Battery pack voltage fault diagnosis method and device, electronic equipment, readable storage medium and computer program product

Abstract

This application discloses a battery pack voltage fault diagnosis method. The method includes: acquiring the charging voltage sequence of each cell in the battery pack; calculating the normalized anomaly score feature and normalized Z-score feature of each cell based on the charging voltage sequence; constructing a normalized two-dimensional feature for each cell based on the normalized anomaly score feature and normalized Z-score feature; performing fault diagnosis using a pre-constructed adaptive Bayesian-Gaussian mixture model and the normalized two-dimensional feature to identify faulty cells; and outputting a first warning message for the faulty cell. This method can solve the problems of low diagnostic accuracy and the tendency for false alarms or missed detections when facing unknown or complex fault scenarios.

Description

Technical Field

[0001] This application relates to the field of battery technology, specifically to a method, apparatus, electronic device, readable storage medium, and computer program product for diagnosing battery pack voltage faults. Background Technology

[0002] Currently, the electric vehicle industry is developing rapidly, and the battery pack, as the core power source, directly affects vehicle performance and driving safety in terms of safety and reliability. Battery pack voltage failure is one of the key causes of performance degradation and even safety accidents, making data-driven voltage fault diagnosis technology a research focus. Existing technologies mostly rely on time-domain analysis methods to identify sudden anomalies by extracting instantaneous features of voltage signals (such as peaks and abrupt changes), but they are difficult to capture slowly evolving potential faults (such as long-term voltage drift and progressive aging), leading to a significant risk of missed fault detection. It is evident that existing methods generally suffer from low diagnostic accuracy, especially when facing unknown or complex fault scenarios, easily resulting in false alarms or missed detections. Summary of the Invention

[0003] In view of the above problems, this application provides a battery pack voltage fault diagnosis method, device, electronic device, readable storage medium and computer program product, which can solve the problems of low diagnostic accuracy and the tendency to false alarm or missed detection when facing unknown or complex fault scenarios.

[0004] In a first aspect, this application provides a method for diagnosing battery pack voltage faults, including:

[0005] Obtain the charging voltage sequence of each cell in the battery pack;

[0006] Based on the charging voltage sequence, calculate the normalized anomaly fraction feature and normalized Z-score feature for each cell;

[0007] Based on the normalized anomaly score feature and the normalized Z score feature, a normalized two-dimensional feature for each battery cell is constructed.

[0008] Fault diagnosis is performed using a pre-constructed Bayesian-Gaussian mixture model based on an adaptive mechanism and the normalized two-dimensional features to identify faulty battery cells.

[0009] Output the first warning information for the faulty battery cell.

[0010] In the above technical solution, the method can improve the accuracy and coverage of battery pack voltage fault diagnosis, effectively identify and distinguish multiple types of voltage faults (including progressive faults and sudden faults), and enhance the discrimination and diagnostic reliability of fault type identification, thereby timely outputting early warning information for faulty cells.

[0011] In some implementations, calculating the normalized anomaly fraction feature and normalized Z-score feature for each cell based on the charging voltage sequence includes:

[0012] The charging voltage sequence is preprocessed to obtain a preprocessed voltage sequence;

[0013] The preprocessed voltage sequence is divided by a preset sliding window to obtain multiple division sequences corresponding to each cell;

[0014] The normalized anomaly score feature corresponding to each cell is calculated based on the multiple partitioning sequences; and the normalized Z score feature corresponding to each cell is calculated based on the multiple partitioning sequences.

[0015] In the above technical solution, the method can effectively extract the normalized anomaly fraction feature and normalized Z-score feature of each cell from the charging voltage sequence through preprocessing and sliding window partitioning technology, providing accurate feature data for subsequent fault diagnosis.

[0016] In some implementations, calculating the normalized anomaly score feature corresponding to each cell based on the plurality of partition sequences includes:

[0017] Kernel density estimation is performed on the multiple partition sequences to obtain the voltage value of each cell at each time point;

[0018] Based on the voltage value of each cell at each time point, calculate the anomaly score of each cell at each time point to obtain a time-based anomaly score sequence for each cell;

[0019] The maximum value on the anomaly score sequence is determined as the normalized anomaly score feature for each cell.

