Lithium battery self-discharge detection method, system and equipment

Abstract

The invention relates to the technical field of battery detection, in particular to a lithium battery self-discharge detection method, system and equipment. According to the detection method, firstly, feature extraction is carried out on formation data of a to-be-detected sample to obtain a feature set of a to-be-detected sample set, then physical self-discharge detection is carried out on the feature set of the to-be-detected sample set through a physical self-discharge cell detection model, and after first abnormal cell data and first cell data are output, the first abnormal cell data and the second abnormal cell data are obtained. Carrying out chemical self-discharge detection on the first battery cell data through a chemical self-discharge battery cell detection model, outputting second abnormal battery cell data and normal battery cell data, and taking the battery cell corresponding to the first abnormal battery cell data and the second abnormal battery cell data as an abnormal battery cell through the above detection. And finally outputting the battery cell corresponding to the normal battery cell data. According to the scheme, different types of abnormal battery cells can be accurately distinguished, and the detection precision of the abnormal battery cells is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery detection, and in particular to a lithium battery self-discharge detection method, system and equipment. Background Art

[0002] Each lithium battery has self-discharge, but the self-discharge rate is different. Therefore, during the production process, the self-discharge rate of lithium batteries must be tested. The self-discharge phenomenon of batteries mainly includes physical self-discharge and chemical self-discharge. Therefore, during the process of testing the self-discharge rate of lithium batteries, physical self-discharge cells and chemical self-discharge cells are mainly tested.

[0003] Due to the difference in the formation mechanism of physical self-discharge cells and chemical self-discharge cells, the characteristic laws they present are different. Therefore, physical self-discharge cells (physical NG), chemical self-discharge cells (chemical NG) and normal cells (A cells) can be regarded as three categories, and a three-classification detection model is established. However, due to the relatively small number of physical NG and chemical NG cells, the training samples are unbalanced, making the three-classification model ineffective.

[0004] At present, in order to solve the problem of few abnormal samples, there are also methods that use single classification models to combine physical NG and chemical NG into abnormal cells (NG cells) and distinguish them from normal cells (A). However, due to the difference between physical NG features and chemical NG features, chemical NG cells are easily misjudged as normal cells. Summary of the invention

[0005] The present invention provides a lithium battery self-discharge detection method, system and device, which solves the problem in the prior art that the detection accuracy of abnormal batteries is low due to the differences in the formation mechanisms of physical NG and chemical NG, resulting in different characteristic laws.

[0006] According to the first aspect, an embodiment provides a lithium battery self-discharge detection method, comprising:

[0007] Acquire formation data of the sample set to be tested; the formation data includes constant current charging step data and static step data;

[0008] Extracting features from the formation data to obtain a feature set of the sample set to be tested; the feature set includes voltage features, temperature features, and capacity features;

[0009] Detect the feature set of the sample set to be detected by using a physical self-discharge cell detection model, and output first abnormal cell data and first cell data;

[0010] The feature set of the first battery cell data is detected by a chemical self-discharge battery cell detection model, and second abnormal battery cell data and normal battery cell data are output.

[0011] In one embodiment, after extracting features from the chemical formation data to obtain the feature set of the sample set to be detected, the method further includes:

[0012] The feature set of the sample set to be detected is subjected to feature preprocessing to screen out important feature sets.

[0013] In one embodiment, the feature preprocessing of the feature set of the sample set to be detected includes:

[0014] Standardizing the feature set of the sample set to be detected to obtain a standard feature set;

[0015] Calculating the mutual information entropy value of each feature in the standard feature set;

[0016] The mutual information entropy value is compared with a preset threshold, and a feature set corresponding to the mutual information entropy value greater than the preset threshold is selected as an important feature set.

[0017] In one embodiment, the method further comprises:

[0018] According to the feature set of normal battery cells and the feature set of physical self-discharge battery cells, multiple single classifier detection models are trained and tested, and the single classifier detection model with the highest recall rate is selected as the physical self-discharge battery cell detection model;

[0019] Multiple single classifier detection models are trained and tested according to the feature sets of normal batteries and chemical self-discharge batteries, and the single classifier detection model with the highest recall rate is selected as the chemical self-discharge battery detection model.

