the application, by segmenting the operating state of battery pack, respectively analyzing operating parameter, identifying abnormal alarm and periodic abnormal feature, realizes accurate positioning of abnormal cell.

Description

Technical Field

[0001] This invention belongs to the field of battery cell testing technology, specifically relating to a method, system, and equipment for detecting abnormalities in battery cells. Background Technology

[0002] A battery pack is composed of multiple cells connected in series and parallel, and is widely used in the power and energy storage industries. The consistency of the cells within a battery pack significantly impacts its overall performance. Poor cell consistency exacerbates the performance degradation of weaker cells, thus limiting the overall performance of the battery pack and even posing a risk of thermal runaway. Cell inconsistencies are primarily manifested in voltage, capacity, and internal resistance, with voltage consistency being the most direct indicator of the battery pack's overall consistency.

[0003] Currently, the detection of cell voltage consistency in battery management systems (BMS) is mainly based on the voltage difference between individual cells in the battery pack under static conditions. However, there is no method for detecting voltage consistency during dynamic operation, especially during the dynamic operation of power batteries. Therefore, there is a need to provide a new method for evaluating cell voltage. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of this invention is to solve the problems in the prior art and provide a method, system and device for abnormal detection of battery cells. By dividing the operating state of the battery cell into sections and performing segmented detection of the operating state of the battery cell in different sections, a more accurate abnormal battery cell identification effect can be achieved.

[0005] This invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for detecting anomalies in battery cells, comprising the following steps: The operating parameters of the battery pack are collected and filtered and cleaned; the operating parameters include current, cell voltage, temperature, and total voltage. The operating parameters after filtration and cleaning are divided into multiple operating ranges of the battery pack; the multiple operating ranges include a charging range, a discharging range, a post-discharge resting range, and a post-charge resting range; Based on multiple operating ranges, the cell voltage consistency coefficient and adaptive voltage difference are calculated, and real-time abnormal alarm identification is performed. Extract voltage features and periodic signal statistical feature vectors from the charging range, and identify periodic anomalies. Abnormal battery cells can be located based on real-time abnormal alarms and periodic abnormal characteristics.

[0006] To optimize the above technical solution, the specific measures also include: Further, the cell voltage consistency coefficient is calculated as follows: Extract cell voltage data within the charging and discharging ranges of the battery pack, and set a dynamic sampling window to collect all valid cell voltage data; Within the dynamic sampling window, the effective cell voltage data corresponding to every two cells are combined, and the corresponding real-time Pearson correlation coefficient matrix is ​​calculated. The cell voltage consistency coefficient is then calculated based on the Pearson correlation coefficient matrix.

[0007] Furthermore, within the dynamic sampling window, the effective cell voltage data corresponding to every two cells are combined, and the corresponding real-time Pearson correlation coefficient matrix is ​​calculated. Based on the Pearson correlation coefficient matrix, the cell voltage consistency coefficient is calculated, specifically as follows: Calculate any two battery cells and Pearson correlation coefficient of effective cell voltage data ; ; in: Indicates the first The energy-saving cell is in the first dynamic sampling window. Voltage value at each sampling time. Indicates the first The energy-saving cell is in the first dynamic sampling window. Voltage value at each sampling time. Indicates the first The average value of N voltage values ​​of the energy-saving cell within the dynamic sampling window. Indicates the first The average value of N voltage values ​​of the energy-saving cell within the dynamic sampling window, where N represents the number of valid sampling points within the sliding window; By iterating through all battery cells, several Pearson correlation coefficients are calculated, and a Pearson correlation coefficient matrix is ​​established. ; ; ; ; Extracting the Pearson correlation coefficient matrix Minimum value of off-diagonal elements in , ; With the minimum value C is used as the cell voltage consistency coefficient.

[0008] Further, the adaptive voltage difference and real-time anomaly alarm identification are calculated as follows: Collect relevant parameters of the battery cell in offline mode; the relevant parameters include battery cell voltage, current, temperature, SOC, and total voltage; Based on the relevant parameters of each cell, the voltage consistency tolerance curve of the cell under multiple operating ranges was fitted. Calculate the tolerance threshold of the battery cell under different operating ranges. ; Compare the voltage consistency tolerance curves of multiple battery cells with the tolerance threshold. Compare and determine the adaptive voltage difference S of the entire cell voltage; The cell voltage consistency coefficient C is compared with the corresponding threshold range. If the cell voltage consistency coefficient C does not meet the threshold range, a serious alarm is issued. The adaptive voltage difference S is compared with the tolerance threshold. By comparison, if the adaptive voltage difference does not meet the tolerance threshold... If so, a serious alarm will be issued; If neither the cell voltage consistency coefficient C nor the adaptive voltage difference S is in a critical alarm state, then the cell voltage consistency coefficient C and the adaptive voltage difference S are fused together, and the fused comprehensive data is analyzed. The value is compared with the corresponding set threshold. If the value exceeds the set threshold, an abnormal alarm is triggered.

