Method for judging abnormal attenuation of battery cells of battery pack

Abstract

The invention belongs to the field of battery anomaly detection, and particularly relates to a method for judging abnormal attenuation of battery cells of a battery pack, which comprises the following steps of: S1, acquiring vehicle operation data from a cloud data platform, removing charging data which does not conform to a preset condition in a full life cycle, and grouping and sequencing the charging data according to a charging working condition; s2, preprocessing each piece of charging data, and calculating the charging process capacity corresponding to each charging working condition; s3, respectively calculating the characteristic capacity difference of the charging data of each battery cell under the same charging condition as a voltage characteristic value; s4, calculating the outlier degree of each battery cell under the same charging condition according to the characteristic capacity difference to obtain an outlier degree table of each battery cell; and S5, calculating the variation trend of the outlier degree in the charging working condition meeting the preset condition in the whole period, and analyzing the abnormal condition of the battery cell. The problem that the abnormal attenuation judgment result of the battery cell is unreliable is solved.

Description

Technical Field

[0001] The present invention belongs to the field of battery abnormality detection, and in particular relates to a method for determining abnormal attenuation of battery cells in a battery pack. Background Art

[0002] As the number of new energy vehicles continues to grow, attention to their safety performance is also increasing. As one of the most important components of new energy vehicles, battery safety is particularly important. However, with the increase in usage time and the influence of factors such as driving behavior, battery failure issues have gradually become prominent, which not only affects the user's driving experience, but may also endanger the safety of the car itself.

[0003] The current methods for judging battery pack abnormalities have the following technical problems: (1) Existing judgment methods are mainly based on capturing feature information of a single fragment, and usually do not consider the changing trend of the outlier phenomenon of a single battery cell in the entire pack over the entire cycle, which can easily lead to false alarms or missed alarms; (2) Existing judgment methods are usually calculated based on capacity. This method usually requires charging data with a long SOC change and stable charging current as a basis, but there is less such data on actual vehicles, resulting in poor algorithm coverage and feasibility; (3) Existing judgment methods mainly rely on the absolute value of capacity, which varies with the type of battery pack, resulting in poor generalization ability. In addition, since batteries will also decay during normal use, and batteries with safety risks are mostly single cells in the battery pack with abnormal decay, the judgment method based on the absolute value of capacity cannot accurately identify cells with abnormal decay. Summary of the invention

[0004] The present invention provides a method for judging abnormal attenuation of battery cells of a battery pack, which solves the problem that the judgment result of abnormal attenuation of battery cells is unreliable.

[0005] The basic solution provided by the present invention is a method for judging abnormal attenuation of a battery pack cell, comprising:

[0006] S1: Obtain vehicle operation data from the cloud data platform, remove charging data that does not meet the preset conditions in the entire life cycle, and group and sort the charging data according to charging conditions;

[0007] S2: pre-process each piece of charging data and calculate the capacity corresponding to each charging condition;

[0008] S3: Calculate the voltage characteristic values corresponding to each charging data under the same charging condition respectively. The method for calculating the voltage characteristic values is as follows:

[0009] a. Establish a sliding window at equal voltage intervals for the monomer voltage sequence in the charging data to calculate the capacity change value under the voltage interval, and obtain a capacity change value group through continuous sliding windows;

[0010] b. Calculate the capacity difference of the charging data, and use the capacity difference as the voltage characteristic value. The calculation formula for the capacity difference is as follows:

[0011] dC = C1 - C0

[0012] In the formula, dC is the capacity difference, C0 is the capacity corresponding to the set initial voltage, and C1 is the capacity corresponding to the minimum value of the capacity change in the capacity change value group;

[0013] The set initial voltage is set according to the battery type and charging current;

[0014] S4: Calculate the outlier degree of each battery cell under the same charging condition according to the voltage characteristic value, and obtain the outlier degree table of each battery cell;

[0015] S5: Calculate the change trend of the outlier degree in the charging conditions that meet the preset conditions in the full cycle, and analyze the abnormal situation of the battery cells.

