Lithium ion battery pack multi-fault diagnosis method and device, terminal equipment and computer readable storage medium

Abstract

The invention discloses a lithium ion battery pack multi-fault diagnosis method, a lithium ion battery pack multi-fault diagnosis device, terminal equipment and a computer readable storage medium. The lithium ion battery pack multi-fault diagnosis method comprises the steps of acquiring terminal voltage of each first battery monomer in a lithium ion battery pack to be diagnosed; extracting a first feature value of each first battery monomer under the target time domain feature; inputting the first feature value into a preset fault diagnosis model, so that the model calculates the distance between the first feature value and a second feature value of each second battery monomer in the model under the target time domain feature, and screening out a plurality of target feature values with the closest distance according to a preset closest number; voting is carried out on the target characteristic values, and the fault type of the second single battery corresponding to the target characteristic value with the maximum votes is used as the fault type of the current first single battery to be diagnosed; and determining the fault of the lithium ion battery pack according to the fault types of all the to-be-diagnosed first battery cells. According to the invention, multi-fault diagnosis in the lithium ion battery can be realized.

Description

Technical Field

[0001] The present invention relates to the field of detecting various types of faults of energy storage lithium-ion battery packs, and in particular to a lithium-ion battery pack multi-fault diagnosis method, device, terminal equipment and computer-readable storage medium. Background Art

[0002] In recent years, lithium-ion batteries have been widely used in electric vehicles and grid energy storage due to their advantages such as high power density and long cycle life. However, due to their poor tolerance to abuse, external environment and operating conditions, lithium-ion battery systems may experience various faults, leading to accelerated battery degradation and even safety accidents characterized by thermal runaway, which has aroused public concerns about the safety of lithium-ion batteries. Therefore, diagnosing various faults is crucial to improving the safety of battery systems.

[0003] Common faults in battery systems can be divided into battery abuse faults and sensor faults. Battery abuse faults mainly include overcharging, over-discharging, external short circuits and internal short circuits. Among them, internal short circuits are one of the most common causes of thermal runaway in lithium-ion batteries. In early internal short circuits (i.e., micro short circuits), the changes in battery electrical and thermal parameters caused by the short circuit are not obvious, and are highly concealed. This also makes micro short circuit faults difficult to detect. In practical applications, due to inherent defects, aging and harsh working environments, sensors will have faults such as bias and decreased accuracy of measurement values. The most direct impact of sensor failure is that the battery management system cannot obtain the accurate working status of the battery, and then issues erroneous control instructions, accelerating the degradation of the battery system. In addition, the inconsistency of the state of charge (SOC) and the state of health (SOH) between monomers will also manifest as voltage abnormalities, which leads to inconsistency and coupling with the aforementioned faults, increasing the difficulty of distinguishing the types of faults.

[0004] Existing multi-fault diagnosis methods for lithium-ion battery packs include model-based methods. Model-based methods estimate the internal parameters of the battery by establishing an electrothermal coupling model of the lithium-ion battery and combining filtering algorithms such as unscented particle filtering. If the rate of change of the internal parameters exceeds a predetermined threshold, it indicates that the battery has failed. However, model-based methods target battery cells rather than battery packs. In addition, this type of method is limited to detecting battery faults and cannot distinguish between different fault types. Therefore, it is urgent to solve the problem of multi-fault diagnosis in lithium-ion battery systems. Summary of the invention

[0005] Embodiments of the present invention provide a lithium-ion battery pack multi-fault diagnosis method, apparatus, terminal device and computer-readable storage medium, which can distinguish different types of faults of battery cells in a lithium-ion battery pack, and help to promptly discover and solve problems in the battery pack.

[0006] An embodiment of the present invention provides a lithium-ion battery pack multi-fault diagnosis method, comprising:

[0007] Acquiring the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed;

[0008] Extracting, according to the terminal voltage of each first battery cell, a first characteristic value corresponding to each first battery cell under the target time domain characteristic;

[0009] For each first characteristic value of each first battery cell in the lithium-ion battery pack to be diagnosed, the first characteristic value of each first battery cell is input into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, and selects several second characteristic values ​​with the closest distance according to the preset nearest neighbor number to obtain several target characteristic values; votes for all target characteristic values, and uses the fault type of the second battery cell corresponding to the target characteristic value with the most votes as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed;

[0010] According to the fault types of all first battery cells in the lithium-ion battery pack to be diagnosed, the fault of the lithium-ion battery pack is determined.

[0011] Furthermore, the target time domain characteristics of the lithium-ion battery pack sample are determined, including:

[0012] Obtaining the terminal voltage of each second battery cell in the lithium-ion battery pack sample and the corresponding fault type;

[0013] Extracting time domain features of the terminal voltage of each second battery cell in the lithium-ion battery pack sample to obtain feature values ​​under different categories of time domain features;

[0014] Calculate the redundancy between different categories of time domain features and the correlation between each category of time domain features and the failure results of lithium-ion battery pack samples;

[0015] The target time domain features of the lithium-ion battery pack samples are screened out from various time domain features according to redundancy and correlation.

[0016] Further, time domain features, including: approximate entropy, peak distance, energy, skewness and kurtosis;

[0017] The terminal voltage of each second battery cell in the lithium-ion battery pack sample is extracted in the time domain to obtain characteristic values ​​under different categories of time domain characteristics, including:

[0018] The characteristic value of the approximate entropy of each second battery cell is calculated by the following formula:

[0019] ApEn(m,r,N)=Φ m (r)-Φ m+1 (r);

[0020]

[0021] dis(x m [i],x m [j])=max k＝1,2,...,m (|y[i+k-1]-y[j+k-1]|);

[0022] x m [i]={y[i],y[i+1],...,y[i+m-1]};

[0023] The characteristic value of the peak distance of each second battery cell is calculated by the following formula:

[0024] p_p=|max(y)-min(y)|;

[0025] The characteristic value of the energy of each second battery cell is calculated by the following formula:

[0026]

[0027] The characteristic value of the skewness of each second battery cell is calculated by the following formula:

[0028]

[0029] The characteristic value of the kurtosis of each second battery cell is calculated by the following formula:

[0030]

[0031] Among them, ApEn(m,r,N) represents the eigenvalue of approximate entropy, m represents the window length, r represents the similarity comparison threshold, Φ m (r) represents the average similarity when the embedding dimension is m, Represents a subsequence x m The template matching probability of [i], represents the number of subsequences that meet the conditions in N-m+1, N represents the length of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, x m[i] represents the i-th terminal voltage subsequence of the lithium-ion battery pack, dis(x m [i],x m [j]) represents the distance between two subsequences, p_p represents the characteristic value of the peak distance, y represents the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, energy represents the characteristic value of energy, skewness represents the characteristic value of skewness, kurtosis represents the characteristic value of kurtosis, Represents the average value of the terminal voltage sequence value of each second battery cell in the lithium-ion battery pack, y i represents the value of the i-th data point in the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, and σ represents the standard deviation of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack.