[0009] Further, the voltage characteristics of the charging range are extracted, specifically as follows: Extract the cell voltage data during the charging period and perform filtering processing; The cell voltage data is converted into an incremental capacity curve, and the voltage characteristics corresponding to the main peak of the incremental capacity curve are extracted. The voltage characteristics include peak voltage, peak height, peak area, and peak voltage span.

[0010] Further, the statistical feature vector of the periodic signal is extracted, specifically as follows: The cell voltage data in the charging range is segmented to obtain a voltage signal with a set period. Multi-dimensional statistical features of voltage signals are extracted to form a periodic signal statistical feature vector. The multi-dimensional statistical features include features describing the signal reference and range, features describing the data dispersion, features describing the data shape, and features describing the waveform characteristics. The characteristics describing the reference and range of the signal include the maximum value, minimum value, mean value, and peak value; the characteristics describing the dispersion of the data include the maximum voltage difference ΔVmax between the resting voltage and the standard cell voltage, the standard deviation σ, and the coefficient of variation CV=σ / μ, where μ represents the mean value; the characteristics describing the shape of the data include skewness and kurtosis; the characteristics describing the waveform characteristics include waveform factor, peak factor, impulse factor, and margin factor.

[0011] Furthermore, the periodic anomaly feature identification specifically includes: Extract the periodic signal statistical feature vector for each battery cell; The periodic signal statistical feature vectors of all battery cells are used to construct a feature matrix X, and then the data is standardized. Calculate the covariance matrix of the feature matrix X after data standardization. The covariance matrix is ​​decomposed into eigenvalues ​​to obtain several eigenvalues ​​and their corresponding eigenvectors of the principal component directions; Select the eigenvalues ​​of the first K principal component directions whose variance accounts for more than 85%, and construct the eigenspace K; Projecting the feature matrix X onto the feature subspace K yields the dimension-reduced score matrix T; each row of the score matrix T represents the principal component score of a sample. Transform the score matrix T in reverse into N-dimensional space to obtain the reconstructed feature matrix. ; Calculate the characteristic matrix X and the characteristic matrix The T² statistic and the squared prediction error; The T² statistic and the squared prediction error are compared with preset thresholds to identify periodic anomalies.

[0012] Furthermore, based on the aforementioned real-time anomaly alarms and periodic anomaly characteristics, the abnormal battery cells are located, specifically as follows: Extract the operating range corresponding to real-time anomaly alarms; Locate the time period of abnormal voltage based on the operating range; Read the time window information of the abnormal voltage period, the time window information including the start timestamp and the end timestamp; Extract the position of the eigenvalue corresponding to the periodic anomaly feature in the feature matrix X; Locate abnormal voltage data based on location; Read the cell identifier corresponding to the abnormal voltage data; The abnormal battery cell can be located based on the time window information and the cell identifier.

[0013] A second aspect of the present invention provides an anomaly detection system for battery cells, comprising: The data acquisition module is used to collect the operating parameters of the battery pack and perform filtering and cleaning; the operating parameters include current, cell voltage, temperature, and total voltage. The data segmentation module is used to segment the filtered and cleaned operating parameters according to multiple operating intervals of the battery pack; the multiple operating intervals include a charging interval, a discharging interval, a post-discharge resting interval, and a post-charge resting interval; The first processing module is used to calculate the consistency coefficient of the cell voltage; The second processing module is used to calculate the adaptive voltage difference of the battery cell voltage; The third processing module is used to extract the voltage characteristics of the charging range; The fourth processing module is used to extract statistical feature vectors of periodic signals; The first analysis module is used to identify real-time abnormal alarms based on the consistency coefficient of the cell voltage and the adaptive voltage difference. The second analysis module is used to identify periodic abnormal features based on the voltage characteristics of the charging range and the statistical feature vector of the periodic signal. The alarm module is used to locate abnormal battery cells based on real-time abnormal alarms and periodic abnormal characteristics.