[0020] In one embodiment, multiple single classifier detection models are trained and tested according to the feature set of normal cells and the feature set of physical self-discharge cells, and the single classifier detection model with the highest recall rate is selected as the physical self-discharge cell detection model, including:

[0021] Selecting feature sets of multiple normal cells and feature sets of multiple physical self-discharging cells from the training test feature set, and performing physical preprocessing on the selected feature sets of the multiple normal cells and feature sets of the multiple physical self-discharging cells to obtain a first feature set;

[0022] Dividing the first feature set into a first training set and a first test set according to a preset ratio;

[0023] The single classifier detection model is trained respectively using the first training set to obtain a trained detection model;

[0024] Testing the trained detection models respectively using the first test set to obtain first test results;

[0025] Calculating the recall rate of the physical self-discharging battery cells and the recall rate of the normal battery cells according to the first test result;

[0026] A single classifier detection model with the highest recall rate of the physical self-discharging battery cell and the highest recall rate of the normal battery cell is selected as the physical self-discharging battery cell detection model.

[0027] In one embodiment, the method of training and testing multiple single classifier detection models according to the feature set of normal battery cells and the feature set of chemical self-discharge battery cells, and selecting the single classifier detection model with the highest recall rate as the chemical self-discharge battery cell detection model includes:

[0028] Selecting a feature set of multiple normal batteries and a feature set of multiple chemical self-discharge batteries from the training test feature set, and performing chemical preprocessing on the feature set of the multiple normal batteries and the feature set of the multiple chemical self-discharge batteries to obtain a second feature set;

[0029] Dividing the second feature set into a second training set and a second test set according to a preset ratio;

[0030] The single classifier detection model is trained respectively using the second training set to obtain a trained detection model;

[0031] Testing the trained detection models respectively using the second test set to obtain second test results;

[0032] Calculating the recall rate of the chemical self-discharge battery cells and the recall rate of the normal battery cells according to the second test result;

[0033] A single classifier detection model with the highest recall rate of the chemical self-discharge battery cell and the highest recall rate of the normal battery cell is selected as the chemical self-discharge battery cell detection model.

[0034] According to the second aspect, an embodiment provides a lithium battery self-discharge detection system, comprising:

[0035] An acquisition module, used for acquiring formation data of the sample set to be detected; the formation data includes constant current charging step data and static step data;

[0036] A feature extraction module, used for extracting features from the formation data to obtain a feature set of the sample set to be detected; the feature set includes voltage features, temperature features and capacity features;

[0037] A physical self-discharge detection module, used to detect the feature set of the sample set to be detected by using a physical self-discharge cell detection model, and output the first abnormal cell data and the first cell data;

[0038] The chemical self-discharge detection module is used to detect the feature set of the first battery cell data through a chemical self-discharge battery cell detection model, and output second abnormal battery cell data and normal battery cell data.

[0039] According to the third aspect, an embodiment provides a lithium battery self-discharge detection device, including:

[0040] Memory for storing computer programs;

[0041] A processor is used to implement the lithium battery self-discharge detection method as described above by executing the computer program stored in the memory.

[0042] According to a fourth aspect, an embodiment provides a computer-readable storage medium, on which a computer program is stored, and the computer program can be executed by a processor to implement the method as described above.

[0043] According to the fifth aspect, an embodiment provides a computer program product, comprising a computer program and / or instructions, wherein the computer program and / or instructions implement the method as described above when executed by a processor.

[0044] According to the lithium battery self-discharge detection method, system / equipment of the above-mentioned embodiment, firstly, the formation data of the sample to be detected is subjected to feature extraction to obtain the feature set of the sample set to be detected, and then the physical self-discharge detection is performed on the feature set of the sample set to be detected through the physical self-discharge cell detection model, and after the first abnormal cell data and the first cell data are output, the first cell data is subjected to chemical self-discharge detection through the chemical self-discharge cell detection model, and the second abnormal cell data and the normal cell data are output. Through the above detection, the cell corresponding to the output first abnormal cell data and the second abnormal cell data is used as the abnormal cell, and the cell corresponding to the output normal cell data is finally used. By adopting the above-mentioned scheme of the present application, different types of abnormal cells can be accurately distinguished, and the detection accuracy of abnormal cells is effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flow chart of the lithium battery self-discharge detection method provided in this embodiment;