[0032] Furthermore, the redundancy between different categories of time domain features is calculated, including:

[0033] For each second battery cell, the mutual information between various time domain features of each second battery cell is calculated one by one according to the characteristic values ​​under various time domain features by using the following formula:

[0034]

[0035] According to the mutual information between various types of time domain features of each second battery cell, the redundancy between different types of time domain features is calculated by the following formula:

[0036]

[0037] Where x represents the characteristic value of a second battery cell in one type of time domain feature, y represents the characteristic value of the same second battery cell in another type of time domain feature, and x i represents the characteristic value of the second battery cell of the i-th type in the time domain feature, y i represents the characteristic value of the second battery cell of the i-th type in another type of time domain feature, p(x i ) represents the probability density function of the characteristic value of the second battery cell in a type of time domain feature, p(y i ) represents the probability density function of the characteristic value of the second battery cell of the i-th type in the time domain feature, p(x i ,y i ) represents the joint probability density function between the characteristic value of the i-th second battery cell in one type of time domain feature and the characteristic value of the i-th second battery cell in another type of time domain feature, I(x i ;y i) represents the mutual information between the i-th eigenvalues ​​of various types of time domain features of the second battery cell, R represents the redundancy between different categories of time domain features, S represents the set composed of the eigenvalues ​​of various types of time domain features of each second battery cell, and |S| represents the number of features in the set composed of the eigenvalues ​​of various types of time domain features of each second battery cell.

[0038] Furthermore, the correlation between each type of time domain feature and the fault results of the lithium-ion battery pack sample is calculated, including:

[0039] For each type of time domain feature, calculating the mutual information between the characteristic value of each second battery cell under the type of time domain feature and the fault type of the second battery cell;

[0040]

[0041] According to the mutual information between the characteristic value of each second battery cell under the time domain characteristic and the fault type of the second battery cell, the correlation between the time domain characteristic and the fault result of the lithium-ion battery pack sample is calculated by the following formula:

[0042]

[0043] Where x represents the characteristic value of a second battery cell in a type of time domain feature, c represents the fault type of the same second battery cell, and x i represents the characteristic value of the second battery cell of the i-th type in the time domain feature, c i represents the fault type of the second battery cell of the ith cell, p(x i ) represents the probability density function of the characteristic value of the second battery cell in a type of time domain feature, p(c i ) represents the probability density function of the fault type of the i-th second battery cell, p(x i ,c i ) represents the joint probability density function of the characteristic value of the i-th second battery cell and the fault type of the i-th second battery cell in a type of time domain feature, I(x i ;c i ) represents the mutual information between the characteristic value of the second battery cell under a type of time domain feature and the fault type of the second battery cell, D represents the correlation between a type of time domain feature of the second battery cell and the fault type of the second battery cell, S represents the set composed of the characteristic values ​​of each type of time domain feature of each second battery cell, and |S| represents the number of features of the set composed of the characteristic values ​​of each type of time domain feature of each second battery cell.

[0044] Furthermore, target time domain features of lithium-ion battery pack samples are screened out from various time domain features based on redundancy and correlation, including:

[0045] According to the redundancy, the minimum redundancy between different categories of time domain features in the lithium-ion battery pack sample is solved;

[0046] The correlation and minimum redundancy between each type of time domain feature and the fault result of the lithium-ion battery pack sample are respectively calculated to obtain a set of correlation and minimum redundancy difference values ​​of each type of time domain feature;

[0047] Sort the difference set from large to small;

[0048] Starting from the maximum difference in the difference set, select the preset number of features to obtain the target difference;

[0049] According to the target difference, the time domain feature category corresponding to the target difference is used as the target time domain feature of the lithium-ion battery pack sample.

[0050] Further, the first characteristic value of each first battery cell is input into a preset fault diagnosis model so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, including:

[0051] inputting the first characteristic value into a preset fault diagnosis model;

[0052] The preset fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model by the following formula:

[0053]

[0054] Wherein, d(h,f) represents the distance between the first eigenvalue of each first battery cell under a target feature and the second eigenvalue corresponding to each second battery cell in each lithium-ion battery pack sample in the fault diagnosis model under a target time domain feature, h represents the first eigenvalue of each first battery cell under the target feature, f represents the second eigenvalue corresponding to each second battery cell in each lithium-ion battery pack sample in the fault diagnosis model under a target time domain feature, and m represents the spatial dimension.

[0055] Based on the above method embodiment, the present invention provides a corresponding device embodiment, including: a terminal voltage acquisition module, a target time domain feature extraction module, a battery cell fault diagnosis module and a battery pack fault diagnosis module;

[0056] A feature value extraction module, used to obtain the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed;

[0057] A target time domain feature extraction module, used to extract a first feature value corresponding to each first battery cell under the target time domain feature according to the terminal voltage of each first battery cell;

[0058] A fault diagnosis module is used to input the first characteristic value of each first battery cell in the lithium-ion battery pack to be diagnosed into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, selects several second characteristic values ​​with the closest distance according to a preset nearest neighbor number, and obtains several target characteristic values; votes for all target characteristic values, and uses the fault type of the second battery cell corresponding to the target characteristic value with the most votes as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed;

[0059] The battery pack fault diagnosis module is used to determine the fault of the lithium-ion battery pack according to the fault types of all first battery cells in the lithium-ion battery pack to be diagnosed.

[0060] Based on the above-mentioned method embodiment, the present invention provides a corresponding terminal device embodiment, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the lithium-ion battery pack multi-fault diagnosis method as described in the present invention are implemented.

[0061] Based on the above method embodiment, the present invention provides a corresponding computer-readable storage medium embodiment, including: a stored computer program, which controls the device where the computer-readable storage medium is located to execute the steps of the lithium-ion battery pack multi-fault diagnosis method as described in the present invention when the computer program is running.

