An abnormal battery cell screening method, device and equipment based on a pre-charging process

Abstract

The present invention discloses a method, device and equipment for screening abnormal battery cells based on a pre-charging process. The method includes: extracting a plurality of non-overlapping sub-samples in time from the pre-charging voltage data of the currently to-be-detected battery cell by using a plurality of time-equal windows, where the pre-charging voltage data is the relationship data between the battery cell voltage and time during the pre-charging process; extracting the differential eigenvalue corresponding to each sub-sample based on the differential operation between the sampled voltage values within each sub-sample; taking the differential eigenvalue corresponding to each sub-sample as the coordinate of the feature space, and extracting the element corresponding to the coordinate from the feature space as the data feature of the currently to-be-detected battery cell, where the dimension of the feature space is equal to the number of time-equal windows; and determining whether the currently to-be-detected battery cell contains foreign matter based on the data feature. The technical solution provided by the present invention improves the detection accuracy of abnormal battery cells with foreign matter.

Description

Technical Field

[0001] The present invention relates to the field of defect detection of battery cells, and in particular to a method, device and equipment for screening abnormal battery cells based on a pre-charging process. Background Art

[0002] Safety issues are the main obstacles preventing the large-scale application of lithium-ion batteries in electric vehicles. With the continuous popularization of the application of lithium-ion batteries, improving their safety has become increasingly urgent for the development of electric vehicles. Among all lithium-ion battery safety accidents, thermal runaway is the most serious one, which can cause the lithium-ion battery to catch fire or even explode, directly threatening the safety of users.

[0003] There are many factors leading to thermal runaway, which can generally be divided into two categories: internal factors and external factors. The internal factors mainly include: ① internal short circuit caused by battery production defects; ② improper use of the battery, resulting in the generation of lithium dendrites inside, causing short circuit between the positive and negative electrodes. The external factors mainly include: ① external factors such as extrusion and needle puncture cause short circuit of the lithium-ion battery; ② external short circuit of the battery causes excessive heat accumulation inside the battery; ③ too high external temperature causes the decomposition of the SEI film and the positive electrode material, etc. Among them, the main reasons for battery production defects are foreign objects, diaphragm wrinkles, and electrode tab folding, and metal foreign objects are the most main reason. If a metal foreign object is located outside the electrode group, it may pierce the protective tape outside the electrode group, and then make the positive and negative electrode plates conduct with the positive electrode housing, thus leading to thermal runaway. The prior art mainly detects whether there are foreign objects in the battery cell by analyzing the difference in trace element composition and charge-discharge power on the outer surface of the battery cell, but the accuracy of the prior art detection method still needs to be improved. Summary of the Invention

[0004] In view of this, the embodiments of the present invention provide a method, device and equipment for screening abnormal battery cells based on a pre-charging process, thereby improving the detection accuracy of abnormal battery cells with foreign objects.

[0005] According to a first aspect, the present invention provides a method for screening abnormal battery cells based on a pre-charging process, the method comprising: extracting a plurality of non-overlapping sub-samples in time from the pre-charging voltage data of the currently to-be-detected battery cell by using a plurality of time-equal windows, the pre-charging voltage data being the relationship data between the battery cell voltage and time in the pre-charging process, and the time length range of the time-equal windows being 200 - 500 seconds; extracting the differential eigenvalue corresponding to each sub-sample based on the differential operation between the sampled voltage values within each sub-sample; using the differential eigenvalues corresponding to the respective sub-samples as the coordinates of the feature space, and extracting the elements corresponding to the coordinates from the feature space as the data features of the currently to-be-detected battery cell, the dimension of the feature space being equal to the number of the time-equal windows; and determining whether the currently to-be-detected battery cell contains a foreign object based on the data features.

[0006] Optionally, extracting the differential feature value corresponding to each sub-sample based on the differential operation between the sampled voltage values within each sub-sample includes: sampling multiple voltage values at equal time intervals within the current sub-sample; calculating the differences between adjacent voltage values to obtain multiple voltage differences; calculating the mean of the multiple voltage differences, and using the obtained mean as the differential feature value of the current sub-sample.