[0046] Figure 2A voltage data curve diagram of the formation pre-charge stage provided in this embodiment;

[0047] Figure 3 (a), (b), (c), (d), (e), and (f) are the confusion matrix diagrams of detecting physical self-discharge cells using the detection models COSVM, IF, LOF, DIF, ABOD, and Deep SVDD, respectively;

[0048] Figure 4 (a), (b), (c), (d), (e), and (f) are the confusion matrix diagrams of detecting chemical self-discharge cells using the detection models COSVM, IF, LOF, DIF, ABOD, and Deep SVDD, respectively;

[0049] Figure 5 (a), (b), and (c) are the comparison diagrams of the confusion matrices of the detection methods corresponding to the fusion single classification model, the three-classification model, and the single classification model, respectively;

[0050] Figure 6 This is a structural block diagram of the lithium battery self-discharge detection system provided in this embodiment.

[0051] Figure numerals: 10, acquisition module; 20, feature extraction module; 30, physical self-discharge detection module; 40, chemical self-discharge detection module. DETAILED DESCRIPTION

[0052] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0053] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.

[0054] The serial numbers of the components in this document, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning. The "connection" and "coupling" mentioned in this application, unless otherwise specified, include direct and indirect connections (couplings).

[0055] In view of the differences in the formation mechanisms of physical self-discharge cells (hereinafter referred to as physical NG) and chemical self-discharge cells (hereinafter referred to as chemical NG), and the different characteristic laws, resulting in different characteristic performances. For example, the three-classification model detection is not effective due to sample imbalance; the single-classification model is used to merge physical NG and chemical NG into abnormal cells, which can easily cause chemical NG to be misjudged as normal cells. Therefore, the present application proposes a lithium battery self-discharge detection method, which first detects abnormal cells of physical NG based on the physical NG self-discharge detection model, and then inputs all the remaining cells into the chemical NG self-discharge detection model to detect abnormal cells of chemical NG, thereby detecting all NG cells. This scheme provides a feasible and advanced way to accurately distinguish different types of abnormal cells.

[0056] The lithium battery self-discharge detection method of this application is as follows:

[0057] like Figure 1 As shown, a lithium battery self-discharge detection method provided in this embodiment includes the following steps:

[0058] Step 100: Acquire formation data of the sample set to be tested through an acquisition module; the formation data includes constant current charging step data and static step data.

[0059] In actual testing, the formation data of the sample set to be tested includes multiple constant current charging steps CC and static steps CV. Figure 2 The voltage data curve of a battery cell in the formation pre-charging stage includes the steps of static step 1, constant current charging step 2, static step 3, constant current charging step 4, static step 5, etc. (all including time series, voltage series, capacity series, temperature series, etc.). In view of the actual process requirements, this embodiment only takes the data of constant current charging step 2, static step 3, constant current charging step 4, and static step 5 for subsequent research and analysis.

[0060] In addition, the formation data also needs to be preprocessed. Due to factors such as the equipment itself and environmental interference, the voltage, capacity and other curve data collected by the equipment during the formation stage may contain noise, such as channel anomalies, process step delays and other problems. Therefore, the original formation curve data needs to be properly preprocessed.

[0061] Specifically, for channel abnormality, that is, during the constant current charging step, individual cells may experience a disordered increase in voltage and capacity. At this time, when preprocessing the data, the cells with channel abnormalities need to be removed to ensure the consistency and accuracy of the data.

[0062] For the step delay problem, that is, when the constant current charging step is transferred to the static step, when the current is zero (actually in the static step), the equipment records still count the battery cell in the constant current charging step. When processing data, it is necessary to divide the real constant current charging step and the static step according to the condition that the current is cut off to zero to ensure the accuracy of subsequent analysis.

[0063] Step 200: extracting features from the formation data through a feature extraction module to obtain a feature set of the sample set to be detected; the feature set includes voltage features, temperature features and capacity features.