[0062] Compared with the prior art, the beneficial effects of the embodiment of this solution are:

[0063] The present invention obtains the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed; according to the terminal voltage of each first battery cell, extracts the first characteristic value corresponding to each first battery cell under the target time domain feature. Different fault types often produce different manifestations in different time domain feature dimensions. Converting the terminal voltage of the battery cell into a time domain feature can introduce more characteristic dimensions to increase the ability to distinguish different faults; then, the first characteristic value of each first battery cell is input into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain feature in the fault diagnosis model, selects several second characteristic values ​​with the closest distance according to a preset nearest neighbor number, and obtains several target characteristic values; votes for all target characteristic values, and uses the fault type of the battery cell corresponding to the target characteristic value with the most votes as the fault type of the current battery cell in the lithium-ion battery pack to be diagnosed. Specific fault types show unique patterns in time domain characteristics. The distance between the characteristic value of the terminal voltage of the battery cell of the battery pack to be diagnosed and the characteristic value of the sample is calculated by the K-nearest neighbor algorithm (kNN). The similarity between each battery cell of the lithium-ion battery pack to be diagnosed and each lithium-ion battery pack sample in the fault diagnosis model can be quantified. The nearest neighbor sample selection and voting rules are adopted, and the existing fault classification information of each lithium-ion battery pack sample in the fault diagnosis model is used to determine the most likely fault type by majority voting, thereby realizing the distinction between different fault types in the lithium-ion battery pack. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a flow chart of a multi-fault diagnosis method for a lithium-ion battery pack provided by an embodiment of the present invention;

[0065] Figure 2 is a confusion matrix diagram of a lithium-ion battery test diagnosis result provided by an embodiment of the present invention;

[0066] Figure 3 is a confusion matrix diagram of the diagnosis result of a lithium-ion battery pack provided by an embodiment of the present invention before target time domain feature selection;

[0067] Figure 4 It is a schematic diagram of the structure of a multi-fault diagnosis device for a lithium-ion battery pack provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features.

[0070] Embodiment 1:

[0071] like Figure 1 As shown, an embodiment of the present invention provides a lithium-ion battery pack multi-fault diagnosis method, the method comprising at least the following steps:

[0072] Step S1: obtaining the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed;

[0073] In one embodiment of the present invention, a lithium-ion battery pack consisting of 8 battery cells connected in series is taken as an example, wherein the specification parameters of the lithium-ion battery cells are shown in Table 1.

[0074] In step S1, a battery management system (BMS) or other monitoring equipment is used to obtain the terminal voltage of each first battery cell in a lithium-ion battery pack composed of 8 battery cells connected in series, and the terminal voltages of the 8 first battery cells {V 1 ,V 2 ,V 3 ,...,V 8}.

[0075] Table 1 Specifications of lithium-ion batteries

[0076]

[0077] Step S2: extracting a first characteristic value corresponding to each first battery cell under a target time domain characteristic according to the terminal voltage of each first battery cell;

[0078] In one embodiment of the present invention, the approximate entropy is used as the target time domain feature for description:

[0079] According to the terminal voltage {V 1 ,V 2 ,V 3 ,...,V 8}, the terminal voltage of each first battery cell is regarded as time series data, and the characteristic value of the approximate entropy of each first battery cell is calculated to obtain the characteristic value of the approximate entropy of each first battery cell as the first characteristic value {T 1 ,T 2 ,T 3 ,...,T 8};

[0080] Step S3: for each first characteristic value of each first battery cell in the lithium-ion battery pack to be diagnosed, input the first characteristic value of each first battery cell into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, and selects several second characteristic values ​​with the closest distance according to the preset nearest neighbor number to obtain several target characteristic values; votes for all target characteristic values, and uses the fault type of the second battery cell corresponding to the target characteristic value with the most votes as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed;

[0081] In one embodiment of the present invention, the fault diagnosis model includes multiple groups of lithium-ion battery pack samples, each group of lithium-ion battery pack samples includes multiple second battery cells, and each second battery cell includes an eigenvalue of the approximate entropy {t m1 ,t m2 ,t m3 ,...,t mn} and its corresponding fault type {c m1 ,c m2 ,c m3 ,...,c mn}, where t m1 represents the approximate entropy characteristic value of the first second battery cell in the mth group of lithium-ion battery pack samples, t mn represents the approximate entropy characteristic value of the nth second battery cell in the mth lithium-ion battery pack sample, c m1 Indicates the fault type label of the first second battery cell in the mth group of lithium-ion battery pack samples, c mn It represents the fault type label of the nth second battery cell in the mth group of lithium-ion battery pack samples. The labels corresponding to various fault types are shown in Table 2.

[0082] Table 2 Labels corresponding to various failure modes

[0083]

[0084] The approximate entropy characteristic value {T 1 ,T 2 ,T 3 ,...,T8} is input into the preset fault diagnosis model, and the fault diagnosis model calculates the approximate entropy characteristic value T of the first battery cell respectively. 1 The approximate entropy eigenvalue of each second battery cell in the lithium-ion battery pack sample in the fault diagnosis model The characteristic difference of the first battery cell {e m1 ,e m2 ,e m3 ,...,e mn}, in this embodiment, the preset number of nearest neighbors is 5, then the 5 smallest data among the characteristic differences of the first battery cell are selected according to the preset number of nearest neighbors, assuming that the selected 5 smallest differences are {e 11 ,e 46 ,e 63 ,e 67 ,e 84}, according to the selected minimum difference, the approximate entropy characteristic value of the second battery cell can be obtained The corresponding target feature value is obtained by filtering and its corresponding fault type {c 11 ,c 46 ,c 63 ,c 67 ,c 84}, then, vote for the fault type corresponding to the target eigenvalue. Assuming that the fault type of the target eigenvalue obtained by screening is {1,1,1,2,1}, through voting, the result with the most votes, label 1, is normal, which is the fault type of the first battery cell. Then, the fault diagnosis model calculates the approximate entropy eigenvalue T of the second first battery cell in the lithium-ion battery pack to be diagnosed. 2 The process is performed until the approximate entropy characteristic value {T 1 ,T 2 ,T 3 ,...,T 8} is processed to obtain the fault type of each first battery cell in the lithium-ion battery pack to be diagnosed.