[0007] Optionally, determining whether the current battery cell to be detected contains foreign matter based on the data features includes: inputting the data features into a pre-trained binary classification model, and determining whether the current battery cell to be detected contains foreign matter through the output result of the pre-trained binary classification model; wherein, the pre-trained binary classification model is generated by training with the data features corresponding to a number of positive samples and a number of negative samples, the negative samples are the pre-charging voltage data of abnormal battery cells with foreign matter, and the positive samples are the pre-charging voltage data of normal battery cells.

[0008] Optionally, the steps of obtaining the positive samples and the negative samples include: obtaining abnormal battery cells known to have foreign matter, and tracing the pre-charging voltage data of the abnormal battery cells known to have foreign matter as negative samples; constructing similar pre-charging voltage data based on the abnormal battery cells known to have foreign matter as negative samples; obtaining known normal battery cells, and tracing the pre-charging voltage data of the known normal battery cells as positive samples; obtaining dissimilar battery cells whose internal structure has a structural similarity lower than a first preset threshold with the abnormal battery cells known to have foreign matter, and tracing the pre-charging voltage data of the dissimilar battery cells as positive samples.

[0009] Optionally, constructing similar pre-charging voltage data based on the abnormal battery cells known to have foreign matter as negative samples includes: obtaining similar battery cells whose internal structure has a structural similarity higher than a second preset threshold with the abnormal battery cells known to have foreign matter, and tracing the pre-charging voltage data of the similar battery cells as negative samples; constructing similar pre-charging voltage data based on the abnormal battery cells known to have foreign matter as negative samples by the SMOTE oversampling method.

[0010] Optionally, before training and generating the binary classification model with a number of positive samples and a number of negative samples, the method further includes: removing the relationship data between the charging voltage and time corresponding to the negative pressure pumping step and the standing step in the pre-charging voltage data, and performing time alignment; removing the positive samples and negative samples with rework.

[0011] Optionally, the pre-charging voltage data selects the data of the battery cell at 1500 seconds - 2000 seconds in the pre-charging process.

[0012] According to a second aspect, an embodiment of the present invention provides an abnormal battery cell screening device based on a pre-charging process. The device includes: a partitioning unit configured to extract a plurality of non-overlapping sub-samples in time from the pre-charging voltage data of a currently to-be-detected battery cell by using a plurality of time-equal windows, where the pre-charging voltage data is the relationship data between the battery cell voltage and time during the pre-charging process, and the time length range of the time-equal windows is 200 - 500 seconds; a difference unit configured to extract difference eigenvalue corresponding to each sub-sample based on the difference operation between the sampled voltage values within each sub-sample; a feature extraction unit configured to use the difference eigenvalue corresponding to each sub-sample as the coordinates of a feature space, and extract the elements corresponding to the coordinates from the feature space as the data features of the currently to-be-detected battery cell, where the dimension of the feature space is equal to the number of the time-equal windows; a detection unit configured to determine whether the currently to-be-detected battery cell contains foreign matter based on the data features.

[0013] According to a third aspect, an embodiment of the present invention provides an abnormal battery cell screening device based on a pre-charging process, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the method described in the first aspect or any optional implementation manner of the first aspect.

[0014] According to a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the method described in the first aspect or any optional implementation manner of the first aspect.

[0015] The technical solution provided by this application has the following advantages:

[0016] It is found through research on the technical solution provided by this application that there are obvious differences in the pre-charging voltage data between normal battery cells and abnormal battery cells with foreign matter. Among them, the voltage rising speed of abnormal battery cells during the pre-charging stage is significantly lower than that of normal battery cells. Based on this, the pre-charging voltage data of the battery cells is sampled according to multiple time-equal windows, and then the difference eigenvalue corresponding to each window is calculated by using the difference operation. Then, each difference eigenvalue is mapped as spatial coordinates into a multi-dimensional space, so as to extract the elements corresponding to the coordinates in the multi-dimensional space as the data features of the battery cells, further deepening the difference between the data features of normal battery cells and abnormal battery cells. Therefore, the normal battery cells and abnormal battery cells are discriminated based on the obtained data features, further improving the accuracy of abnormal battery cell screening.