[0064] In this step, in order to accurately characterize the self-discharge state of the lithium battery, it is necessary to extract characteristic factors based on the electrochemical reaction mechanism inside the battery. Specifically, the characteristics of the constant current charging step that need to be extracted include: capacity cumulative value, capacity average value, capacity end value, temperature end value, temperature mean, voltage variation coefficient, voltage start value, voltage mean, voltage end value, and diffusion polarization pressure difference. The characteristics of the static step that need to be extracted include: voltage end value, ohmic polarization pressure difference. Among them, for voltage characteristics, the voltage statistical characteristics include: voltage skewness, voltage kurtosis, voltage standard deviation, differential voltage maximum value, differential voltage variation coefficient, differential voltage skewness, differential voltage kurtosis, differential voltage standard deviation, and differential voltage mean.

[0065] When extracting features from the formation data, the above features are extracted from each step (in this embodiment, only the data of constant current charging step 2, static step 3, constant current charging step 4, and static step 5 are taken).

[0066] It should be noted that the sum of all values ​​of the capacity sequence of the constant current charging step is the cumulative capacity value; the average value of the capacity sequence is the average capacity value, and the last value of the capacity sequence is the end capacity value; other feature extractions are similar.

[0067] Step 300: The physical self-discharge detection module detects the feature set of the sample set to be detected by using the physical self-discharge cell detection model, and outputs the first abnormal cell data and the first cell data.

[0068] Step 400: The chemical self-discharge detection module detects a feature set of the first battery cell data through a chemical self-discharge battery cell detection model, and outputs second abnormal battery cell data and normal battery cell data.

[0069] In this embodiment, after feature extraction is performed on the formation data of the sample set to be detected to obtain a feature set, the feature set is first input into the physical self-discharge cell detection model for detection, and the first abnormal cell data and the first cell data are output. At this time, the output first abnormal cell data is a physical self-discharge cell, and the output first cell data includes normal cells and chemical self-discharge cells. Therefore, the first cell data also needs to be input into the chemical self-discharge cell detection model for detection. After the feature set of the first cell data is detected by the chemical self-discharge cell detection model, the second abnormal cell data and the normal cell data are output. At this time, the output second abnormal cell data is a chemical self-discharge cell, and the output normal cell data is a normal cell. By adopting the above-mentioned scheme of the present application, physical self-discharge cells, chemical self-discharge cells and normal cells can be accurately distinguished, which effectively improves the detection accuracy of abnormal cells.

[0070] In this embodiment, after extracting features from the chemical formation data to obtain a feature set of the sample set to be detected, the method further includes: performing feature preprocessing on the feature set of the sample set to be detected to screen out important feature sets.

[0071] In this step, feature preprocessing is performed on the feature set, mainly to measure the correlation between each feature and the self-discharge state, screen out features with strong characterization capabilities, thereby improving the detection accuracy and operating efficiency of the model, and avoiding interference of invalid or redundant features on model detection.

[0072] Specifically, feature preprocessing is performed on the feature set of the sample set to be detected, including the following steps:

[0073] Standardizing the feature set of the sample set to be detected to obtain a standard feature set;

[0074] Calculate the mutual information entropy value of each feature in the standard feature set;

[0075] The mutual information entropy value is compared with a preset threshold, and the feature set corresponding to the mutual information entropy value greater than the preset threshold is selected as the important feature set.

[0076] Standardize the feature set according to the formula: X ′ =(X i -X mean ) / X std ;

[0077] Where, X ′ represents the standardized feature value (i.e., the standard feature set), X i represents the i-th eigenvalue in the original data set, X mean represents the mean of all feature values ​​in the original data set, X stdRepresents the standard deviation of all feature values ​​in the original dataset.

[0078] After obtaining the standard feature set, the feature set (denoted as APN1) is screened for important features using mutual information entropy. Mutual information entropy aims to capture the relationship between each feature and the label, including linear and nonlinear relationships. By calculating the mutual information entropy value of each feature and sorting all the mutual information entropy values, the top-ranked features have a stronger correlation with the target variable. In this embodiment, the preset threshold is set to 0, and the calculated mutual information entropy value is compared with the preset threshold to eliminate features whose mutual information entropy values ​​are less than or equal to the preset threshold, so as to achieve the purpose of screening important features.