[0085] In a preferred embodiment, the determination of the target time domain characteristics of the lithium-ion battery pack sample includes:

[0086] Obtaining the terminal voltage of each second battery cell in the lithium-ion battery pack sample and the corresponding fault type;

[0087] Extracting time domain features of the terminal voltage of each second battery cell in the lithium-ion battery pack sample to obtain feature values ​​under different categories of time domain features;

[0088] Calculate the redundancy between different categories of time domain features and the correlation between each category of time domain features and the failure results of lithium-ion battery pack samples;

[0089] The target time domain features of the lithium-ion battery pack samples are screened out from various time domain features according to redundancy and correlation.

[0090] In a preferred embodiment, the time domain features include: approximate entropy, peak distance, energy, skewness and kurtosis;

[0091] The terminal voltage of each second battery cell in the lithium-ion battery pack sample is extracted in the time domain to obtain characteristic values ​​under different categories of time domain characteristics, including:

[0092] The characteristic value of the approximate entropy of each second battery cell is calculated by the following formula:

[0093] ApEn(m,r,N)=Φ m (r)-Φ m+1 (r);

[0094]

[0095] dis(x m [i],x m [j])=max k＝1,2,...,m (|y[i+k-1]-y[j+k-1]|);

[0096] x m [i]={y[i],y[i+1],...,y[i+m-1]};

[0097] The characteristic value of the peak distance of each second battery cell is calculated by the following formula:

[0098] p_p=|max(y)-min(y)|;

[0099] The characteristic value of the energy of each second battery cell is calculated by the following formula:

[0100]

[0101] The characteristic value of the skewness of each second battery cell is calculated by the following formula:

[0102]

[0103] The characteristic value of the kurtosis of each second battery cell is calculated by the following formula:

[0104]

[0105] Among them, ApEn(m,r,N) represents the eigenvalue of approximate entropy, m represents the window length, r represents the similarity comparison threshold, Φ m (r) represents the average similarity when the embedding dimension is m, Represents a subsequence x m The template matching probability of [i], represents the number of subsequences that meet the conditions in N-m+1, N represents the length of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, x m [i] represents the i-th terminal voltage subsequence of the lithium-ion battery pack, dis(x m [i],x m [j]) represents the distance between two subsequences, p_p represents the characteristic value of the peak distance, y represents the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, energy represents the characteristic value of energy, skewness represents the characteristic value of skewness, kurtosis represents the characteristic value of kurtosis, Represents the average value of the terminal voltage sequence value of each second battery cell in the lithium-ion battery pack, y i represents the value of the i-th data point in the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, and σ represents the standard deviation of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack.

[0106] In one embodiment of the present invention, in order to mine rich fault information, the terminal voltage of the battery cell is decomposed into multiple time domain features such as approximate entropy, peak distance, energy, skewness and kurtosis. By analyzing the changes in different time domain features, the fault type of the battery can be identified. However, if the battery operating status is diagnosed based on the above-mentioned multiple time domain features, it will undoubtedly increase the calculation complexity. In order to solve this problem, the present invention uses the maximum relevance minimum redundancy (MRMR) algorithm to analyze the redundancy between features and the correlation between features and fault modes, and then screen out the features that are most beneficial to the classification of fault modes.

[0107] It should be noted that the target time domain features in the fault diagnosis model of the present invention are the feature subsets of the best time domain features after being screened by the Maximum Relevance Minimum Redundancy (MRMR) algorithm.

[0108] Therefore, the determination of the target time domain features within the fault diagnosis model is described in detail:

[0109] First, the fault diagnosis model includes multiple groups of lithium-ion battery pack samples, each group of lithium-ion battery pack samples includes multiple second battery cells and the fault type corresponding to each second battery cell, and the terminal voltage {V 11 ,V 12 ,V 13,...,V 1n ,...V m1 ,V m2 ,V m3 ,...,V mn} and the corresponding fault type {c 11 ,c 12 ,c 13 ,...,c 1n ,...c m1 ,c m2 ,c m3 ,...,c mn};

[0110] Secondly, the terminal voltage {V 11 ,V 12 ,V 13 ,...,V 1n ,...V m1 ,V m2 ,V m3 ,...,V mn}Perform time domain feature extraction and extract feature values ​​under different categories of time domain features in, The characteristic value representing the first type of time domain characteristics of the first second battery cell in the first group of lithium-ion battery pack samples, The characteristic value of the fifth type of time domain feature of the nth second battery cell in the mth group of lithium-ion battery pack samples, and the labels corresponding to each time domain feature category are shown in Table 3;

[0111] Table 3 Labels corresponding to each time domain feature category

[0112]

[0113] Next, according to the characteristic values ​​of different categories of time domain features in the lithium-ion battery pack sample The redundancy between time domain features of different categories is calculated, that is, the redundancy between the characteristic value data of the first category of time domain features and the characteristic value data of the second category of time domain features, the redundancy between the characteristic value data of the first category of time domain features and the characteristic value data of the third category of time domain features, ..., the redundancy between the characteristic value data of the second category of time domain features and the characteristic value data of the third category of time domain features, ..., and so on; the minimum redundancy value is selected from the calculated redundancy between the time domain features of different categories; then, according to the characteristic values ​​of the time domain features of different categories in the lithium-ion battery pack sample, and the failure type of each second battery cell in the lithium-ion battery pack sample {c 11 ,c 12 ,c 13,...,c 1n ,...c m1 ,c m2 ,c m3 ,...,c mn}, calculate the correlation between each type of time domain feature and the failure result of the lithium-ion battery pack sample, and obtain the correlation between the five types of time domain features and the failure result of the lithium-ion battery pack sample, namely, the correlation between the approximate entropy and the failure result of the lithium-ion battery pack sample, the correlation between the peak distance and the failure result of the lithium-ion battery pack sample, the correlation between the energy and the failure result of the lithium-ion battery pack sample, the correlation between the skewness and the failure result of the lithium-ion battery pack sample, and the correlation between the kurtosis and the failure result of the lithium-ion battery pack sample;

[0114] Then, the correlation between the time domain features of the five categories and the fault results of the lithium-ion battery pack samples is subtracted from the minimum redundancy value to obtain the maximum correlation minimum redundancy difference set {s 1 ,s 2 ,s 3 ,s 4 ,s 5}; Sort the maximum correlation minimum redundancy difference set from small to large, and select the corresponding number from small to large according to the preset target time domain feature screening number. Assume that the maximum correlation minimum redundancy difference set is sorted from small to large as {s 1 ,s 5 ,s 4 ,s 2 ,s 3}, assuming that the number of target time domain feature screening is set to 1, the target time domain feature obtained is the approximate entropy of the first type of time domain feature. Assuming that the number of target time domain feature screening is set to 2, the target time domain feature obtained has two types of time domain features, namely the approximate entropy of the first type of time domain feature and the kurtosis of the fifth type of time domain feature.