[0017] In addition, based on the data features extracted by the above means, a binary classification model was trained in combination with data classification models such as neural networks to detect abnormal battery cells, further improving the accuracy of abnormal battery cell screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0019] Figure 1 A schematic diagram showing the steps of a method for screening abnormal cells based on a pre-charging process in one embodiment of the present invention is shown;

[0020] Figure 2 A schematic diagram showing pre-charge voltage data of a battery cell in one embodiment of the present invention is shown;

[0021] Figure 3 A schematic diagram showing data characteristics of a batch of normal and abnormal battery cells in one embodiment of the present invention is shown;

[0022] Figure 4 A schematic structural diagram of an abnormal battery cell screening device based on a pre-charging process in one embodiment of the present invention is shown;

[0023] Figure 5 A schematic structural diagram of an abnormal cell screening device based on a pre-charging process in one embodiment of the present invention is shown;

[0024] Figure 6 A schematic structural diagram of a square battery cell in one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0026] See also Figure 1 In one embodiment, a method for screening abnormal cells based on a pre-charging process specifically includes the following steps:

[0027] Step S101: extracting a plurality of sub-samples with non-overlapping time from the pre-charge voltage data of the battery cell to be tested using a plurality of time windows of equal length. The pre-charge voltage data is the relationship data between the battery cell voltage and time during the pre-charge process.

[0028] Step S102: Extract the differential eigenvalue corresponding to each sub-sample based on the differential operation between the sampled voltage values within each sub-sample.

[0029] Step S103: Use the differential eigenvalue corresponding to each sub-sample as the coordinate in the feature space, and extract the element corresponding to the coordinate from the feature space as the data feature of the current battery cell to be detected. The dimension of the feature space is equal to the number of time equal-length windows.

[0030] Step S104: Determine whether the current battery cell to be detected contains foreign objects based on the data feature.

[0031] Specifically, in the pre-charging stage of the battery cell, there are mainly 9 working steps. Among them, the 1st, 3rd, 5th, 7th, and 9th working steps are the static steps of the battery cell, and the 2nd, 4th, 6th, and 8th working steps are the charging steps of the battery cell. Among them, the 2nd step and the 4th step are mainly the voltage ranges for the formation of the SEI film (Solid electrolyte interface solid electrolyte interface film), which have a very large impact on the battery's electrochemical performance. By the 6th step, the SEI film has been formed, mainly modulating the aging SOC and deepening the formation of the SEI film at the same time. In this process, mainly in the 2nd working step, it is found through analysis that: there is a difference in the trend of the pre-charging voltage data between normal battery cells and abnormal battery cells with foreign objects. The voltage rise rate of abnormal battery cells with foreign objects is significantly lower than that of normal batteries. In other words, if a battery cell has a foreign object, the voltage rise rate of that battery cell will significantly decrease around 1500 seconds - 2000 seconds during the pre-charging stage. As Figure 2 shown, it is a schematic diagram of the pre-charging voltage data of a certain battery cell, that is, the relationship between the battery cell voltage and time.

[0032] Based on the above principle, in this embodiment, the pre-charging voltage data is used to extract features, and whether the battery cell contains foreign objects is determined according to the extracted features. Considering that the pre-charging voltage data is a strictly increasing curve, in this embodiment, a method is used to manually construct data features based on the time series. After trying features such as the area under the pre-charging curve and difference values in the experiment, it is found that the data features obtained based on the differential operation of the voltage have a large correlation with the abnormal situation of the battery.