[0079] Specifically, in actual use, since the feature set of the sample set to be tested includes physical NG, chemical NG and normal cells, when performing feature preprocessing, the feature set of the sample set to be tested is first physically preprocessed. It should be noted that physical preprocessing only performs feature preprocessing on cells with physical NG. For example, assuming that there are four input features: voltage difference, voltage mean, capacity accumulation value, and capacity start value, after the physical preprocessing process, the obvious features (voltage difference, voltage mean) that distinguish physical NG from normal cells are preserved.

[0080] Then, the feature set is chemically preprocessed. Similarly, chemical preprocessing is only performed on chemically NG cells. For example, assuming there are four input features: voltage difference, voltage mean, capacity accumulation value, and capacity start value, after the chemical preprocessing process, the obvious features (capacity accumulation value, capacity start value) that distinguish chemically NG cells from normal cells are preserved.

[0081] For physical pretreatment and chemical pretreatment, the processing process is the same (for example, in the above steps, both are through standardization - information entropy - de-standardization), but because the input data is different, the important features of the physical NG and chemical NG outputs are different.

[0082] Furthermore, the lithium battery self-discharge detection method also includes the following steps:

[0083] According to the feature set of normal battery cells and the feature set of physical self-discharge battery cells, multiple single classifier detection models are trained and tested, and the single classifier detection model with the highest recall rate is selected as the physical self-discharge battery cell detection model;

[0084] Multiple single classifier detection models are trained and tested according to the feature sets of normal batteries and chemical self-discharge batteries, and the single classifier detection model with the highest recall rate is selected as the chemical self-discharge battery detection model.

[0085] In this embodiment, in order to ensure the accuracy of model detection, it is necessary to test and screen the physical self-discharge battery cell detection model and the chemical self-discharge battery cell detection model to select a model with the best detection effect as the physical self-discharge battery cell detection model and / or chemical self-discharge battery cell detection model of this embodiment.

[0086] Specifically, the selection of the physical self-discharge cell detection model includes the following steps:

[0087] Selecting feature sets of multiple normal cells and feature sets of multiple physical self-discharging cells from the training test feature set, and performing physical preprocessing on the selected feature sets of the multiple normal cells and feature sets of the multiple physical self-discharging cells to obtain a first feature set;

[0088] Dividing the first feature set into a first training set and a first test set according to a preset ratio;

[0089] The single classifier detection model is trained respectively by using the first training set to obtain a trained detection model;

[0090] The trained detection models are tested respectively using the first test set to obtain first test results;

[0091] Calculating the recall rate of the physical self-discharged battery cells and the recall rate of the normal battery cells according to the first test result;

[0092] The single classifier detection model with the highest recall rate of physical self-discharge cells and normal cells is selected as the physical self-discharge cell detection model.

[0093] First, feature sets of multiple normal cells and feature sets of multiple physical self-discharge cells are selected from the training test feature set, and then the selected normal cell feature sets and physical self-discharge cell feature sets are physically preprocessed to obtain the first feature set. Specifically, the physical preprocessing mainly outputs an obvious feature set that distinguishes physical NG from normal cells through standardization, information entropy screening of important features, and de-standardization (that is, the above-mentioned feature preprocessing steps), and the obvious feature set is used as the first feature set.

[0094] Take the following Table 1 as an example, which shows the number of cells in the training and testing feature set:

[0095]

[0096]

[0097] Table 1

[0098] In the first feature set of this embodiment, the total number of normal cells is 900, the total number of physical NGs is 66, and the total number of chemical NGs is 250. The first feature set is randomly divided into a first test set and a first training set according to a certain ratio. For example, 600 A cells, 36 physical NGs, and 130 chemical NGs are taken as the first training set; 300 A cells, 30 physical NGs, and 120 chemical NGs are taken as the first test set.

[0099] It should be noted that before training and testing, the first feature set is also preprocessed, mainly to measure the correlation between each feature and the self-discharge state, screen out features with strong characterization capabilities, thereby improving the detection accuracy and operating efficiency of the model, while avoiding the interference of invalid or redundant features on model detection. The specific preprocessing method is the same as the feature preprocessing method for the feature set of the sample set to be detected, so as to screen out important feature sets.