[0020] In the above technical solution, the method can quantify the degree of voltage anomaly of each cell at each time point through kernel density estimation technology, and extract the maximum anomaly score that can reflect the instantaneous anomaly characteristics of the cell voltage as the normalized anomaly score feature, thereby effectively capturing the instantaneous anomaly changes in the voltage sequence and providing key information for fault diagnosis.

[0021] In some implementations, calculating the normalized Z-score feature corresponding to each cell based on the plurality of partition sequences includes:

[0022] Based on the plurality of division sequences, the cell voltage deviation corresponding to each division sequence is calculated;

[0023] Based on the cell voltage deviation, calculate the average value of the absolute value of the voltage deviation corresponding to each of the division sequences and the standard deviation of the absolute value of the voltage deviation corresponding to each of the division sequences;

[0024] Calculate the standardized Z-score for each partition sequence based on the mean and the standard deviation;

[0025] The normalized Z-score characteristics of each cell are determined based on the standardized Z-score.

[0026] In the above technical solution, the method can effectively quantify the drift trend and fluctuation of cell voltage, providing important statistical characteristics for battery pack voltage fault diagnosis.

[0027] In some implementations, the fault diagnosis using a pre-built adaptive Bayesian-Gaussian mixture model and the normalized two-dimensional features to identify faulty battery cells includes:

[0028] Obtain a pre-constructed Bayesian Gaussian mixture model based on an adaptive mechanism; wherein, the posterior distribution of the Bayesian Gaussian mixture model is determined using a variational distribution algorithm and a KL divergence minimization algorithm;

[0029] The model parameters of the Bayesian-Gaussian mixture model are adaptively adjusted to obtain a fault identification model;

[0030] Clustering is performed based on the fault identification model and the normalized two-dimensional features to obtain the clustering result for each battery cell; wherein, the clustering result includes clustering labels and clustering parameters;

[0031] Based on the clustering results, the categories of fault-free battery cells and multiple categories of battery cells to be identified are determined.

[0032] Based on the clustering results, calculate the inter-class distance between each of the cell classes to be identified and the class of fault-free cells;

[0033] The cell categories to be identified that have an inter-class distance greater than a preset distance threshold are classified as faulty cells.

[0034] The faulty cells are identified based on the aforementioned faulty cell type.

[0035] In the above technical solution, the method can accurately and effectively distinguish between fault-free cells and faulty cells, thereby achieving accurate diagnosis of battery pack voltage faults.

[0036] In some embodiments, the method further includes:

[0037] Obtain the normalized anomaly score feature and normalized Z score feature of the faulty battery cell;

[0038] The state space of each faulty cell is constructed based on the normalized anomaly score feature and the normalized Z score feature;

[0039] The faulty cells are classified according to the state space to obtain sudden faulty cells and gradual faulty cells;

[0040] Output a second early warning message for the sudden failure battery cell and the gradual failure battery cell.

[0041] In the above technical solution, this method can improve the precision of battery pack voltage fault diagnosis and the accuracy of early warning.

[0042] Secondly, this application provides a battery pack voltage fault diagnosis device, comprising:

[0043] The acquisition unit is used to acquire the charging voltage sequence of each cell in the battery pack;

[0044] The calculation unit is used to calculate the normalized anomaly fraction feature and the normalized Z-score feature of each cell based on the charging voltage sequence.

[0045] A construction unit is used to construct normalized two-dimensional features for each cell based on the normalized anomaly score features and the normalized Z score features.

[0046] The diagnostic unit is used to perform fault diagnosis using a pre-built Bayesian-Gaussian mixture model based on an adaptive mechanism and the normalized two-dimensional features to identify faulty battery cells.

[0047] The output unit is used to output the first warning information for the faulty battery cell.

[0048] In the above technical solution, the device can improve the accuracy and coverage of battery pack voltage fault diagnosis, effectively identify and distinguish multiple types of voltage faults (including progressive faults and sudden faults), and enhance the discrimination and diagnostic reliability of fault type identification, thereby timely outputting early warning information for faulty cells.