[0115] Finally, according to the selected target time domain feature categories, the feature values ​​in the lithium-ion battery pack samples are The corresponding characteristic value is selected from the fault diagnosis model as the second characteristic value corresponding to the second battery cell in the lithium-ion battery pack under the target time domain feature. In this embodiment, the approximate entropy is used as the target time domain feature, and the second characteristic value selected in the fault diagnosis model is

[0116] In a preferred embodiment, calculating the redundancy between different categories of time domain features includes:

[0117] For each second battery cell, the mutual information between various time domain features of each second battery cell is calculated one by one according to the characteristic values ​​under various time domain features by using the following formula:

[0118]

[0119] According to the mutual information between various types of time domain features of each second battery cell, the redundancy between different types of time domain features is calculated by the following formula:

[0120]

[0121] Where x represents the characteristic value of a second battery cell in one type of time domain feature, y represents the characteristic value of the same second battery cell in another type of time domain feature, and x i represents the characteristic value of the second battery cell of the i-th type in the time domain feature, y i represents the characteristic value of the second battery cell of the i-th type in another type of time domain feature, p(x i ) represents the probability density function of the characteristic value of the second battery cell in a type of time domain feature, p(y i ) represents the probability density function of the characteristic value of the second battery cell of the i-th type in the time domain feature, p(x i ,y i ) represents the joint probability density function between the characteristic value of the i-th second battery cell in one type of time domain feature and the characteristic value of the i-th second battery cell in another type of time domain feature, I(x i ;y i ) represents the mutual information between the i-th eigenvalues ​​of various types of time domain features of the second battery cell, R represents the redundancy between different categories of time domain features, S represents the set composed of the eigenvalues ​​of various types of time domain features of each second battery cell, and |S| represents the number of features in the set composed of the eigenvalues ​​of various types of time domain features of each second battery cell.

[0122] In one embodiment of the present invention, for each second battery cell in each lithium-ion battery pack sample, the mutual information between each pair of eigenvalues ​​of each second battery cell in each lithium-ion battery pack sample can be calculated through the probability density function of each eigenvalue and the joint probability density function of each pair of eigenvalues, that is, the eigenvalues ​​are calculated. and The mutual information between and The mutual information between, ..., eigenvalues and The mutual information between, ..., eigenvalues and Then, for each second battery cell, the mutual information of the same type is added, that is, the eigenvalue and Mutual information and eigenvalues ​​between and Mutual information and eigenvalues ​​between and The mutual information between, ..., and the eigenvalues and The total mutual information between the two categories of time domain features is added together, and the redundancy between the two categories of time domain feature values ​​can be calculated by dividing the total mutual information by the square of the number of eigenvalues. The redundancy between each category of time domain feature values ​​is obtained through the mutual information. The higher the redundancy, the stronger the correlation between the feature and other features, and there may be redundant information.

[0123] In a preferred embodiment, calculating the correlation between each type of time domain feature and the fault result of the lithium-ion battery pack sample includes:

[0124] For each type of time domain feature, calculating the mutual information between the characteristic value of each second battery cell under the type of time domain feature and the fault type of the second battery cell;

[0125]

[0126] According to the mutual information between the characteristic value of each second battery cell under the time domain characteristic and the fault type of the second battery cell, the correlation between the time domain characteristic and the fault result of the lithium-ion battery pack sample is calculated by the following formula:

[0127]

[0128] Where x represents the characteristic value of a second battery cell in a type of time domain feature, c represents the fault type of the same second battery cell, and x i represents the characteristic value of the second battery cell of the i-th type in the time domain feature, c i represents the fault type of the second battery cell of the ith cell, p(x i ) represents the probability density function of the characteristic value of the second battery cell in a type of time domain feature, p(c i ) represents the probability density function of the fault type of the i-th second battery cell, p(x i ,c i ) represents the joint probability density function of the characteristic value of the i-th second battery cell and the fault type of the i-th second battery cell in a type of time domain feature, I(x i ;c i) represents the mutual information between the characteristic value of the second battery cell under a type of time domain feature and the fault type of the second battery cell, D represents the correlation between a type of time domain feature of the second battery cell and the fault type of the second battery cell, S represents the set composed of the characteristic values ​​of each type of time domain feature of each second battery cell, and |S| represents the number of features of the set composed of the characteristic values ​​of each type of time domain feature of each second battery cell.

[0129] In one embodiment of the present invention, for each second battery cell in each lithium-ion battery pack sample, the mutual information between each eigenvalue of each second battery cell in each lithium-ion battery pack sample and the fault type of the corresponding battery cell can be calculated through the probability density function of each eigenvalue, the probability density function of the fault type corresponding to each battery cell, and the joint probability density function thereof, that is, the eigenvalue is calculated. and c 11 The mutual information between and c 11 The mutual information between, ..., eigenvalues and c 11 The mutual information between, ..., eigenvalues and c mn Then, for each battery cell, the mutual information of the same type of time domain features is added, that is, the feature value and c 11 Mutual information and eigenvalues ​​between and c 12 Mutual information and eigenvalues ​​between and c 13 The mutual information between, ..., and the eigenvalues and c mn The mutual information between them is added to obtain the total mutual information between the time domain features of this category and the fault type. The total mutual information is divided by the number of eigenvalues ​​to calculate the correlation between the time domain features of this category and the fault type of the second battery cell. The correlation between each type of time domain feature and the fault type of the second battery cell is obtained through the mutual information. The higher the correlation, the stronger the correlation between the feature and the fault result of the battery cell, which may contain independent fault information.

[0130] Through the maximum correlation between various time domain features and battery cell fault types and the minimum redundancy between various time domain features, we can identify which features are most critical for fault prediction and diagnosis, and further screen out the target time domain features in the fault diagnosis model. These target time domain features have higher importance and information content in the fault diagnosis of battery cells.