[0016] The principle and advantages of the present invention are as follows: By obtaining vehicle operation data from the cloud, screening the charging data that meets the preset conditions in the full life cycle, and grouping and sorting according to the charging conditions for subsequent analysis and judgment, even short-term or unstable charging segments can be effectively utilized, ensuring that the data covers the complete cycle and the accuracy of the judgment result is relatively high; By calculating the capacity difference at equal voltage intervals through a sliding window, directly extracting features using the voltage sequence, avoiding dependence on long-term stable charging data, compared with the existing capacity-based judgment algorithm, it has low requirements for the SOC change, that is, the charging current stability charging data, and more data can be involved in the judgment method, and the algorithm coverage and feasibility are relatively high; When performing outlier degree trend analysis, the outlier degree of each battery cell under each charging condition is calculated to form an outlier degree table of the battery cells, so as to be able to analyze the change trend of the outlier degree of all battery cells under each condition rather than a single-point judgment, accurately capture the dynamic process of abnormal attenuation of the battery cells, distinguish normal attenuation from abnormal attenuation, improve the sensitivity and specificity to the abnormality of a single battery cell, reduce false alarms or missed alarms, and further improve the accuracy and reliability of the judgment result; Identify abnormalities through the relative outlier degree between battery cells under the same condition, avoiding the problem of threshold failure caused by battery type or overall attenuation

[0017] This solution uses three core technologies: full-cycle data integration, voltage characteristic value replacing capacity, and outlier degree dynamic analysis, to solve the core problems of fragmented false alarms, high data dependence, and poor threshold generalization ability in the background technology respectively, thus realizing more accurate, robust and practical battery anomaly detection.

[0018] Preferably, in S1, the preset condition includes that the SOC change value of the charging data is not less than 20%.

[0019] Beneficial effects: By selecting charging data according to preset conditions, sufficient SOC changes can make the data more accurately reflect the charge and discharge characteristics of the battery, thus providing a more reliable data basis for subsequent analysis and judgment, and reducing misjudgments or inaccurate judgments caused by incomplete or unrepresentative data; at the same time, it is convenient for the algorithm to capture the characteristic information of the battery at different charging stages, avoid large fluctuations in the algorithm's judgment of the battery state due to overly limited data, help improve the accuracy and reliability of the algorithm's judgment of abnormal attenuation of battery pack cells, and enhance the stability and robustness of the entire judgment method.

[0020] Preferably, in S2, the preprocessing includes deleting abnormal characters and invalid data, and selecting the single-cell voltage sequence, current, SOC, and temperature.

[0021] Beneficial effects: Deleting abnormal characters and invalid data can remove noise and error information in the data, make the data cleaner and more accurate, help avoid interference caused by abnormal or invalid data to subsequent calculations and analysis, improve the accuracy and reliability of the entire judgment method, and focus on key information such as the single-cell voltage sequence, current, SOC, and temperature, enabling more targeted judgment of abnormal attenuation of battery pack cells, improving the efficiency and accuracy of judgment, and avoiding the processing of a large amount of irrelevant data, thereby reducing the complexity and computational amount of data processing, increasing the running speed of the algorithm, reducing the requirements for hardware devices, and reducing system costs.

[0022] Preferably, in S4, at least two of the interquartile range method, three times standard deviation rule, standardized score method, and median outlier method are used to calculate the outlier degree of the battery cells.

[0023] Beneficial effects: By cross-analyzing the output results of different methods, the reasons for abnormalities (such as consistency deterioration, single-cell failure, etc.) can be clarified; and it adapts to different battery working conditions: in different scenarios such as fast charging, slow charging, high and low temperatures, the distribution of the characteristic values of the battery cell voltage may be different, and the combination of multiple methods can dynamically adapt to the data characteristics and improve the generalization ability.

[0024] Preferably, the results of various outlier degree calculation methods are weighted and summed for fusion to obtain the final outlier degree. Among them, the weights of the outlier degree algorithms are trained through machine learning algorithms to generate a weight mapping table. When judging the attenuation of the battery cells, the weights are determined according to the feature matching of the current charging working condition.

[0025] Beneficial effects: Through the pre-trained weight mapping table, the optimal weight combination can be automatically matched according to the current, temperature, and voltage distributions in the charging data. The weight mapping table is pre-generated through historical data and does not require online operation of complex models, which is conducive to improving the speed of the entire method.

[0026] Preferably, in S3, the set initial voltage is 3.5 - 3.7V.

[0027] Beneficial effects: The initial voltage range is determined to be 3.5 - 3.7V, which not only includes the key voltage stage during the normal charging start of the battery, but also avoids the excessive computational complexity or information redundancy caused by selecting too high or too low voltages. The interference of current fluctuations within this range to the voltage-capacity relationship is smaller, ensuring the stability of capacity difference calculation.