[0014] In a third aspect, the present invention provides an apparatus comprising a processor and a memory; the memory stores a set of program instructions; when the set of program instructions stored in the memory is loaded and executed by the processor, the aforementioned method for detecting abnormalities in battery cells can be implemented.

[0015] The beneficial effects of this invention are: Compared with existing technologies, the battery cell anomaly detection method, system and equipment disclosed in this invention can accurately locate abnormal cells by segmenting the operating state of the battery pack, analyzing the operating parameters of each operating state, identifying abnormal alarms and periodic abnormal features. Attached Figure Description

[0016] Figure 1 This is a flowchart of an abnormality detection method for battery cells according to the present invention; Figure 2 This is a voltage band diagram showing the voltage consistency tolerance of the battery cell in different SOC ranges of the present invention; Figure 3 This is the incremental capacity IC curve of the present invention; Figure 4 This is the consistency coefficient alarm division curve diagram of the present invention; Figure 5 This is a module connection diagram of an abnormality detection system for battery cells according to the present invention; Figure 6 This is a schematic diagram of the structure of the device of the present invention. Detailed Implementation

[0017] To clarify the technical solution and working principle of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0018] Example 1: As Figure 1 As shown, this embodiment provides a method for detecting abnormalities in battery cells, mainly targeting battery packs containing multiple individual cells, and specifically includes the following steps: S10: Collect the operating parameters of the battery pack and perform filtration and cleaning; the operating parameters include current, cell voltage, temperature, and total voltage; Cell voltage refers to the voltage value of each individual cell, while total voltage refers to the overall voltage value of the entire battery pack. Operating parameters are primarily collected by various sensors, with conventional sensor types selected, which will not be elaborated upon in this embodiment. During operation, individual cells within the battery pack may generate abnormal values ​​due to electromagnetic interference. Therefore, it is necessary to filter out these abnormal values. This is achieved through two methods: firstly, by limiting threshold ranges and rates of change to filter out extreme or transient values ​​that are significantly deviating from physical reality; secondly, by using moving averages or low-pass filtering to smooth out high-frequency random noise from various sensors. Then, combined with the inherent physical laws of the battery pack system (such as verification of the sum of total voltage and individual cell voltages, and judgment of voltage stability during rest), data with obvious abnormal values ​​in the operating parameters are removed, as are duplicate data such as state of health (SOH) and state of charge (SOC).

[0019] S20: The operating parameters after filtration and cleaning are divided into multiple operating intervals of the battery pack; the multiple operating intervals include a charging interval, a discharging interval, a post-discharge resting interval, and a post-charge resting interval.

[0020] Because the operating information of battery cells varies under different operating conditions, the classification standard is mainly based on the changes in current and voltage. The changes in current mainly include four stages: the current is continuously positive and the voltage gradually rises to the cutoff condition; the current is continuously negative and the voltage gradually decreases to the cutoff condition; the current returns to zero and the voltage slowly recovers from the end of discharge and tends to stabilize; the current returns to zero and the voltage slowly decreases from the end of charging and tends to stabilize. These four stages are defined as follows: charging interval, discharging interval, post-discharge resting interval, and post-charging resting interval.

[0021] The charging and discharging ranges can fully demonstrate the energy transfer process and dynamic characteristics of the battery. The two resting ranges can provide information on the electrochemical relaxation process and open-circuit voltage, which can be used for accurate SOC calibration and state of health analysis.

[0022] S30: Calculates the cell voltage consistency coefficient and adaptive voltage difference based on multiple operating ranges, and performs real-time abnormal alarm identification.

[0023] The cell voltage consistency coefficient can quantify the differences between individual cells in the battery pack, while the adaptive voltage difference can achieve differentiated alarm thresholds for different operating ranges. Finally, by combining the cell voltage consistency coefficient and the adaptive voltage difference, accurate identification and judgment of abnormal alarms can be achieved.

[0024] Step 1: Calculate the cell voltage consistency coefficient.

[0025] First, cell voltage data is extracted from the charging and discharging ranges of the battery pack. For the charging range, voltage data stability is considered; for example, charging ranges exceeding 95% SOC are excluded. Similarly, for the discharging range, ranges with an SOC of around 5% are excluded. Then, a dynamic sampling window is set to collect all valid cell voltage data. The criteria for setting the dynamic sampling window are: firstly, the window length needs to cover a certain operating cycle (e.g., 1-2 complete charging / discharging ranges) to ensure sufficient data volume for calculating stable correlations; secondly, the window length should not be too long to avoid failing to capture sudden faults. The step size of the dynamic sampling window is set to one sampling interval in this embodiment, but the specific step size can be selected and set according to actual needs.