[0033] However, relying solely on the difference values between the voltage values on the curve as features makes it somewhat difficult to distinguish in scenarios with a large number of normal battery cells and a large number of abnormal battery cells. On this basis, the embodiments of the present invention first use multiple windows of equal time length to intercept on the pre-charge voltage curve (mainly intercepting data around 1500 seconds - 2000 seconds, and a window time length range of 200 - 500 seconds has a better effect.), obtaining multiple relatively short pre-charge voltage curves, that is, multiple sub-samples. Then, differential operations are respectively performed on the voltages of each sub-sample, so that each sub-sample corresponds to a differential eigenvalue. Then, each differential eigenvalue is used as the coordinate in space and mapped in a high-dimensional space, where the dimension of the high-dimensional space is equal to the number of windows. For example: 3 windows calculate 3 differential eigenvalues, and then the 3 calculated differential eigenvalues are respectively used as the coordinates of the x, y, and z axes and mapped in a three-dimensional space, thereby obtaining an element in the three-dimensional space. In other words, the element can be regarded as a coordinate combination in the three-dimensional space. After that, the element obtained from the high-dimensional space is used as the data feature of a battery cell, further deepening the difference between normal battery cells and abnormal battery cells. Thus, based on the obtained data features, subsequent discrimination is performed on whether the battery cell contains foreign objects, significantly improving the screening accuracy of abnormal battery cells.

[0034] Specifically, in one embodiment, step S102 above specifically includes the following steps:

[0035] Step 1: Sample multiple voltage values at equal time intervals from the current sub-sample.

[0036] Step 2: Calculate the differences between adjacent voltage values to obtain multiple voltage differences.

[0037] Step 3: Calculate the average value of the multiple voltage differences, and use the obtained average value as the differential eigenvalue of the current sub-sample.

[0038] Specifically, differential operations include first-order differences, second-order differences, etc. Through experiments, it is found that there are relatively obvious differences in the first-order differential data of normal battery cells and abnormal battery cells. Therefore, in this embodiment, based on the first-order differential operation, the differential eigenvalues of each sub-sample are created. First, sample multiple voltage values at equal time intervals from the current sub-sample. For example: the current sub-sample is a 50-second pre-charge voltage curve, and a voltage value can be collected from the curve every 1 second. Then, calculate the differences between adjacent voltage values to perform differential operations. Finally, calculate the average value of all the calculated differences, that is, the differential average value. In this embodiment, the differential average value is used as the differential eigenvalue, further improving the feature discrimination degree between normal battery cells and abnormal battery cells.

[0039] The following uses a specific embodiment to comprehensively explain the above steps S101 - S104:

[0040] Obtain the pre-charge voltage data of multiple known normal battery cells and abnormal battery cells with known foreign objects, and intercept the pre-charge voltage data of each battery cell respectively. Two windows are used, namely the window from 1000 to 1500 seconds and the window from 1500 to 2000 seconds. Then, for each battery cell, calculate the differential mean corresponding to each window respectively. Then, take the differential mean corresponding to the window from 1000 to 1500 seconds as the ordinate, and take the differential mean corresponding to the window from 1500 to 2000 seconds as the abscissa, and map them into a two-dimensional space, as Figure 3 shown, so as to obtain multiple elements in the two-dimensional space. Then, mark each element according to the known normal battery cells and abnormal battery cells. It is not difficult to find that there are obvious differences between normal battery cells and abnormal battery cells.

[0041] Specifically, in one embodiment, step S104 above specifically includes the following steps:

[0042] Step Four: Input the data features into a pre-trained binary classification model, and determine whether the currently to-be-detected battery cell contains foreign objects through the output result of the pre-trained binary classification model. Among them, the pre-trained binary classification model is generated by training with the data features corresponding to a number of positive samples and a number of negative samples. The negative samples are the pre-charge voltage data of abnormal battery cells with foreign objects, and the positive samples are the pre-charge voltage data of normal battery cells.