[0028] Preferably, in S5, the specific steps are as follows:

[0029] S5-1) Calculate the difference between the outlier degree of the battery cell in the last three working conditions and the outlier degree of the battery cell in the first three working conditions;

[0030] S5-2) Compare the difference with the set outlier degree threshold. If it exceeds the set outlier degree threshold, it is considered that the battery cell has the risk of accelerated abnormal attenuation.

[0031] Beneficial effects: By calculating the difference between the outlier degrees of the battery cell in the last three working conditions and the first three working conditions, the change trend of the outlier degree of the battery cell over time or working conditions can be effectively captured, and the abnormal changes in the state of the battery cell can be detected in a timely manner, rather than just focusing on the outlier situation at a certain moment, so as to more comprehensively and dynamically evaluate the health status of the battery cell.

[0032] Further preferably, in S5-2), the set outlier degree threshold is 3 - 5.

[0033] Beneficial effects: A reasonably set outlier degree threshold can detect the abnormal situation of the battery cell in a timely manner, avoid the further deterioration of the problem, and ensure the safe and stable operation of the battery system.

[0034] Preferably, in S3, the width of the sliding window is 0.01 - 0.1V.

[0035] Beneficial effects: Setting the width of the voltage sliding window to 0.01 - 0.1V ensures that the length of the voltage window is appropriate, reduces the number of redundant calculations, and at the same time avoids the distortion of the calculated capacity difference caused by too long a voltage length. Description of the Drawings

[0036] Figure 1 is the flow chart of the present invention;

[0037] Figure 2 is the schematic diagram of the principle for calculating the capacity difference of the present invention;

[0038] Figure 3 is the trend chart of the change of the outlier degree of the present invention. Detailed Embodiments

[0039] The following is a further detailed description through specific embodiments:

[0040] Example 1

[0041] The specific implementation process is as follows: Refer to Figure 1 , a method for judging abnormal attenuation of battery pack cells, including:

[0042] S1: Obtain vehicle operation data from the cloud data platform, remove charging data that does not meet the preset conditions throughout the life cycle, group the charging data that meets the preset conditions according to the charging conditions, and sort each group according to the driving mileage or time. In this embodiment, the charging condition is the number of charge-discharge cycles experienced by the battery pack. Specifically, the charging data is grouped according to one charge-discharge cycle of the battery pack, and its flag bit is from the start of charging to the end of charging (corresponding to "charging status" == 1 in the national standard 32960 data), denoted as process_1 to process_m. Taking ternary battery cells as an example, each group is sorted according to the driving mileage;

[0043] In S1, the preset conditions include that the SOC change value of the charging data is not less than 20%.

[0044] S2: Preprocess each piece of charging data and calculate the capacity corresponding to each charging condition;

[0045] In S2, the preprocessing includes deleting abnormal characters and invalid data, and selecting the single-cell voltage sequence, current, SOC, and temperature.

[0046] The capacity calculation formula is as follows:

[0047]

[0048] In the formula, C is the capacity, t is the time, and I is the charging current;

[0049] S3: Calculate the voltage characteristic values corresponding to each piece of charging data under the same charging condition respectively. The method for calculating the voltage characteristic values is as follows:

[0050] a. Establish a sliding window for the single-cell voltage sequence in the charging data at equal voltage value intervals to calculate the capacity change value d under this voltage interval, and obtain a group of capacity change values list_d through continuous sliding windows;

[0051] The width of the sliding window is 0.01 - 0.1V. In this embodiment, the width of the sliding window is 0.1V.

[0052] b. Calculate the capacity difference of the charging data and use the capacity difference as the voltage characteristic value,

[0053] b1. Find the minimum capacity change value in the group of capacity change values list_d and record the corresponding sliding window start voltage V1 and capacity C1;

[0054] b2. Obtain the voltage value V0 of the set initial voltage and its corresponding capacity C0 in the charging data;

[0055] b3. The capacity difference calculation formula is as follows:

[0056] dC = C1 - C0

[0057] In the formula, dC is the capacity difference, C0 is the capacity corresponding to the set initial voltage, and C1 is the capacity corresponding to the minimum capacity change value in the capacity change value group;

[0058] The set initial voltage is set according to the battery type and the charging current in the charging data;

[0059] In S3, the set initial voltage is 3.5 - 3.7V. In this embodiment, the set initial voltage is 3.65V. In other embodiments, the set initial voltage is 3.5V.