[0131] In a preferred embodiment, the first characteristic value of each first battery cell is input into a preset fault diagnosis model so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, including:

[0132] inputting the first characteristic value into a preset fault diagnosis model;

[0133] The preset fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model by the following formula:

[0134]

[0135] Wherein, d(h,f) represents the distance between the first eigenvalue of each first battery cell under a target feature and the second eigenvalue corresponding to each second battery cell in each lithium-ion battery pack sample in the fault diagnosis model under a target time domain feature, h represents the first eigenvalue of each first battery cell under the target feature, f represents the second eigenvalue corresponding to each second battery cell in each lithium-ion battery pack sample in the fault diagnosis model under a target time domain feature, and m represents the spatial dimension.

[0136] In one embodiment of the present invention, the terminal voltage {V 1 ,V 2 ,V 3 ,...,V 8} is input into the fault diagnosis model, and the fault diagnosis model is based on the terminal voltage {V 1 ,V 2 ,V 3 ,...,V 8} to extract the eigenvalue of the target time domain feature, that is, the eigenvalue of the approximate entropy {T 1 ,T 2 ,T 3 ,...,T 8}; For the approximate entropy eigenvalue of each first battery cell, the fault diagnosis model calculates the approximate entropy eigenvalue of each first battery cell through the Euclidean distance calculation formula, and compares it with the second eigenvalue corresponding to each battery cell of each lithium-ion battery pack sample in the fault diagnosis model under the target time domain feature-approximate entropy. The distance between them is calculated to calculate the approximate entropy eigenvalue T of the first battery cell. 1 , respectively, with each second eigenvalue in the lithium-ion battery pack sample in the fault diagnosis model The distance between them, get the distance value Similarly, the approximate entropy eigenvalue T of the second first battery cell is calculated respectively. 2 , the approximate entropy eigenvalue T of the third first battery cell 3 ,......, the approximate entropy characteristic value T of the eighth first battery cell 8 Distance values, get several groups of distance values Then, the obtained distance value is filtered out through a preset number of nearest neighbors to obtain the second eigenvalue with the closest distance; finally, through a voting mechanism, the fault type of the second battery cell corresponding to the target eigenvalue with the most votes is used as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed.

[0137] Step S4: determining the fault of the lithium-ion battery pack according to the fault types of all first battery cells in the lithium-ion battery pack to be diagnosed.

[0138] In one embodiment of the present invention, according to the fault types {c 1 ,c 2 ,c 3 ,...,c 8}, assuming that the fault types of all the first battery cells in the lithium-ion battery pack to be diagnosed are {1,1,3,5,5,5,2,1}, it can be determined that the lithium-ion battery pack to be diagnosed has voltage sensor faults, micro-short circuit faults and SOH inconsistency faults, and the faulty battery cells can be located, which is helpful to promptly discover and solve the fault problems of the battery pack, thereby improving the reliability and performance of the battery system.

[0139] Embodiment 2:

[0140] In another embodiment of the present invention, assuming that there are multiple target time domain features, two types of time domain features, namely, approximate entropy and kurtosis, are used as target features for illustration:

[0141] According to the terminal voltage {V 1 ,V 2 ,V 3 ,...,V 8}, the terminal voltage of each first battery cell is regarded as time series data, and the characteristic value of the approximate entropy of each first battery cell and the characteristic value of the kurtosis of each first battery cell are calculated to obtain the characteristic value of the approximate entropy of each first battery cell and the eigenvalue of the kurtosis of each first battery cell As the first eigenvalue The first characteristic value of the first battery cell to be diagnosed Input into the preset fault diagnosis model, at this time, the target time domain eigenvalue in the fault diagnosis model, that is, the second eigenvalue contains The fault diagnosis model calculates the approximate entropy eigenvalue of the first battery cell respectively The approximate entropy eigenvalue of each second battery cell in the lithium-ion battery pack sample in the fault diagnosis model The distance, and the kurtosis eigenvalue of the first battery cell The kurtosis eigenvalue of each second battery cell in the lithium-ion battery pack sample in the fault diagnosis model The distance is obtained to obtain the characteristic difference set of the first battery cell In this embodiment, the preset number of nearest neighbors is 5, and the 5 smallest data among the characteristic differences of the first battery cell are selected according to the preset number of nearest neighbors. Assuming that the selected 5 smallest differences are According to the selected minimum difference, the approximate entropy characteristic value of the second battery cell can be obtained. The corresponding target feature value is obtained by filtering and its corresponding fault type {c 11 ,c 46 ,c 31 ,c 44 ,c 82}, then, vote on the fault type corresponding to the target characteristic value. Assuming that the fault type of the screened target characteristic value is {1,1,2,2,2}, through voting, the result of the fault type of label 2 with the most votes being a micro-short circuit fault is taken as the fault type of the first battery cell.

[0142] In order to further verify the effectiveness of the multi-fault diagnosis method of lithium-ion battery packs of the present invention, we conducted a set of comparative tests:

[0143] First, two battery packs consisting of eight lithium-ion battery cells connected in series were prepared for testing. The first battery pack was used as a training set, and the second battery pack was used as a test set.

[0144] For the first battery pack, before the battery cells are connected in series, the battery cell 8 is cycle aged to make its SOH degrade to 95%, which is used to simulate the aging of the battery after a period of use to increase the authenticity of the experiment. The other cells remain as new batteries that have not been aged; on this basis, the cell 7 is charged with constant current to set its initial SOC to 35%, and the initial SOC of the other cells is 20%; on this basis, the battery pack is charged and discharged 40 times, and different modes of faults are injected into the battery pack during the charging and discharging process, specifically: a 50Ω resistor is connected in parallel at both ends of the cell 2 to simulate a micro-short circuit fault, and random noise is added to the measured value of the voltage sensor corresponding to the cell 5 to simulate the voltage sensor fault. In order to facilitate subsequent analysis, digital labels are added to different fault modes according to Table 3.

[0145] For the second battery pack, the configuration, experimental test procedure, and fault injection conditions are exactly the same as those of the first battery pack.