[0043] Specifically, in the embodiment of the present invention, a neural network model is pre-created to train the data features corresponding to abnormal battery cells and normal battery cells to obtain a binary classification model. Among them, the data features of normal battery cells can be labeled as 1, and the data features of abnormal battery cells can be labeled as 0. Furthermore, data classification is performed through the pre-trained binary classification model, which further improves the screening accuracy of abnormal battery cells. In the embodiment of the present invention, during the working step 2 time period, an equal-length window method is adopted, and 6 equal-length windows are selected. Calculate the voltage differential mean within the window respectively, and use them as 6 features in turn, and directly input them into the model. Compared with selecting other numbers of windows, the detection accuracy is higher. In this embodiment, a random forest model is adopted for training, and the hyperparameters therein are determined by Bayesian hyperparameter tuning: the maximum depth of the tree max_depth = 8, the maximum number of features that a single decision tree in the random forest is allowed to use max_features = 0.28, the minimum number of samples required to split the internal node min_samples.split = 10, and the number of trees n_estimators = 130. Through verification, the effects achieved by each index of the binary classification model provided in this embodiment are as follows:

[0044] Accuracy = 0.9710588235294118;

[0045] Precision = 0.9621068376068375;

[0046] Recall = 0.8733333333333334;

[0047] F1 value = (2 * precision * recall) / (precision + recall) = 0.9128019602858314;

[0048] AUC (Area under Curve ROC, the area under the ROC curve) = 0.9326666666666666.

[0049] Specifically, in one embodiment, the steps of obtaining positive samples and negative samples include:

[0050] Step Five: Obtain abnormal battery cells known to have foreign objects, and trace the pre-charge voltage data of the abnormal battery cells known to have foreign objects as negative samples.

[0051] Step Six: Construct similar pre-charge voltage data based on the abnormal battery cells known to have foreign objects as negative samples.

[0052] Step Seven: Obtain known normal battery cells, and trace the pre-charge voltage data of the known normal battery cells as positive samples.

[0053] Step Eight: Obtain dissimilar battery cells whose internal structure similarity to the abnormal battery cells known to have foreign objects is lower than the first preset threshold, and trace the pre-charge voltage data of the dissimilar battery cells as positive samples.

[0054] Among them, in one embodiment, Step Six further includes:

[0055] Step Nine: Obtain similar battery cells whose internal structure similarity to the abnormal battery cells known to have foreign objects is higher than the second preset threshold, and trace the pre-charge voltage data of the similar battery cells as negative samples.

[0056] Step Ten: Construct similar pre-charge voltage data based on the abnormal battery cells known to have foreign objects as negative samples by the SMOTE oversampling method.

[0057] Specifically, in order to train a binary classification model so that the model can learn the difference between normal and abnormal cells, a certain number of normal and abnormal cell samples are first required. It is difficult to know whether a cell is abnormal before it is disassembled. Moreover, the manpower and material costs of disassembling the cell are also quite huge. Even if disassembly is possible, the proportion of abnormal cells in all batteries is relatively small. Therefore, the collection of abnormal samples in the prior art is relatively difficult. In this embodiment, the main sources of abnormal samples are: vehicles operating on the market have malfunctions, or faulty batteries are sampled, and then the battery packs are disassembled according to the faults, and the abnormal cells containing foreign objects are located. The pre-charge voltage data of the cell during the production process is then traced through the data system. The method collects a total of between 60 and 100 abnormal sample data, and the samples obtained in this way are not enough to meet the large number of sample requirements for model training. It is not realistic to disassemble all normally produced batteries in actual applications. Based on this, this embodiment also constructs data similar to the pre-charge voltage data of abnormal cells based on the SMOTE oversampling method on the one hand; on the other hand, for cells produced normally, based on the characteristics of the internal structure of the cells, the cells whose similarity with the abnormal cell structure is above a second preset threshold value is calculated based on the Euclidean distance, so that these cells are used as abnormal cells to trace the pre-charge voltage data, thereby achieving the purpose of expanding negative samples. Similarly, the method for obtaining positive samples, on the one hand, traces data through known normal cells, and on the other hand, finds cells with a long Euclidean distance from the abnormal cell structure in a large number of batteries produced normally as normal cells, and achieves the effect of expanding samples by tracing the pre-charge voltage data. Through the above steps, the number of positive and negative samples is increased, thereby further improving the accuracy of binary classification model training.