[0049] Thirdly, this application provides an electronic device including a memory and a processor, the memory storing a computer program, and the processor running the computer program to cause the electronic device to perform the battery pack voltage fault diagnosis method described in any one of the first aspects.

[0050] Fourthly, this application provides a readable storage medium storing a computer program, which, when executed by a processor, performs the battery pack voltage fault diagnosis method described in any one of the first aspects.

[0051] Fifthly, this application provides a computer program product, which includes a computer program that, when executed by a processor, performs the battery pack voltage fault diagnosis method described in any one of the first aspects.

[0052] The beneficial effects of this application are: it can extract fused features that take into account both time-domain features and statistical features to capture sudden anomalies and quantify voltage drift trends, thereby significantly improving accuracy and coverage compared to methods that directly use raw voltage data for fault analysis and diagnosis.

[0053] It can also improve its adaptability to the diversity and dynamic evolution of fault modes based on the adaptive Bayesian Gaussian Mixture Model (BGMM). Compared with models that pre-determine the number of clusters, it can adaptively determine the number of fault categories, thereby significantly reducing the probability of false alarms or missed detections when facing unknown or complex faults.

[0054] It can further subdivide the fault type into sudden faults and gradual faults, which enhances the ability to identify battery fault types and the reliability of diagnosis. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart illustrating a battery pack voltage fault diagnosis method in some embodiments of this application;

[0057] Figure 2 The voltage fault clustering results based on feature fusion and hybrid models in some embodiments of this application;

[0058] Figure 3 The voltage fault clustering results based on feature fusion and hybrid models in some embodiments of this application;

[0059] Figure 4 The voltage fault clustering results based on feature fusion and hybrid models in some embodiments of this application;

[0060] Figure 5 The voltage fault clustering results based on feature fusion and hybrid models in some embodiments of this application;

[0061] Figure 6This is a flowchart illustrating the voltage fault diagnosis process based on feature fusion and hybrid models in some embodiments of this application.

[0062] Figure 7 This is a schematic diagram of the battery pack voltage fault diagnosis device in some embodiments of this application;

[0063] Figure 8 This is a schematic diagram of the structure of an electronic device in some embodiments of this application. Detailed Implementation

[0064] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0066] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more (including two), similarly, "multiple sets" refers to two or more sets (including two sets), and "multiple pieces" refers to two or more pieces (including two pieces) unless otherwise explicitly defined.

[0067] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0068] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0069] With the increasing demands for battery pack safety and reliability in electric vehicles, data-driven battery pack voltage fault diagnosis methods have become a research hotspot. Classical time-domain analysis methods can sensitively capture sudden anomalies by extracting instantaneous voltage features, but they are insufficient for responding to slowly evolving faults. Statistical feature methods can quantify voltage drift trends, but they are difficult to accurately locate instantaneous faults. At the same time, common data-driven models (such as traditional GMMs) require pre-setting the number of clusters, lack the ability to adapt to the diversity of fault modes and dynamic evolution, and are prone to false alarms or missed detections when facing unknown or complex faults.

[0070] To address the aforementioned technical problems, this application provides a battery pack voltage fault diagnosis method. This method integrates time-domain features and statistical features, comprehensively considering both the instantaneous changes and long-term trends of the voltage signal. This allows for full utilization of the synergistic effect of these two types of features, thereby improving the comprehensiveness and effectiveness of fault diagnosis. Furthermore, by introducing an adaptive Bayesian Gaussian Mixture Model (BGMM), this method not only combines the performance advantages of traditional diagnostic methods and enhances lightweight application, but also adaptively determines the number of fault categories, achieving more accurate hybrid model diagnosis. In addition, this method can further distinguish between sudden and progressive faults by analyzing the fused features of fault classes, thereby refining the fault type and significantly improving the accuracy and diagnostic coverage of battery pack fault detection.

[0071] like Figure 1 As shown, some embodiments of this application provide a battery pack voltage fault diagnosis method, which includes:

[0072] S101. Obtain the charging voltage sequence of each cell in the battery pack;

[0073] S102. Based on the charging voltage sequence, calculate the normalized anomaly fraction feature and normalized Z-score feature of each cell.