[0060] As Figure 2 shown, it is a schematic diagram of the principle for calculating the capacity difference of the present invention. Among them, the abscissa is the charging capacity, the ordinate is the capacity difference calculated by the equal voltage window, and the gray curve clusters represent different battery cells. In the figure, the initial voltage V0 and the capacity C0 corresponding to the initial voltage are the initial characteristic points of the initial charge, reflecting the start stage of the electrochemical reaction, while the voltage V1 and the capacity C1 reflect the peak stage of the electrochemical reaction. Therefore, the capacity difference dC can reflect the characteristic capacity attenuation process.

[0061] S4: Calculate the outlier degree of each battery cell under the same charging condition according to the voltage characteristic value to obtain an outlier degree table for each battery cell;

[0062] In S4, at least two methods among the interquartile range method, the three - standard - deviation rule, the standardized score method, and the median outlier method are used to calculate the outlier degree of the battery cells, and the results of multiple outlier degree calculation methods are fused to obtain the final outlier degree;

[0063] The following formula is used for fusion:

[0064]

[0065] In the formula, O final is the final outlier degree, n is the number of outlier degree algorithms used, O i is the calculation result of the i - th outlier degree algorithm, and ω i is the weight of the i - th outlier degree algorithm;

[0066] The weights of the outlier degree algorithms are trained through machine learning algorithms to generate a weight mapping table. When judging the attenuation of the battery cells, the pre - stored weights are matched according to the current working condition characteristics.

[0067] In this embodiment, the standardized score method and the median outlier method are adopted. Specifically, the fusion formula is expressed as

[0068] O final1 = ω1O1 + ω2O2

[0069] In the formula, O final1 is the final outlier degree after fusion, O1 is the calculation result of the standardized score method, ω1 is the weight of the standardized score method, O2 is the calculation result of the median outlier method, and ω2 is the weight of the median outlier method;

[0070] Among them, the weight ω1 of the standardized score method and the weight ω2 of the median outlier method generate a weight mapping table through a BP neural network, which is used to query and match the weights during judgment;

[0071] In the BP neural network: the input layer takes the average current of the charging data, the skewness of the voltage curve, the standard deviation of the capacity between battery cells, and the ambient temperature as input parameters, and four neurons are set; the output layer corresponds to the output parameters of the weight of the standardized score method and the weight of the median outlier method, and two neurons are set, and the Softmax function is used as the activation function.

[0072] The calculation formula of the median outlier method for the outlier degree of battery cells

[0073]

[0074] In the formula, outier_n is the outlier degree of the battery cell, and dCn is the voltage eigenvalue.

[0075] S5: Calculate the change trend of the outlier degree and analyze the abnormal situation of the battery cell.

[0076] In S5, the specific steps are as follows:

[0077] S5-1) Calculate the difference between the outlier degree of the battery cell in the last three working conditions and the outlier degree of the battery cell in the first three working conditions;

[0078] S5-2) Compare the difference with the set outlier degree threshold. If it exceeds the set outlier degree threshold, it is considered that the battery cell has the risk of accelerated abnormal attenuation.

[0079] In S5-2), the set outlier degree threshold is 3-5. In this embodiment, the outlier degree threshold is 3.

[0080] The outlier degree threshold is obtained by collecting the data of abnormal batteries and statistically analyzing the outlier degree distribution law of abnormal batteries, so as to determine the outlier degree threshold adopted in the test method.

[0081] The statistical distribution method for analyzing the outlier degree threshold can adopt the standard deviation method, the interquartile range, and clustering methods such as the KMeans and DBSCAN algorithms. In this embodiment, it is determined by the interquartile range method, and the set threshold is 3, so as to ensure that 99.8% of the data is within 3σ and reduce the false alarm risk.

[0082] In this embodiment, as Figure 3 shown, it is a graph of the outlier degree change trend, where the abscissa is time and the ordinate is the average outlier degree calculated adjacent. The initial outlier degree of the 20th battery cell is 0.7, and the final outlier degree is -3.8. The absolute value of the change rate is 4.5, exceeding the set change rate threshold. Therefore, the 20th battery cell is an abnormally decaying battery cell.