[0026] Before acquiring valid cell voltage data within the dynamic sampling window, the dynamic sampling window needs to be calibrated using a square wave correction signal. The amplitude of the square wave correction signal needs to be significantly higher than the sampling noise, but lower than the operating voltage of the battery pack cells. The period of the square wave correction signal is an odd number of seconds to avoid the normal operating frequency of the battery pack, facilitating extraction and separation. The actual correction process of the square wave signal is as follows: A pre-set standard square wave signal is synchronously injected into all sampling channels. Then, the response waveform of the square wave is extracted from the data of each sampling channel. The difference between the response waveform of each sampling channel and the injected standard square wave waveform is then compared to accurately calculate the gain error and offset error corresponding to each sampling channel. Finally, all voltage data subsequently acquired by the corresponding sampling channel are corrected in real time based on the error value, thereby eliminating measurement deviations caused by inconsistencies in circuit components.

[0027] Within the dynamic sampling window, the effective cell voltage data corresponding to every two cells are combined, and the corresponding real-time Pearson correlation coefficient matrix is ​​calculated. The cell voltage consistency coefficient is then calculated based on the Pearson correlation coefficient matrix.

[0028] When calculating the cell voltage consistency coefficient, all cells in the battery pack need to be grouped into pairs. Then, a comprehensive index is extracted from the real-time correlation coefficients of all cell pairs. This comprehensive index is then used to calculate the cell voltage consistency coefficient. Specifically: Calculate any two battery cells and Pearson correlation coefficient of effective cell voltage data ; ; in: Indicates the first The energy-saving cell is in the first dynamic sampling window. Voltage value at each sampling time. Indicates the first The energy-saving cell is in the first dynamic sampling window. Voltage value at each sampling time. Indicates the first The average value of N voltage values ​​of the energy-saving cell within the dynamic sampling window. Indicates the first The average value of N voltage values ​​of the energy-saving cell within the dynamic sampling window, where N represents the number of valid sampling points within the sliding window; By iterating through all battery cells, several Pearson correlation coefficients are calculated, and a Pearson correlation coefficient matrix is ​​established. ; ; ; ; Extracting the Pearson correlation coefficient matrix Minimum value of off-diagonal elements in , ; with minimum value C is used as the cell voltage consistency coefficient; like Figure 4 As shown, the cell voltage consistency coefficient is divided into three stages: normal, mild alarm, and severe alarm, based on different threshold ranges of the consistency coefficient. Different countermeasures are then taken for each stage.

[0029] Step 2: Calculate the adaptive voltage difference.

[0030] Considering that the voltage difference of battery cells varies greatly in different SOC ranges during charging, discharging, and resting, voltage consistency cannot be simply determined by a threshold. Therefore, by analyzing the battery cell voltage, current, temperature, SOC, and total voltage data offline, the voltage consistency tolerance of the battery cell under different operating ranges is fitted, and the adaptive voltage difference of the battery cell in different operating ranges is set based on this tolerance.

[0031] Collect relevant parameters of the battery cells in offline mode; these parameters include cell voltage, current, temperature, SOC, and total voltage; fit voltage consistency tolerance curves for each cell across multiple operating ranges based on the relevant parameters; and calculate the tolerance thresholds for each cell in different operating ranges. .

[0032] Specifically: tolerance threshold To determine the dynamic range of SOC and operating conditions, within the post-charge resting range and post-discharge resting range, for the same SOC point, the open-circuit voltage (OCV) of all cells is very close, and its tolerance threshold is... The tolerance threshold is relatively narrow, ranging from a few millivolts to tens of millivolts. Within the charging and discharging ranges, the difference in polarization resistance during charging and discharging leads to a significant increase in the voltage difference between cells. The range is wider, and varies significantly across different SOC ranges. Therefore, tolerance thresholds are calculated segment by segment based on the cell's operating range and SOC range. For example, the SOC range can be divided into 5% segments, with the specific percentage for each segment adjustable based on actual conditions. Specifically, the charging, discharging, post-charging resting, and post-discharging resting periods can each be divided into 5% segments, and the tolerance threshold for each segment can be calculated. ;

[0033] like Figure 2 As shown, the voltage consistency tolerance curves of multiple battery cells are compared with the tolerance threshold. Matching analysis is performed to obtain the consistency tolerance voltage band diagram, and the adaptive voltage difference of the entire cell voltage is determined based on the voltage band diagram.