[0074] S103. Based on the normalized anomaly score characteristics and the normalized Z-score characteristics, construct the normalized two-dimensional features for each cell.

[0075] S104. Fault diagnosis is performed using a pre-constructed Bayesian-Gaussian mixture model based on an adaptive mechanism and normalized two-dimensional features to identify faulty battery cells.

[0076] S105, Output the first warning information for the faulty battery cell.

[0077] In some embodiments, the method is applied to scenarios such as batteries, fault diagnosis and early warning, aging analysis and prediction, simulation analysis, and BMS software development.

[0078] In these embodiments, the method can diagnose battery pack voltage faults based on feature fusion and an adaptive mechanism-based Bayesian-Gaussian mixture model, thereby improving the accuracy and coverage of battery system diagnosis for multiple types of voltage faults in practical vehicles.

[0079] In some embodiments, the normalized anomaly fraction feature and normalized Z-score feature of each cell are calculated based on the charging voltage sequence, including:

[0080] The charging voltage sequence is preprocessed to obtain a preprocessed voltage sequence.

[0081] The preprocessed voltage sequence is divided by a preset sliding window to obtain multiple division sequences corresponding to each cell;

[0082] The normalized anomaly score feature for each cell is calculated based on multiple partition sequences; and the normalized Z-score feature for each cell is calculated based on multiple partition sequences.

[0083] In some embodiments, the method uses a sliding window to divide charging data segments.

[0084] In these embodiments, the method can effectively extract the normalized anomaly score features and normalized Z score features of each cell from the charging voltage sequence through preprocessing and sliding window partitioning techniques, providing accurate feature data for subsequent fault diagnosis.

[0085] In some embodiments, the normalized anomaly score feature corresponding to each cell is calculated based on multiple partition sequences, including:

[0086] Kernel density estimation is performed on multiple partition sequences to obtain the voltage value of each cell at each time point;

[0087] Based on the voltage value of each cell at each time point, calculate the anomaly score of each cell at each time point to obtain the time-based anomaly score sequence of each cell;

[0088] The maximum value on the anomaly score sequence is determined as the normalized anomaly score feature for each cell.

[0089] In some embodiments, the method calculates the kernel density estimate of the voltage sequence of each cell, obtains the anomaly score of each cell at each time point, and extracts the maximum outlier in the time series as a time-domain feature.

[0090] For example, firstly, the kernel density estimate (KDE) is defined as shown in Equation (1).

[0091]

[0092] Where h is the bandwidth parameter;

[0093] n is the number of samples in the window;

[0094] K(u) is the Gaussian kernel function, which can be expressed as formula (2).

[0095]

[0096] Secondly, the voltage value x of battery i at time t. i，t Processing it into the form of formula (3) can highlight the relative differences between battery cells.

[0097]

[0098] Where n is the total number of batteries.

[0099] Then, for the data within the time window, anomaly scores are calculated based on formula (4).

[0100]

[0101] Among them, X t It is the set of voltage values ​​of all batteries at time t.

[0102] Finally, the maximum value in the time series is determined as the outlier score feature, as shown in formula (5).

[0103]

[0104] In these embodiments, the method can quantify the degree of voltage anomaly of each cell at each time point through kernel density estimation technology, and extract the maximum anomaly score that can reflect the instantaneous anomaly characteristics of the cell voltage as a normalized anomaly score feature, thereby effectively capturing instantaneous anomaly changes in the voltage sequence and providing key information for fault diagnosis.

[0105] In some embodiments, the normalized Z-score feature corresponding to each cell is calculated based on multiple partition sequences, including:

[0106] Calculate the cell voltage deviation corresponding to each of the multiple partitioning sequences;

[0107] Based on the cell voltage deviation, calculate the average value of the absolute value of the voltage deviation corresponding to each division sequence and the standard deviation of the absolute value of the voltage deviation corresponding to each division sequence;

[0108] Calculate the standardized Z-score for each partition sequence based on the mean and standard deviation;

[0109] The normalized Z-score characteristics of each cell are determined based on the standardized Z-score.