[0146] In this experiment, the voltage sequence of the first battery group is used as the training set of the fault diagnosis model. The fault diagnosis model extracts time domain features from it and determines the target time domain features. During the target feature selection and determination process, it can be found that the time domain features with the smallest redundancy with other features and the greatest correlation with the fault type are the approximate entropy and kurtosis of the voltage sequence. Therefore, the characteristic values ​​of the approximate entropy and kurtosis are extracted from the voltage sequence of the first battery group.

[0147] Next, the approximate entropy and kurtosis of the voltage sequence of the second battery pack are used as the input of the fault diagnosis model, and the fault type is used as the output of the fault diagnosis model. The model output results are compared with the actual results of the battery pack, and the following are obtained: Figure 2 The confusion matrix diagram of the test diagnosis results shown in the figure visualizes the diagnosis results, in which we can observe the proportion of each fault mode being correctly diagnosed or misdiagnosed as other fault modes. The true attribute of the data is represented by the vertical axis, and the predicted state is represented by the horizontal axis. Each element u of the confusion matrix ij It represents the proportion of the i-th fault mode being diagnosed as the j-th fault mode. The value on the diagonal represents the correct diagnosis result. The larger the value on the diagonal, the higher the diagnostic accuracy of the diagnostic model. Figure 2 It can be seen that the element values ​​on the diagonal of the confusion matrix of the obtained result are all equal to or close to 1, which can prove that the diagnosis model constructed by the present invention can accurately detect and isolate most of the fault modes.

[0148] Figure 3 It is the confusion matrix diagram of the diagnosis results before the target time domain feature selection. Figure 2 and Figure 3 The comparison shows that selecting target time domain features can effectively eliminate redundancy between features, screen out the best features, and improve diagnostic accuracy.

[0149] like Figure 4 As shown, based on the above method embodiment, a corresponding device embodiment is provided; an embodiment of the present invention provides a lithium-ion battery pack multi-fault diagnosis device, including: a terminal voltage acquisition module, a target time domain feature extraction module, a battery cell fault diagnosis module and a battery pack fault diagnosis module;

[0150] A feature value extraction module, used to obtain the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed;

[0151] A target time domain feature extraction module, used to extract a first feature value corresponding to each first battery cell under the target time domain feature according to the terminal voltage of each first battery cell;

[0152] A fault diagnosis module is used to input the first characteristic value of each first battery cell in the lithium-ion battery pack to be diagnosed into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, selects several second characteristic values ​​with the closest distance according to a preset nearest neighbor number, and obtains several target characteristic values; votes for all target characteristic values, and uses the fault type of the second battery cell corresponding to the target characteristic value with the most votes as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed;

[0153] The battery pack fault diagnosis module is used to determine the fault of the lithium-ion battery pack according to the fault types of all first battery cells in the lithium-ion battery pack to be diagnosed.

[0154] It can be understood that the above-mentioned device item embodiments correspond to the method item embodiments of the present invention, and can implement the lithium-ion battery pack multi-fault diagnosis method provided by any of the above-mentioned method item embodiments of the present invention.

[0155] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art may understand and implement the present invention without creative work.

[0156] Based on the above-mentioned embodiment of the multi-fault diagnosis method for lithium-ion battery packs, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the multi-fault diagnosis method for lithium-ion battery packs of any embodiment of the present invention is implemented.

[0157] Exemplarily, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more module elements may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0158] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0159] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0160] Based on the above method embodiment, another embodiment is provided: another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the lithium-ion battery pack multi-fault diagnosis method described in any one of the above method embodiments of the present invention.

[0161] Wherein, the module / unit integrated with the multi-fault diagnosis device / terminal device of the lithium-ion battery pack, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0162] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-fault diagnosis method for a lithium-ion battery pack, characterized in that: include: Acquiring the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed; Extracting, according to the terminal voltage of each first battery cell, a first characteristic value corresponding to each first battery cell under the target time domain characteristic; For each first characteristic value of each first battery cell in the lithium-ion battery pack to be diagnosed, the first characteristic value of each first battery cell is input into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, and selects a number of second characteristic values ​​with the closest distance according to a preset nearest neighbor number to obtain a number of target characteristic values; Voting for all target characteristic values, and taking the fault type of the second battery cell corresponding to the target characteristic value with the most votes as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed; According to the fault types of all first battery cells in the lithium-ion battery pack to be diagnosed, the fault of the lithium-ion battery pack is determined.

2. The lithium-ion battery pack multi-fault diagnosis method according to claim 1, characterized in that: Determination of target time domain characteristics of lithium-ion battery pack samples, including: Obtaining the terminal voltage of each second battery cell in the lithium-ion battery pack sample and the corresponding fault type; Extracting time domain features of the terminal voltage of each second battery cell in the lithium-ion battery pack sample to obtain feature values ​​under different categories of time domain features; Calculate the redundancy between different categories of time domain features and the correlation between each category of time domain features and the failure results of lithium-ion battery pack samples; The target time domain features of the lithium-ion battery pack sample are screened out from various types of time domain features according to the redundancy and the correlation.

3. The lithium-ion battery pack multi-fault diagnosis method according to claim 2, characterized in that: The time domain features include: approximate entropy, peak distance, energy, skewness and kurtosis; The terminal voltage of each second battery cell in the lithium-ion battery pack sample is extracted in the time domain to obtain characteristic values ​​under different categories of time domain characteristics, including: The characteristic value of the approximate entropy of each second battery cell is calculated by the following formula: ApEn(m, r， N)=Φ m (r)-Φ m+1 (r); dis(x m [i],x m [j])=max k＝1，2，...，m (|y[i+k-1]-y[j+k-1]|); X m [i]={y[i],y[i+1],...,y[i+m-1]}; The characteristic value of the peak distance of each second battery cell is calculated by the following formula: p_p=|max(y)-min(y)|; The characteristic value of the energy of each second battery cell is calculated by the following formula: The characteristic value of the skewness of each second battery cell is calculated by the following formula: The characteristic value of the kurtosis of each second battery cell is calculated by the following formula: Among them, ApEn(m, r, N) represents the eigenvalue of approximate entropy, m represents the window length, r represents the similarity comparison threshold, Φ m (r) represents the average similarity when the embedding dimension is m, Represents a subsequence x m The template matching probability of [i], represents the number of subsequences that meet the conditions in N-m+1, N represents the length of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, x m [i] represents the i-th terminal voltage subsequence of the lithium-ion battery pack, dis(x m [i],x m [j]) represents the distance between two subsequences, p_p represents the eigenvalue of the peak distance, y represents the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, energy represents the eigenvalue of energy, skewness represents the eigenvalue of skewness, and kurtosis represents the eigenvalue of kurtosis. represents the average value of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, yi represents the value of the ith data point in the terminal voltage sequence of each second battery cell in the lithium-ion battery pack, and σ represents the standard deviation of the terminal voltage sequence of each second battery cell in the lithium-ion battery pack.