[0058] Specifically, in one embodiment, before training the binary classification model, the abnormal cell screening method based on the pre-charging process provided by the embodiment of the present invention further includes the following steps:

[0059] Step 11: Eliminate the charging voltage and time relationship data corresponding to the negative pressure pumping step and the static step in the pre-charging voltage data, and perform time alignment.

[0060] Step 12: Eliminate positive or negative samples that have been reworked.

[0061] Specifically, before model training, a data cleaning step is also performed to eliminate abnormal data and further ensure the accuracy of model training. Specifically, the pre-charge voltage data of the negative pressure waiting step (the first step before pre-charging begins) and the static step during pre-charging are eliminated. Because when the battery cell is pumped with negative pressure, there are large differences in the trays and the suction nozzles cannot be aligned. The time for pumping negative pressure varies. This part of the time has no effect on pre-charging and needs to be eliminated to align the time. In addition, the static step has no effect on the change in charging voltage, so it also needs to be eliminated to avoid too many redundant curves in the pre-charge voltage data. Among the constructed positive and negative samples, it is very likely that there will be reworked samples, so the samples with obvious abnormal rework are eliminated to further improve the credibility of the constructed data. In addition, abnormal data that may be caused by equipment reasons or battery cell data that can be intercepted by traditional production line methods are also eliminated.

[0062] Through the above steps, the technical solution provided by this application has been found through research that there is a relatively obvious difference in the pre-charge voltage data between normal cells and abnormal cells with foreign matter, among which the voltage rise rate of abnormal cells in the pre-charge stage is significantly lower than that of normal cells. Based on this, the pre-charge voltage data of the battery cells are sampled according to multiple windows of equal time length, and then the differential eigenvalues corresponding to each window are calculated respectively using differential operations, and then each differential eigenvalue is mapped to a multidimensional space as a spatial coordinate, so that the elements corresponding to the coordinates in the multidimensional space are extracted as the data features of the battery cells, thereby further deepening the difference between the data features of normal cells and abnormal cells. Therefore, normal cells and abnormal cells are distinguished based on the obtained data features, further improving the accuracy of abnormal cell screening.

[0063] In addition, based on the data features extracted by the above means, a binary classification model was trained in combination with data classification models such as neural networks to detect abnormal battery cells, further improving the accuracy of abnormal battery cell screening.

[0064] like Figure 4 As shown, this embodiment also provides an abnormal cell screening device based on a pre-charging process, the device comprising:

[0065] The division unit 101 is configured to extract multiple non-overlapping subsamples from the pre-charge voltage data of the battery cell to be tested using a plurality of equal-length time windows. The pre-charge voltage data is the relationship between the battery cell voltage and time during the pre-charge process. For details, see the description of step S101 in the above method embodiment and will not be repeated here.

[0066] The difference unit 102 is used to extract the differential characteristic value corresponding to each sub-sample based on the difference operation between the sampled voltage values in each sub-sample. For details, please refer to the relevant description of step S102 in the above method embodiment, which will not be repeated here.

[0067] The feature extraction unit 103 is configured to use the differential eigenvalue corresponding to each sub-sample as the coordinate of the feature space, and extract the element corresponding to the coordinate from the feature space as the data feature of the current battery cell to be detected. The dimension of the feature space is equal to the number of time-equal-length windows. For the detailed content, please refer to the relevant description of step S103 in the above method embodiment, and details will not be repeated here.

[0068] The detection unit 104 is configured to determine whether the current battery cell to be detected contains foreign matter based on the data feature. For the detailed content, please refer to the relevant description of step S104 in the above method embodiment, and details will not be repeated here.

[0069] The abnormal battery cell screening device based on the pre-charging process provided by the embodiment of the present invention is used to execute the abnormal battery cell screening method based on the pre-charging process provided by the above embodiment. The implementation manner and principle are the same. For the detailed content, please refer to the relevant description of the above method embodiment, and details will not be repeated.