[0110] In some embodiments, the method employs a sliding window technique to obtain the average value and standard deviation of the voltage deviation, further calculates the Z-score, and extracts its maximum value as a statistical feature.

[0111] For example, for cell i, the Z-score calculation process within the time window W is as follows.

[0112] First, calculate the cell voltage deviation, as shown in formula (6).

[0113]

[0114] Among them, V i (t) is the voltage value of battery i at time t, and N is the total number of batteries.

[0115] Next, the standardized scores are calculated within window W, as shown in formula (7).

[0116]

[0117] in, It is the average of the absolute values ​​of the voltage deviation within window W;

[0118] It is the standard deviation of the voltage deviation within window W;

[0119] The calculation methods for these two parameters are shown in formulas (8) and (9), respectively.

[0120]

[0121] in, It is the average value of the deviation within the window.

[0122] Finally, the maximum Z-score of all windows is determined as the Z-score feature, as shown in formula (10).

[0123]

[0124] In these embodiments, the method can effectively quantify the drift trend and fluctuation of cell voltage, providing important statistical characteristics for battery pack voltage fault diagnosis.

[0125] In some embodiments, a normalized two-dimensional feature for each cell is constructed based on the normalized anomaly score feature and the normalized Z-score feature, including:

[0126] The normalized Z-score features and the normalized outlier score features are fused to obtain fused two-dimensional features;

[0127] The fused two-dimensional features are normalized to obtain the normalized two-dimensional features of each cell.

[0128] In some embodiments, the method defines the fused features of the normalized Z-score features and the outlier scores as a two-dimensional feature vector.

[0129] In some embodiments, this feature fusion method can extract instantaneous and statistical features of voltage while taking into account time-domain analysis methods, thereby quantifying voltage drift trends.

[0130] For example, for each cell i, its fused two-dimensional feature vector is defined by formula (11).

[0131] F i =[f z,i ,f a,i ] T (11)

[0132] Among them, f z,i It is the normalized Z-score feature, f a,i These are the abnormal score features after normalization.

[0133] Then, in the two-dimensional features of the fused features, each cell is represented as a point, as shown in Equation (12).

[0134]

[0135] The feature normalization process uses Min-Max standardization, as shown in formula (13).

[0136]

[0137] In these embodiments, the method can effectively integrate normalized Z-score features and normalized anomaly score features through feature fusion processing to form fused two-dimensional features, and further normalize the fused features to obtain more representative and consistent normalized two-dimensional features, providing more accurate and reliable feature input for subsequent fault diagnosis.

[0138] In some embodiments, fault diagnosis is performed using a pre-built adaptive Bayesian-Gaussian mixture model and normalized two-dimensional features to identify faulty battery cells, including:

[0139] Obtain a pre-constructed Bayesian Gaussian mixture model based on an adaptive mechanism; wherein, the posterior distribution of the Bayesian Gaussian mixture model is determined using a variational distribution algorithm and a KL divergence minimization algorithm;

[0140] The model parameters of the Bayesian-Gaussian mixture model are adaptively adjusted to obtain the fault identification model;

[0141] Clustering is performed based on the fault identification model and normalized two-dimensional features to obtain the clustering results for each battery cell; the clustering results include cluster labels and clustering parameters.

[0142] Based on the clustering results, the categories of fault-free battery cells and multiple categories of battery cells to be identified were determined.

[0143] Based on the clustering results, calculate the inter-class distance between each cell class to be identified and the class of fault-free cells;

[0144] The cell categories to be identified that have an inter-class distance greater than a preset distance threshold are classified as faulty cells.

[0145] The faulty cells are identified based on their type.

[0146] In some embodiments, the method can construct a Bayesian Gaussian mixture model based on an adaptive mechanism according to the Gaussian mixture model.