[0100] Then, the first training set is used to train different single classifier detection models to obtain trained detection models. Specifically, the single classifier detection models include: One-Class SVM, Isolation Forest (IF), Local Outlier Factor (LOF), Deep Isolation Forest (DIF), Angle-Based Outlier Detection (ABOD), Deep One-Class Classification (Deep SVDD).

[0101] After that, the first test set is input into different single classifier detection models respectively, so that each single classifier detection model detects the cells of the first test set respectively to output the first test result. The output first test result is compared with the number of A cells and physical NG in the actual test to obtain the recall rate of physical self-discharge cells and the recall rate of normal cells.

[0102] The confusion matrix of the above different single classifier detection models is as follows Figure 3 As shown, the detection effects of the above different single classifier detection models are analyzed to obtain a detection effect comparison table as shown in Table 2.

[0103] Single classifier detection model Physical NG recall rate Normal battery cell recall rate OCSVM 88.89％ 85.50％ IF 97.22％ 87.67％ LOF 97.22％ 91.17％ DIF 91.67％ 88.16％ ABOD 35 / 36 90.33％ Deep SVDD 88.89％ 84.66％

[0104] Table 2

[0105] Comparative analysis shows that the physical NG self-discharge detection model based on LOF has the highest accuracy, with a recall rate of 97.22% for physical NG cells and 91.17% for normal cells. Therefore, the physical NG self-discharge detection model based on LOF is used as the optimal physical NG detection model. The model and parameters are saved as PNG model name and PNG model params.

[0106] Similarly, the selection of the chemical self-discharge cell detection model includes the following steps:

[0107] Selecting a feature set of multiple normal batteries and a feature set of multiple chemical self-discharge batteries from the training test feature set, and performing chemical preprocessing on the feature set of the multiple normal batteries and the feature set of the multiple chemical self-discharge batteries to obtain a second feature set;

[0108] Dividing the second feature set into a second training set and a second test set according to a preset ratio;

[0109] The single classifier detection model is trained respectively by using the second training set to obtain a trained detection model;

[0110] The trained detection models are tested respectively using the second test set to obtain second test results;

[0111] Calculating the recall rate of chemical self-discharge cells and the recall rate of normal cells according to the second test result;

[0112] The single classifier detection model with the highest recall rate of chemical self-discharge cells and normal cells is selected as the chemical self-discharge cell detection model.

[0113] First, feature sets of multiple normal cells and feature sets of multiple chemical self-discharge cells are selected from the training test feature set, and then the selected normal cell feature sets and chemical self-discharge cell feature sets are chemically preprocessed to obtain the second feature set. Specifically, the chemical preprocessing mainly outputs an obvious feature set that distinguishes chemical NG from normal cells through standardization, information entropy screening of important features, and de-standardization (that is, the above-mentioned feature preprocessing steps), and the obvious feature set is used as the second feature set.

[0114] The training and testing are also performed using the above Table 1 as an example. The second training set is used to train different single classifier detection models to obtain the trained detection models. Specifically, the single classifier detection models include: One-ClassSVM, Isolation Forest (IF), Local Outlier Factor (LOF), Deep Isolation Forest (DIF), Angle-Based Outlier Detection (ABOD), Deep One-Class Classification (DeepSVDD).

[0115] After that, the second test set is input into different single classifier detection models respectively, so that each single classifier detection model detects the cells of the second test set respectively to output the second test result. The output second test result is compared with the number of A cells and chemical NG cells actually tested to obtain the recall rate of chemical self-discharge cells and the recall rate of normal cells.

[0116] The confusion matrix of the above different single classifier detection models is as follows Figure 4 As shown, the detection effects of the above different single classifier detection models are analyzed to obtain a detection effect comparison table as shown in Table 3.

[0117]

[0118]

[0119] Comparative analysis shows that the chemical NG self-discharge detection model based on ABOD has the highest accuracy, with a recall rate of 93.85% for chemical NG cells and 64.3% for normal cells. The chemical NG self-discharge detection model based on ABOD is taken as the optimal chemical NG detection model. The model and parameters are saved as CNG model name and CNG model params.