[0034] Step 3: Anomaly Alarm Identification.

[0035] The cell voltage consistency coefficient C is compared with the corresponding threshold range. If the cell voltage consistency coefficient C falls within the threshold range for a serious alarm, a serious alarm is triggered. The adaptive voltage difference S is compared with the tolerance threshold. For comparison, if the adaptive voltage difference exceeds the tolerance threshold... If so, a serious alarm will be issued; If neither the voltage consistency coefficient C nor the adaptive voltage difference S triggers a severe alarm, the cell may already be experiencing an abnormal condition, for example, within the range of a minor alarm. However, under normal conditions, this range would not trigger an alarm. Therefore, to improve early warning analysis, the cell voltage consistency coefficient C and the adaptive voltage difference S are fused together, and the fused comprehensive data is analyzed. Anomaly alarm analysis is performed, and this comprehensive data can better identify whether the cell's status is within the range of minor alarms. If so, action is taken, and an actual alarm is triggered. Before the analysis, the adaptive voltage difference S and the cell voltage consistency coefficient C are first processed to achieve dimensional consistency: S = ΔV / θ, where ΔV represents the voltage difference between the resting voltage and the standard cell voltage. Then, the dimensional adaptive voltage difference S and the voltage consistency coefficient C are weighted accordingly. α + β = 1, where the values ​​of weights α and β are set according to the operating state of the battery pack. For example, when the battery cells are in a high-dynamic condition (such as high-current discharge, rapid acceleration): increase the value of α (e.g., α=0.8, β=0.2). Because the voltage change is significant at this time, the correlation coefficient is extremely sensitive to anomalies. This ratio can detect the consistent degradation trend earlier and issue an alarm compared to the static range. In static or steady-state conditions (such as static after charging, low-speed cruising): increase the value of β (e.g., α=0.2, β=0.8). Finally, the combined data... The data is compared with the corresponding set threshold. If the comprehensive data Q exceeds the set threshold, a serious alarm is issued.

[0036] S40: Extract the voltage characteristics and periodic signal statistical feature vectors of the charging range, and perform periodic anomaly feature identification.

[0037] Step 1: Extract the voltage characteristics of the charging process.

[0038] like Figure 3 As shown, it contains charging data of all cells in the same battery pack. By extracting all cell voltage data in the charging range and performing filtering, the cell voltage data is converted into an incremental capacity curve. The voltage characteristics corresponding to the main peak of the incremental capacity curve are extracted. The voltage characteristics include peak voltage, peak height, peak range area, and peak voltage span.

[0039] Step 2: Extract the statistical feature vector of the periodic signal.

[0040] The cell voltage data within the charging interval is segmented to obtain a voltage signal with a set period. This set period is a manually defined periodic interval, such as the voltage plateau region and non-plateau region during charging. Multi-dimensional statistical features of the voltage signal are extracted to form a periodic signal statistical feature vector. These multi-dimensional statistical features include features describing the signal baseline and range, features describing data dispersion, features describing data shape, and features describing waveform characteristics. Features describing the signal baseline and range include maximum value, minimum value, mean, and peak value. Features describing data dispersion include the maximum voltage difference ΔVmax between the resting voltage and the standard cell voltage, the standard deviation σ, and the coefficient of variation CV = σ / μ, where μ represents the mean. Features describing data shape include skewness and kurtosis. Features describing waveform characteristics include waveform factor, peak factor, impulse factor, and margin factor.

[0041] Step 3: Identification of periodic anomaly features.

[0042] A: Construct the feature matrix.

[0043] Extract the periodic signal statistical feature vector of each cell; construct a feature matrix X from the periodic signal statistical feature vectors of all cells, and perform data standardization. Since the dimensions and orders of magnitude of each feature are different, such as voltage value and dimensionless kurtosis, standardization can make the mean of each feature 0 and the variance 1, which is convenient for analysis and calculation.

[0044] B: Calculate the characteristic subspace K.

[0045] Calculate the covariance matrix of the feature matrix X after data standardization; perform eigenvalue decomposition on the covariance matrix to obtain several eigenvalues ​​and their corresponding eigenvectors of the principal component directions; select the eigenvalues ​​of the first K principal component directions whose variance accounts for more than 85% to construct the feature subspace K.

[0046] C: Identification of periodic anomaly features.