[0147] In some embodiments, the adaptive Bayesian Gaussian Mixture Model (GMM) introduces Bayesian priors into the Gaussian Mixture Model (GMM). This model can achieve adaptive clustering by introducing a "separable" clustering structure through Bayesian nonparametric methods (such as the Dirichlet process) (i.e., the model allows the number of clusters to automatically adjust during the learning process), thereby enhancing the model's adaptability. Furthermore, incorporating Bayesian nonparametric methods such as the Dirichlet process into the traditional GMM architecture enables the model to dynamically adjust the number of clusters during training and achieve adaptive clustering of normal and faulty cell classes based on the average inter-class distance.

[0148] For example, Figure 2 , Figure 3 , Figure 4 , Figure 5 Four voltage fault clustering results based on feature fusion and hybrid models are shown (specifically: clustering results of faulty cells in vehicle No. 62, vehicle No. 68, vehicle No. 64, and vehicle No. 67). Based on these results, it can be seen that this application can accurately distinguish faulty cells from normal cells through cluster analysis.

[0149] In these embodiments, the method can accurately and effectively distinguish between fault-free cells and faulty cells, thereby achieving accurate diagnosis of battery pack voltage faults.

[0150] In some embodiments, the model parameters of the Bayesian-Gaussian mixture model are adaptively adjusted to obtain a fault identification model, including:

[0151] Define the prior distribution of the model parameters; where the prior distribution includes the mixing coefficients, mean, and covariance matrix;

[0152] The variational distribution is optimized by minimizing the Kullback-Leibler divergence algorithm to obtain the posterior distribution;

[0153] A fault identification model is constructed based on prior distribution, posterior distribution, optimal cluster number, and Bayesian-Gaussian mixture model.

[0154] In some embodiments, the method may first determine the initial number of clusters, then determine the variational distribution according to the variational inference algorithm, and finally select the optimal number of clusters using the Bayesian Information Criterion (BIC) or the Variational Lower Bound (ELBO).

[0155] For example, in a Gaussian Mixture Model (GMM), suppose the dataset... Composed of K Gaussian distributions, with each component corresponding to a cluster, its probability density function can be expressed as formula (14).

[0156]

[0157] Where, π k Let be the mixing coefficient of the k-th Gaussian distribution, satisfying The mean is μ k The covariance matrix is ​​∑ k The Gaussian distribution.

[0158] For example, in a Bayesian Gaussian mixture model based on an adaptive mechanism, the model parameters can be adjusted based on a Bayesian framework. Assign a prior distribution and observe the data. To update the posterior distribution. The prior distribution is usually chosen as shown in formulas (17) to (18).

[0159] π k ~Dirichlet(α)(17)

[0160]

[0161] Σ k ~Inverse-Wishart(W0,ν0)(19)

[0162] Where α, μ0, κ0, W0 and ν0 are prior hyperparameters.

[0163] To achieve adaptive clustering, the model allows the number of clusters K to automatically adjust during the learning process. It also utilizes a scalable GMM framework, enabling the addition or merging of clusters in a data-driven manner. This is due to the posterior distribution... Since an analytical solution is difficult to obtain, a variational inference method is used for approximation. A variational distribution q(θ) is introduced, and the Kullback-Leibler divergence is minimized. To optimize the parameters.

[0164] In these embodiments, the method can adaptively construct a Bayesian-Gaussian mixture model for battery pack fault identification by setting a reasonable prior distribution of model parameters (including mixing coefficients, mean, and covariance matrix) and optimizing the variational distribution using the minimization of the Kullback-Leibler divergence algorithm to obtain the posterior distribution. Then, by combining the prior distribution, the posterior distribution, and the optimal number of clusters, a Bayesian-Gaussian mixture model for battery pack fault identification can be constructed, thereby effectively improving the accuracy of fault diagnosis and the model's adaptability to complex fault modes.

[0165] In some embodiments, the method further includes:

[0166] Obtain the normalized anomaly score characteristics and normalized Z-score characteristics of the faulty battery cell;

[0167] The state space of each faulty cell is constructed based on the normalized anomaly score feature and the normalized Z-score feature;

[0168] Based on the state space, faulty cells are classified into sudden faulty cells and gradual faulty cells.

[0169] Output a second early warning message for cells experiencing sudden and gradual failures.