4. The lithium-ion battery pack multi-fault diagnosis method according to claim 3, characterized in that: Calculate the redundancy between different categories of time domain features, including: For each second battery cell, the mutual information between various time domain features of each second battery cell is calculated one by one according to the characteristic values ​​under various time domain features by using the following formula: According to the mutual information between various types of time domain features of each second battery cell, the redundancy between different types of time domain features is calculated by the following formula: Where x represents the characteristic value of a second battery cell in one type of time domain feature, y represents the characteristic value of the same second battery cell in another type of time domain feature, and x i represents the characteristic value of the second battery cell of the i-th type in the time domain feature, y i represents the characteristic value of the second battery cell of the i-th type in another type of time domain feature, p(x i ) represents the probability density function of the characteristic value of the second battery cell in a type of time domain feature, p(y i ) represents the probability density function of the characteristic value of the second battery cell of the i-th type in the time domain feature, p(x i ,y i ) represents the joint probability density function between the characteristic value of the i-th second battery cell in one type of time domain feature and the characteristic value of the i-th second battery cell in another type of time domain feature, I(x i ;y i ) represents the mutual information between the i-th eigenvalues ​​of various types of time domain features of the second battery cell, R represents the redundancy between different categories of time domain features, S represents the set composed of the eigenvalues ​​of various types of time domain features of each second battery cell, and |S| represents the number of features in the set composed of the eigenvalues ​​of various types of time domain features of each second battery cell.

5. The lithium-ion battery pack multi-fault diagnosis method according to claim 4, characterized in that: Calculate the correlation between each type of time domain feature and the failure results of the lithium-ion battery pack samples, including: For each type of time domain feature, calculating the mutual information between the characteristic value of each second battery cell under the type of time domain feature and the fault type of the second battery cell; According to the mutual information between the characteristic value of each second battery cell under the time domain characteristic and the fault type of the second battery cell, the correlation between the time domain characteristic and the fault result of the lithium-ion battery pack sample is calculated by the following formula: Where x represents the characteristic value of a second battery cell in a type of time domain feature, c represents the fault type of the same second battery cell, and x i represents the characteristic value of the second battery cell of the i-th type in the time domain feature, c i represents the fault type of the second battery cell of the ith cell, p(x i ) represents the probability density function of the characteristic value of the second battery cell in a type of time domain feature, p(c i ) represents the probability density function of the fault type of the i-th second battery cell, p(x i , c i ) represents the joint probability density function of the characteristic value of the i-th second battery cell and the fault type of the i-th second battery cell in a type of time domain feature, I(x i ;c i ) represents the mutual information between the characteristic value of the second battery cell under a type of time domain feature and the fault type of the second battery cell, D represents the correlation between a type of time domain feature of the second battery cell and the fault type of the second battery cell, S represents the set composed of the characteristic values ​​of each type of time domain feature of each second battery cell, and |S| represents the number of features of the set composed of the characteristic values ​​of each type of time domain feature of each second battery cell.

6. The lithium-ion battery pack multi-fault diagnosis method according to claim 5, characterized in that: The target time domain features of the lithium-ion battery pack sample are screened out from various time domain features according to the redundancy and the correlation, including: According to the redundancy, solving the minimum redundancy between different categories of time domain features in the lithium-ion battery pack sample; Differences are respectively calculated between the correlation between each type of time domain feature and the fault result of the lithium-ion battery pack sample and the minimum redundancy to obtain a set of correlation and minimum redundancy difference values ​​of each type of time domain feature; Sort the difference value set from large to small; Starting from the maximum difference in the difference set, a preset number of features is selected to obtain a target difference; According to the target difference, the time domain feature category corresponding to the target difference is used as the target time domain feature of the lithium-ion battery pack sample.

7. The lithium-ion battery pack multi-fault diagnosis method according to claim 6, characterized in that: Inputting the first characteristic value of each first battery cell into a preset fault diagnosis model so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, including: inputting the first characteristic value into a preset fault diagnosis model; The preset fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model by the following formula: Wherein, d(h,f) represents the distance between the first eigenvalue of each first battery cell under a target feature and the second eigenvalue corresponding to each second battery cell in each lithium-ion battery pack sample in the fault diagnosis model under a target time domain feature, h represents the first eigenvalue of each first battery cell under the target feature, f represents the second eigenvalue corresponding to each second battery cell in each lithium-ion battery pack sample in the fault diagnosis model under a target time domain feature, and m represents the spatial dimension.

8. A multi-fault diagnosis device for a lithium-ion battery pack, characterized in that: include: Terminal voltage acquisition module, target time domain feature extraction module, battery cell fault diagnosis module and battery pack fault diagnosis module; The characteristic value extraction module is used to obtain the terminal voltage of each first battery cell in the lithium-ion battery pack to be diagnosed; The target time domain feature extraction module is used to extract the first feature value corresponding to each first battery cell under the target time domain feature according to the terminal voltage of each first battery cell; The fault diagnosis module is used for inputting the first characteristic value of each first battery cell in the lithium-ion battery pack to be diagnosed into a preset fault diagnosis model, so that the fault diagnosis model calculates the distance between the first characteristic value and the second characteristic value corresponding to each second battery cell in the lithium-ion battery pack sample under the target time domain characteristic in the fault diagnosis model, and selects a number of second characteristic values ​​with the closest distance according to a preset nearest neighbor number to obtain a number of target characteristic values; Voting for all target characteristic values, and taking the fault type of the second battery cell corresponding to the target characteristic value with the most votes as the fault type of the current first battery cell in the lithium-ion battery pack to be diagnosed; The battery pack fault diagnosis module is used to determine the fault of the lithium-ion battery pack according to the fault types of all first battery cells in the lithium-ion battery pack to be diagnosed.

9. A terminal device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the multi-fault diagnosis method for a lithium-ion battery pack as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: include: A stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the lithium-ion battery pack multi-fault diagnosis method as described in any one of claims 1-7.

Images