[0120] In addition, as a further improvement of the present embodiment, the detection method of the present embodiment - the fusion single classification model is compared with the recall rate of the single classification model and the three-classification model. Specifically, the training set includes A (600), physical NG (36), and chemical NG (130). The test set includes 300 A batteries, 30 physical NG batteries, and 120 chemical NG batteries. The methods for extracting physical signs and screening important features of the above-mentioned batteries are the same as above. The training set is used to train the fusion single classification model, the single classification model, and the three-classification model respectively, and then the test set is input into the fusion single classification model, the single classification model, and the three-classification model respectively, and the detection effects of the above three detection methods are evaluated respectively.

[0121] When evaluating the effect, the physical NG recall rate and the chemical NG recall rate are combined into the NG recall rate. The confusion matrix of the detection model corresponding to the above three detection methods is as follows: Figure 5 As shown in the figure, the detection effects of the fusion single classification model, the single classification model and the three-classification model are analyzed, and the detection effect comparison table shown in Table 3 is obtained:

[0122]

[0123] Table 3

[0124] Through the comparative analysis of Table 3, it can be seen that the NG recall rate of the fused single classification model used in the detection method of this application is 94%, and the A detection rate is 68.17%, which has better detection effect than other two-impulse methods.

[0125] In summary, the detection method of the present application can accurately distinguish different types of abnormal battery cells, effectively improving the detection accuracy of abnormal battery cells.

[0126] like Figure 6 As shown, a lithium battery self-discharge detection system provided in this embodiment includes an acquisition module 10, a feature extraction module 20, a physical self-discharge detection module 30 and a chemical self-discharge detection module 40. Among them, the acquisition module 10 is used to acquire the formation data of the sample set to be detected; the formation data includes constant current charging step data and static step data; the feature extraction module 20 is used to extract features from the formation data to obtain a feature set of the sample set to be detected; the feature set includes voltage features, temperature features and capacity features; the physical self-discharge detection module 30 is used to detect the feature set of the sample set to be detected through a physical self-discharge cell detection model, and output the first abnormal cell data and the first cell data; the chemical self-discharge detection module 40 is used to detect the feature set of the first cell data through a chemical self-discharge cell detection model, and output the second abnormal cell data and the normal cell data.

[0127] The lithium battery self-discharge detection system of this embodiment includes an acquisition module 10, a feature extraction module 20, a physical self-discharge detection module 30, and a chemical self-discharge detection module 40. Specifically, the functions and effects of each module are described in detail in the above lithium battery self-discharge detection method embodiment, and this embodiment will not be repeated in detail here.

[0128] The present embodiment provides a lithium battery self-discharge detection device, comprising a memory and a processor. The memory is used to store a computer program; the processor is used to implement the lithium battery self-discharge detection method as described above by executing the computer program stored in the memory.

[0129] It should be noted that the specific operation method and corresponding technical effects of the lithium battery self-discharge detection device can refer to the discussion of the aforementioned lithium battery self-discharge detection method, which will not be repeated here.

[0130] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and the computer program can be executed by a processor to implement the lithium battery self-discharge detection method as described above. Since the lithium battery self-discharge detection method has been described in detail in the above embodiment, this embodiment will not be described in detail here.

[0131] The present embodiment provides a computer program product, including a computer program and / or instructions, which implement the above-mentioned lithium battery self-discharge detection method when executed by a processor. Since the lithium battery self-discharge detection method has been described in detail in the above-mentioned embodiment, this embodiment will not be described in detail here.

[0132] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.

[0133] The above specific examples are used to illustrate the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the art, according to the concept of the present invention, some simple deductions, modifications or substitutions can be made.

Claims

1. A lithium battery self-discharge detection method, characterized in that: include: Acquire formation data of the sample set to be tested; the formation data includes constant current charging step data and static step data; Extracting features from the formation data to obtain a feature set of the sample set to be tested; the feature set includes voltage features, temperature features, and capacity features; Detect the feature set of the sample set to be detected by using a physical self-discharge cell detection model, and output first abnormal cell data and first cell data; The feature set of the first battery cell data is detected by a chemical self-discharge battery cell detection model, and second abnormal battery cell data and normal battery cell data are output.

2. The lithium battery self-discharge detection method according to claim 1, characterized in that: After extracting features from the chemical formation data to obtain a feature set of the sample set to be detected, the method further includes: The feature set of the sample set to be detected is subjected to feature preprocessing to screen out important feature sets.