[0047] Projecting the feature matrix X onto the feature subspace K yields the dimensionality-reduced score matrix T; each row of the score matrix T represents the principal component score of a sample; the score matrix T is then transformed inversely into an N-dimensional space to obtain the reconstructed feature matrix. ; Calculate the characteristic matrix X and the characteristic matrix The T² statistic and the squared prediction error (SPE) are calculated; the T² statistic and the SPE are compared with a preset threshold to identify periodic anomalies.

[0048] The preset threshold is obtained by analyzing and calculating the T² and SPE values ​​of a large amount of sample data. In actual monitoring, for new feature matrices X and X... The calculated T² statistic and squared prediction error (SPE) are compared with the aforementioned preset threshold to determine whether the feature is abnormal.

[0049] S50: Locate abnormal battery cells based on real-time abnormal alarms and periodic abnormal characteristics.

[0050] Extract the operating range corresponding to the real-time abnormal alarm; locate the abnormal voltage time period based on the operating range; read the time window information of the abnormal voltage time period, which includes the start timestamp and the end timestamp; extract the position of the feature value corresponding to the periodic abnormal feature in the feature matrix X; locate the abnormal voltage data based on the position; read the cell identifier corresponding to the abnormal voltage data; locate the abnormal cell based on the time window information and the cell identifier.

[0051] When collecting valid voltage data, each dynamic sampling window corresponds to a specific time window, such as the charging interval from second t1 to second t2. That is, whenever valid voltage data is generated, its metadata includes the start and end timestamps of that data segment.

[0052] For example, the initial data for the battery pack cells is set as follows: 100 cells, with 50 charging cycles. Then, feature extraction is performed on the charging voltage range of each cell in each cycle, resulting in 100 × 50 = 5000 samples. Each sample's metadata includes the cell identifier, cycle number, and time interval. Each sample contains all the aforementioned features. In this case, the number of features is set to 15, thus constructing a 5000 × 15 feature matrix.

[0053] PCA modeling training was performed using normal cyclic data to obtain preset thresholds for T² and SPE values. Then, data from 100 cells in the new cycle (51st cycle) were input into the PCA model to calculate the actual T² and SPE values.

[0054] By comparing the actual T² and SPE values ​​with preset thresholds, the sample SPE value of Cell_023 in the 51st cycle was found to be severely exceeded. Further querying the sample's metadata revealed that the anomaly occurred during the "constant current charging phase, during the voltage rise from 3.30V to 3.35V" in the 51st charge. Thus, the cell causing the abnormal state was ultimately located.

[0055] Example 2: As Figure 5 As shown, this embodiment provides an anomaly detection system for battery cells, including: The acquisition module 101 is used to acquire the operating parameters of the battery pack and perform filtering and cleaning; the operating parameters include current, cell voltage, temperature, and total voltage. The data segmentation module 102 is used to segment the filtered and cleaned operating parameters according to multiple operating intervals of the battery pack; the multiple operating intervals include a charging interval, a discharging interval, a post-discharge resting interval, and a post-charge resting interval; The first processing module 103 is used to calculate the consistency coefficient of the cell voltage; The second processing module 104 is used to calculate the adaptive voltage difference of the battery cell voltage; The third processing module 105 is used to extract the voltage characteristics of the charging range; The fourth processing module 106 is used to extract the statistical feature vector of the periodic signal; The first analysis module 107 is used to identify real-time abnormal alarms based on the consistency coefficient of the cell voltage and the adaptive voltage difference. The second analysis module 108 is used to identify periodic abnormal features based on the voltage characteristics of the charging range and the statistical feature vector of the periodic signal. Alarm module 109 is used to locate abnormal battery cells based on real-time abnormal alarms and periodic abnormal characteristics.

[0056] The battery cell anomaly detection system provided in this embodiment of the invention utilizes a first processing module 103 to calculate the consistency coefficient of the cell voltage, a second processing module 104 to calculate the adaptive voltage difference of the cell voltage, a third processing module 105 to extract voltage features within the charging range, a fourth processing module 106 to extract periodic signal statistical feature vectors, a first analysis module 107 for real-time anomaly alarm identification, a second analysis module 108 for periodic anomaly feature identification, and an alarm module 109 for locating the abnormal cell. Compared with existing technologies, this system can more accurately and efficiently locate abnormal cells.

[0057] The battery cell anomaly detection system provided in this embodiment of the invention can implement the aforementioned method embodiment. For specific functional implementation, please refer to the description in the method embodiment, which will not be repeated here. The battery cell anomaly detection system provided in this embodiment of the invention can be used for battery pack monitoring in energy storage systems, but is not limited to this.