[0070] Through the collaborative cooperation of the above-mentioned various components, it is found through research that there are obvious differences in the pre-charging voltage data between normal battery cells and abnormal battery cells with foreign matter. Among them, the voltage rise speed of abnormal battery cells during the pre-charging stage is significantly lower than that of normal battery cells. Based on this, the pre-charging voltage data of the battery cells are sampled according to multiple windows of equal time length, and then the differential eigenvalues corresponding to each window are calculated respectively by using differential operations. Then, each differential eigenvalue is mapped to a multi-dimensional space as a spatial coordinate, and the element corresponding to the coordinate (the differential mean combination coordinate in this embodiment) is extracted from the multi-dimensional space as the data feature of the battery cell, so as to further deepen the difference between the data features of normal battery cells and abnormal battery cells. Thus, the discrimination between normal battery cells and abnormal battery cells is performed based on the obtained data features, and the accuracy of abnormal battery cell screening is further improved.

[0071] In addition, based on the data features extracted by the above means, a binary classification model is trained in combination with a data classification model such as a neural network for the detection of abnormal battery cells, further improving the accuracy of abnormal battery cell screening.

[0072] Figure 5 shows an abnormal battery cell screening device based on the pre-charging process according to an embodiment of the present invention, which is used to screen a square battery cell similar to Figure 6 as shown. The device includes a processor 901 and a memory 902, and can be connected through a bus or other means. Figure 5 Here, taking the connection through the bus as an example.

[0073] The processor 901 may be a Central Processing Unit (CPU). The processor 901 may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. chips, or a combination of the above types of chips.

[0074] As a non-transitory computer-readable storage medium, the memory 902 can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above method embodiments. The processor 901 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory 902, that is, implements the methods in the above method embodiments.

[0075] The memory 902 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by the processor 901, etc. In addition, the memory 902 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory 902 may optionally include a memory remotely provided relative to the processor 901, and these remote memories can be connected to the processor 901 through a network. Examples of the above networks include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and combinations thereof.

[0076] One or more modules are stored in the memory 902 and, when executed by the processor 901, execute the methods in the above method embodiments.

[0077] The specific details of the above abnormal cell screening device based on the pre-charging process can be understood by referring to the corresponding relevant descriptions and effects in the above method embodiments, and will not be elaborated here.

[0078] Those skilled in the art can understand that to implement all or part of the processes in the above-described embodiment methods, it can be completed by instructing relevant hardware through a computer program. The implemented program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-described method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memories.

[0079] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An abnormal battery cell screening method based on a pre-charging process, characterized in that The method includes: Using a number of time-equal windows to extract multiple non-overlapping sub-samples from the pre-charge voltage data of the current cell to be detected. The pre-charge voltage data is the relationship data between the cell voltage and time during the pre-charge process. The time length range of the time-equal windows is 200 - 500 seconds, and the pre-charge voltage data selects the data of the cell from 1500 seconds to 2000 seconds during the pre-charge process; Based on the differential operation between the sampled voltage values within each sub-sample, extracting the differential feature value corresponding to each sub-sample; Taking the differential feature values corresponding to each sub-sample as the coordinates of the feature space, and extracting the elements corresponding to the coordinates from the feature space as the data features of the current cell to be detected. The dimension of the feature space is equal to the number of the time-equal windows; Inputting the data features into a pre-trained binary classification model, and determining whether the current cell to be detected contains foreign objects through the output result of the pre-trained binary classification model. Among them, the pre-trained binary classification model is generated by training with the data features of a number of positive samples and a number of negative samples. The negative samples are the pre-charge voltage data of abnormal cells with foreign objects, and the positive samples are the pre-charge voltage data of normal cells; The steps of obtaining positive samples and negative samples include: obtaining abnormal cells known to have foreign objects, and tracing back the pre-charge voltage data of the abnormal cells known to have foreign objects as negative samples; obtaining similar cells whose internal structure has a similarity higher than the second preset threshold with the structure of the abnormal cells known to have foreign objects, and tracing back the pre-charge voltage data of the similar cells as negative samples; constructing similar pre-charge voltage data as negative samples based on the abnormal cells known to have foreign objects through the SMOTE oversampling method; obtaining known normal cells, and tracing back the pre-charge voltage data of the known normal cells as positive samples; obtaining dissimilar cells whose internal structure has a similarity lower than the first preset threshold with the structure of the abnormal cells known to have foreign objects, and tracing back the pre-charge voltage data of the dissimilar cells as positive samples; The extracting the differential feature value corresponding to each sub-sample based on the differential operation between the sampled voltage values within each sub-sample includes: sampling multiple voltage values at equal time intervals within the current sub-sample; calculating the difference between adjacent voltage values to obtain a number of voltage differences; calculating the mean value of the number of voltage differences, and taking the obtained mean value as the differential feature value of the current sub-sample.