3. The lithium battery self-discharge detection method according to claim 2, characterized in that: The step of performing feature preprocessing on the feature set of the sample set to be detected includes: Standardizing the feature set of the sample set to be detected to obtain a standard feature set; Calculating the mutual information entropy value of each feature in the standard feature set; The mutual information entropy value is compared with a preset threshold, and a feature set corresponding to the mutual information entropy value greater than the preset threshold is selected as an important feature set.

4. The lithium battery self-discharge detection method according to claim 1, characterized in that: The method further comprises: According to the feature set of normal battery cells and the feature set of physical self-discharge battery cells, multiple single classifier detection models are trained and tested, and the single classifier detection model with the highest recall rate is selected as the physical self-discharge battery cell detection model; Multiple single classifier detection models are trained and tested according to the feature sets of normal batteries and chemical self-discharge batteries, and the single classifier detection model with the highest recall rate is selected as the chemical self-discharge battery detection model.

5. The lithium battery self-discharge detection method according to claim 4, characterized in that: According to the feature set of normal cells and the feature set of physical self-discharge cells, multiple single classifier detection models are trained and tested, and the single classifier detection model with the highest recall rate is selected as the physical self-discharge cell detection model, including: Selecting feature sets of multiple normal cells and feature sets of multiple physical self-discharging cells from the training test feature set, recording the feature sets of the multiple normal cells and the feature sets of the multiple physical self-discharging cells as a first feature set, and performing physical preprocessing on the selected feature sets of the multiple normal cells and the feature sets of the multiple physical self-discharging cells to obtain the first feature set; Dividing the first feature set into a first training set and a first test set according to a preset ratio; The single classifier detection model is trained respectively using the first training set to obtain a trained detection model; Testing the trained detection models respectively using the first test set to obtain first test results; Calculating the recall rate of the physical self-discharging battery cells and the recall rate of the normal battery cells according to the first test result; A single classifier detection model with the highest recall rate of the physical self-discharging battery cell and the highest recall rate of the normal battery cell is selected as the physical self-discharging battery cell detection model.

6. The lithium battery self-discharge detection method according to claim 4, characterized in that: The method of training and testing multiple single classifier detection models according to the feature set of normal cells and the feature set of chemical self-discharge cells, and selecting the single classifier detection model with the highest recall rate as the chemical self-discharge cell detection model, includes: Selecting feature sets of multiple normal cells and feature sets of multiple chemical self-discharge cells from the training test feature set, recording the feature sets of the multiple normal cells and the feature sets of the multiple chemical self-discharge cells as the second feature set; and performing physical preprocessing on the selected feature sets of the multiple normal cells and the feature sets of the multiple physical self-discharge cells to obtain a first feature set; Dividing the second feature set into a second training set and a second test set according to a preset ratio; The single classifier detection model is trained respectively using the second training set to obtain a trained detection model; Testing the trained detection models respectively using the second test set to obtain second test results; Calculating the recall rate of the chemical self-discharge battery cells and the recall rate of the normal battery cells according to the second test result; A single classifier detection model with the highest recall rate of the chemical self-discharge battery cell and the highest recall rate of the normal battery cell is selected as the chemical self-discharge battery cell detection model.

7. A lithium battery self-discharge detection system, characterized in that: include: An acquisition module, used for acquiring formation data of the sample set to be detected; the formation data includes constant current charging step data and static step data; A feature extraction module, used for extracting features from the formation data to obtain a feature set of the sample set to be detected; the feature set includes voltage features, temperature features and capacity features; A physical self-discharge detection module, used to detect the feature set of the sample set to be detected by using a physical self-discharge cell detection model, and output the first abnormal cell data and the first cell data; The chemical self-discharge detection module is used to detect the feature set of the first battery cell data through a chemical self-discharge battery cell detection model, and output second abnormal battery cell data and normal battery cell data.

8. A lithium battery self-discharge detection device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the lithium battery self-discharge detection method as described in any one of claims 1 to 6 by executing the computer program stored in the memory.

9. A computer-readable storage medium, characterized in that: A computer program is stored on the medium, and the computer program can be executed by a processor to implement the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or the instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images