[0058] Example 3: As Figure 6 As shown, this embodiment provides a device 200, which includes a processor 201 and a memory 202; the memory 202 stores a set of program instructions; when the set of program instructions stored in the memory 202 is loaded and executed by the processor 201, the aforementioned abnormal detection method for battery cells can be implemented.

[0059] Device 200 is an electronic device connected to battery pack system 300, including at least a processor 201 and a memory 202 connected to the processor 201. The processor 201 is typically a general-purpose computer processor capable of executing computer program instructions. The memory 202 is typically a medium that can retain data even after power loss, including but not limited to disks, magnetic tapes, and flash memory. The memory 202 is typically used to store computer program instruction sets and data. The processor 201 implements its corresponding automation functions by loading the program instruction set stored in the memory 202. Specifically, in this embodiment, the processor 201 implements the aforementioned abnormal detection method for battery cells by loading and executing the program instruction set stored in the memory 202.

[0060] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for detecting anomalies in battery cells, characterized in that, Includes the following steps: The operating parameters of the battery pack are collected and filtered and cleaned; the operating parameters include current, cell voltage, temperature, and total voltage. The operating parameters after filtration and cleaning are divided into multiple operating ranges of the battery pack; the multiple operating ranges include a charging range, a discharging range, a post-discharge resting range, and a post-charge resting range; Based on multiple operating ranges, the cell voltage consistency coefficient and adaptive voltage difference are calculated, and real-time abnormal alarm identification is performed. Collect relevant parameters of the battery cell in offline mode; the relevant parameters include battery cell voltage, current, temperature, SOC, and total voltage; Based on the relevant parameters of each cell, the voltage consistency tolerance curve of the cell under multiple operating ranges was fitted. Calculate the tolerance threshold of the battery cell under different operating ranges. ; The voltage difference ΔV between multiple cells is compared with the tolerance threshold. Calculations are performed to obtain the adaptive voltage difference S of the entire cell voltage, S = ΔV / , where ΔV represents the voltage difference between the resting voltage and the standard cell voltage; The cell voltage consistency coefficient C is compared with the corresponding threshold range. If the cell voltage consistency coefficient C does not meet the threshold range, a serious alarm is issued. Compare the voltage difference ΔV with the tolerance threshold For comparison, if the voltage difference ΔV does not meet the tolerance threshold... If so, a serious alarm will be issued; If neither the cell voltage consistency coefficient C nor the voltage difference ΔV is in serious alarm state, the cell voltage consistency coefficient C and the adaptive voltage difference S are fused together, and the fused comprehensive data Q is compared with the corresponding set threshold. If it exceeds the set threshold, an abnormal alarm is triggered. Extract voltage features and periodic signal statistical feature vectors from the charging range, and identify periodic anomalies. Abnormal battery cells can be located based on real-time abnormal alarms and periodic abnormal characteristics.

2. The method for detecting abnormalities in a battery cell according to claim 1, characterized in that: The cell voltage consistency coefficient is calculated as follows: Extract cell voltage data within the charging and discharging ranges of the battery pack, and set a dynamic sampling window to collect all valid cell voltage data; Within the dynamic sampling window, the effective cell voltage data corresponding to every two cells are combined, and the corresponding real-time Pearson correlation coefficient matrix is ​​calculated. The cell voltage consistency coefficient is then calculated based on the Pearson correlation coefficient matrix.

3. The method for detecting abnormalities in a battery cell according to claim 2, characterized in that: Within the dynamic sampling window, the effective cell voltage data corresponding to every two cells are combined, and the corresponding real-time Pearson correlation coefficient matrix is ​​calculated. Based on the Pearson correlation coefficient matrix, the cell voltage consistency coefficient is calculated, specifically as follows: Calculate any two battery cells and Pearson correlation coefficient of effective cell voltage data ; ； in: Indicates the first The energy-saving cell is in the first dynamic sampling window. Voltage value at each sampling time. Indicates the first The energy-saving cell is in the first dynamic sampling window. Voltage value at each sampling time. Indicates the first The average value of N voltage values ​​of the energy-saving cell within the dynamic sampling window. Indicates the first The average value of N voltage values ​​of the energy-saving cell within the dynamic sampling window, where N represents the number of valid sampling points within the sliding window; By iterating through all battery cells, several Pearson correlation coefficients are calculated, and a Pearson correlation coefficient matrix is ​​established. ; ； ； ； Extracting the Pearson correlation coefficient matrix Minimum value of off-diagonal elements in , ； With the minimum value C is used as the cell voltage consistency coefficient.