2. The method according to claim 1, characterized in that, Before training and generating the binary classification model with a number of positive samples and a number of negative samples, the method further includes: Removing the relationship data between the charging voltage and time corresponding to the negative pressure extraction step and the standing step in the pre-charge voltage data, and performing time alignment; Removing the positive samples and negative samples with rework.

3. An abnormal battery cell screening device based on a pre-charging process, characterized in that, The device includes: A division unit, configured to use a number of time-equal windows to extract multiple non-overlapping sub-samples from the pre-charge voltage data of the current cell to be detected. The pre-charge voltage data is the relationship data between the cell voltage and time during the pre-charge process. The time length range of the time-equal windows is 200 - 500 seconds, and the pre-charge voltage data selects the data of the cell from 1500 seconds to 2000 seconds during the pre-charge process; A differential unit, configured to extract differential eigenvalue corresponding to each sub-sample based on differential operation between sampled voltage values within each sub-sample; the extracting differential eigenvalue corresponding to each sub-sample based on differential operation between sampled voltage values within each sub-sample includes: sampling a plurality of voltage values at equal time intervals within the current sub-sample; calculating differences between adjacent voltage values to obtain a plurality of voltage differences; calculating an average value of the plurality of voltage differences, and using the obtained average value as the differential eigenvalue of the current sub-sample; A feature extraction unit, configured to use the differential eigenvalue corresponding to each sub-sample as coordinates of a feature space, and extract elements corresponding to the coordinates from the feature space as data features of the current cell to be detected, wherein a dimension of the feature space is equal to a number of the time equal-length windows; A detection unit, configured to determine whether the current cell to be detected contains a foreign object based on the data features; inputting the data features into a pre-trained binary classification model, and determining whether the current cell to be detected contains a foreign object through an output result of the pre-trained binary classification model; wherein, the pre-trained binary classification model is generated by training with data features corresponding to a plurality of positive samples and a plurality of negative samples, the negative samples are pre-charging voltage data of abnormal cells with foreign objects, and the positive samples are pre-charging voltage data of normal cells; The steps of obtaining the positive samples and the negative samples include: obtaining abnormal cells known to have foreign objects, and tracing pre-charging voltage data of the abnormal cells known to have foreign objects as negative samples; obtaining similar cells whose internal structure has a structural similarity higher than a second preset threshold with the internal structure of the abnormal cells known to have foreign objects, and tracing pre-charging voltage data of the similar cells as negative samples; constructing similar pre-charging voltage data as negative samples based on the abnormal cells known to have foreign objects through the SMOTE oversampling method; obtaining known normal cells, and tracing pre-charging voltage data of the known normal cells as positive samples; obtaining dissimilar cells whose internal structure has a structural similarity lower than a first preset threshold with the internal structure of the abnormal cells known to have foreign objects, and tracing pre-charging voltage data of the dissimilar cells as positive samples.

4. An abnormal battery cell screening device based on a pre-charging process, characterized in that, including: A memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to execute the method according to any one of claims 1-2.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the method according to any one of claims 1-2.

Images