[0083] Embodiment 2

[0084] In this embodiment, the three - times standard deviation method and the median outlier method are adopted. Specifically, the fusion formula is expressed as

[0085] O final2 = ω3O3 + ω4O4

[0086] In the formula, O final2 is the final outlier degree after fusion, O3 is the calculation result of the three - times standard deviation method, ω3 is the weight of the three - times standard deviation method, O4 is the calculation result of the median outlier method, and ω4 is the weight of the median outlier method;

[0087] Among them, ω3 is the weight of the three - times standard deviation method, and ω4 is the weight of the median outlier method. Similarly, a weight mapping table is generated through the BP neural network for querying and matching weights during judgment.

[0088] The above are only embodiments of the present invention. Specific structures and common knowledge such as characteristics well - known in the art are not described in detail here. Those of ordinary skill in the art know all the common general technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can know all the existing technologies in this field, and have the ability to apply conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given in this application, combine their own abilities to improve and implement this solution. Some typical well - known structures or well - known methods should not become obstacles for those of ordinary skill in the art to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, which will not affect the implementation effect of the present invention and the practicality of the patent. The protection scope required by this application should be based on the content of its claims, and the specific implementation manners and the like recorded in the specification can be used to explain the content of the claims.

Claims

1. A method for judging abnormal attenuation of battery pack cells, characterized in that, Including: S1: Obtain vehicle operation data from the cloud data platform, remove charging data that does not meet the preset conditions throughout the life cycle, group and sort the charging data according to the charging conditions. S2: Preprocess each piece of charging data and calculate the capacity corresponding to each charging condition. S3: Calculate the voltage characteristic values corresponding to each piece of charging data under the same charging condition respectively. The method for calculating the voltage characteristic values is as follows: a. Establish a sliding window for the single-cell voltage sequence in the charging data at equal voltage value intervals to calculate the capacity change value under this voltage interval, and obtain a group of capacity change values through continuous sliding windows. b. Calculate the capacity difference of the charging data, and use the capacity difference as the voltage characteristic value. The capacity difference calculation formula is as follows: dC = C1 - C0 In the formula, dC is the capacity difference, C0 is the capacity corresponding to the set initial voltage, and C1 is the capacity corresponding to the minimum capacity change value in the group of capacity change values. The set initial voltage is set according to the battery type and charging current. S4: Calculate the outlier degree of each cell under the same charging condition according to the voltage characteristic value, and obtain an outlier degree table for each cell. S5: Calculate the change trend of the outlier degree in the charging conditions that meet the preset conditions throughout the cycle, and analyze the abnormal conditions of the cells.

2. The method for judging abnormal attenuation of a battery pack cell according to claim 1, wherein: In S1, the preset conditions include that the SOC change value of the charging data is not less than 20%.

3. The method for judging abnormal attenuation of battery pack battery cells according to claim 1, wherein: In S2, the preprocessing includes deleting abnormal characters and invalid data, and selecting the single-cell voltage sequence, current, SOC, and temperature.

4. The method for judging abnormal attenuation of battery pack battery cells according to claim 1, characterized in that: In S4, at least two of the interquartile range method, three times the standard deviation rule, standardized score method, and median outlier method are used to calculate the outlier degree of the cells.

5. The method for judging abnormal attenuation of battery pack cells according to claim 4, characterized in that: The results of various outlier degree calculation methods are weighted and summed for fusion to obtain the final outlier degree. Among them, the weights of the outlier degree algorithms are trained through machine learning algorithms to generate a weight mapping table. When judging the cell attenuation, the weights are determined according to the feature matching of the current charging condition.

6. The method for judging abnormal attenuation of battery pack battery cells according to claim 1, characterized in that: In S3, the set initial voltage is 3.5 - 3.7V.

7. The method for judging abnormal attenuation of battery pack battery cells according to claim 1, characterized in that: In S5, the specific steps are as follows: S5-1) Calculate the difference between the outlier degree of the cells in the last three working conditions and the outlier degree of the cells in the first three working conditions. S5-2) Compare the difference with the set outlier degree threshold. If it exceeds the set outlier degree threshold, it is considered that the cell has a risk of accelerated abnormal attenuation.

8. The method for judging abnormal attenuation of battery pack battery cells according to claim 7, characterized in that: In S5-2), the set outlier degree threshold is 3 - 5.

9. The method for judging abnormal attenuation of battery pack cells according to claim 1, wherein: In S3, the width of the sliding window is 0.01 - 0.1V.

Images