4. The method for detecting abnormalities in a battery cell according to claim 3, characterized in that: Extracting the voltage characteristics of the charging range, specifically: Extract the cell voltage data during the charging period and perform filtering processing; The cell voltage data is converted into an incremental capacity curve, and the voltage characteristics corresponding to the main peak of the incremental capacity curve are extracted. The voltage characteristics include peak voltage, peak height, peak area, and peak voltage span.

5. The method for detecting abnormalities in a battery cell according to claim 4, characterized in that: The statistical feature vector of the periodic signal is extracted as follows: The cell voltage data in the charging range is segmented to obtain a voltage signal with a set period. Multi-dimensional statistical features of voltage signals are extracted to form a periodic signal statistical feature vector. The multi-dimensional statistical features include features describing the signal reference and range, features describing the data dispersion, features describing the data shape, and features describing the waveform characteristics. The characteristics describing the reference and range of the signal include the maximum value, minimum value, mean value, and peak value; the characteristics describing the dispersion of the data include the maximum voltage difference ΔVmax between the resting voltage and the standard cell voltage, the standard deviation σ, and the coefficient of variation CV=σ / μ, where μ represents the mean value; the characteristics describing the shape of the data include skewness and kurtosis; the characteristics describing the waveform characteristics include waveform factor, peak factor, impulse factor, and margin factor.

6. The method for detecting abnormalities in a battery cell according to claim 5, characterized in that: The periodic anomaly feature identification specifically includes: Extract the periodic signal statistical feature vector for each battery cell; The periodic signal statistical feature vectors of all battery cells are used to construct a feature matrix X, and then the data is standardized. Calculate the covariance matrix of the feature matrix X after data standardization. The covariance matrix is ​​decomposed into eigenvalues ​​to obtain several eigenvalues ​​and their corresponding eigenvectors of the principal component directions; Select the eigenvalues ​​of the first K principal component directions whose variance accounts for more than 85%, and construct the eigenspace K; Projecting the feature matrix X onto the feature subspace K yields the dimension-reduced score matrix T; each row of the score matrix T represents the principal component score of a sample. Transform the score matrix T in reverse into N-dimensional space to obtain the reconstructed feature matrix. ; Calculate the characteristic matrix X and the characteristic matrix The T² statistic and the squared prediction error; The T² statistic and the squared prediction error are compared with preset thresholds to identify periodic anomalies.

7. The method for detecting abnormalities in a battery cell according to claim 6, characterized in that: The abnormal battery cell is located based on the real-time abnormal alarm and periodic abnormal characteristics, specifically as follows: Extract the operating range corresponding to real-time anomaly alarms; Locate the time period of abnormal voltage based on the operating range; Read the time window information of the abnormal voltage period, the time window information including the start timestamp and the end timestamp; Extract the position of the eigenvalue corresponding to the periodic anomaly feature in the feature matrix X; Locate abnormal voltage data based on location; Read the cell identifier corresponding to the abnormal voltage data; The abnormal battery cell can be located based on the time window information and the cell identifier.

8. A battery cell anomaly detection system employing the anomaly detection method for battery cells according to any one of claims 1-7, characterized in that: include, The data acquisition module is used to collect the operating parameters of the battery pack and perform filtering and cleaning; the operating parameters include current, cell voltage, temperature, and total voltage. The data segmentation module is used to segment the filtered and cleaned operating parameters according to multiple operating intervals of the battery pack; the multiple operating intervals include a charging interval, a discharging interval, a post-discharge resting interval, and a post-charge resting interval; The first processing module is used to calculate the consistency coefficient of the cell voltage; The second processing module is used to calculate the adaptive voltage difference of the battery cell voltage; The third processing module is used to extract the voltage characteristics of the charging range; The fourth processing module is used to extract statistical feature vectors of periodic signals; The first analysis module is used to identify real-time abnormal alarms based on the consistency coefficient of the cell voltage and the adaptive voltage difference. The second analysis module is used to identify periodic abnormal features based on the voltage characteristics of the charging range and the statistical feature vector of the periodic signal. The alarm module is used to locate abnormal battery cells based on real-time abnormal alarms and periodic abnormal characteristics.

9. An apparatus comprising a processor and a memory, characterized in that: The memory stores a set of program instructions; when the set of program instructions stored in the memory is loaded and executed by the processor, the abnormal detection method of the battery cell according to any one of claims 1-7 is implemented.

Images