[0170] In some embodiments, the distinction between progressive and abrupt faults is achieved by further constructing a state space consisting of normalized anomaly scores and normalized Z-scores within the fault class, and then differentiating between abrupt and progressive faults based on the partitioning results. This distinction enhances the discriminative power and diagnostic reliability of battery fault type identification.

[0171] For example, when determining whether a problem is faulty or normal, this method can first consider K categories, each containing |C| samples. i | The category with the most samples is selected as the fault-free class;

[0172] Then, for any two categories, calculate the inter-class distance d. ij And for each fault class, calculate its distance d from the fault-free class. i，normal If d i，normal >d max If d, then the class is determined to be a fault class, where d max The above-mentioned preset distance threshold.

[0173] For example, when determining progressive and abrupt failures, this method can be based on a two-dimensional state space of feature fusion. The (anomaly score-Z-score plane) consists of two dimensions: anomaly score and Z-score. x is defined in the state space. threshold and y threshold Let R1 represent the maximum value of the normalized anomaly score and the normalized Z score in the fault class, respectively. Then the state space can be divided into the sudden fault region R1 and the progressive fault region R2.

[0174] In these embodiments, the method can improve the precision of battery pack voltage fault diagnosis and the accuracy of early warning.

[0175] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in this application will be described clearly and completely below. In some embodiments, the battery pack voltage fault diagnosis method can be implemented according to... Figure 6 The voltage fault diagnosis process flowchart based on feature fusion and hybrid model is shown and implemented.

[0176] Figure 7 A schematic diagram of a battery pack voltage fault diagnosis device is shown. It should be understood that this device is related to... Figure 1 The method executed in the middle corresponds to the steps involved in the aforementioned method. The specific functions and effects of the device can be found in the description above. To avoid repetition, detailed descriptions are omitted here.

[0177] The battery pack voltage fault diagnosis device includes:

[0178] The acquisition unit 210 is used to acquire the charging voltage sequence of each cell in the battery pack;

[0179] The calculation unit 220 is used to calculate the normalized anomaly fraction feature and the normalized Z-score feature of each cell based on the charging voltage sequence.

[0180] Construction unit 230 is used to construct normalized two-dimensional features for each cell based on normalized anomaly score features and normalized Z score features;

[0181] The diagnostic unit 240 is used to perform fault diagnosis through a pre-built Bayesian-Gaussian mixture model based on an adaptive mechanism and normalized two-dimensional features to identify faulty battery cells.

[0182] Output unit 250 is used to output the first warning information for the faulty battery cell.

[0183] In some embodiments, the computing unit 220 includes:

[0184] The preprocessing subunit 221 is used to preprocess the charging voltage sequence to obtain a preprocessed voltage sequence;

[0185] Sub-unit 222 is used to divide the preprocessed voltage sequence through a preset sliding window to obtain multiple division sequences corresponding to each cell;

[0186] The first calculation subunit 223 is used to calculate the normalized anomaly score feature corresponding to each cell based on multiple partitioning sequences; and to calculate the normalized Z score feature corresponding to each cell based on multiple partitioning sequences.

[0187] In some embodiments, the first computing subunit 223 includes:

[0188] The estimation module is used to perform kernel density estimation on multiple partition sequences to obtain the voltage value of each cell at each time point;

[0189] The calculation module is used to calculate the anomaly score of each cell at each time point based on the voltage value of each cell at each time point, and obtain the time-based anomaly score sequence of each cell;

[0190] The determination module is used to determine the maximum value on the anomaly score sequence as the normalized anomaly score characteristic of each cell.

[0191] In some embodiments, the first computing subunit 223 includes:

[0192] The calculation module is also used to calculate the cell voltage deviation corresponding to each of the multiple division sequences.

[0193] The calculation module is also used to calculate the average value of the absolute value of the voltage deviation corresponding to each division sequence and the standard deviation of the absolute value of the voltage deviation corresponding to each division sequence based on the cell voltage deviation.

[0194] The calculation module is also used to calculate the standardized Z-score corresponding to each partition sequence based on the mean and standard deviation;

[0195] The determination module is also used to determine the normalized Z-score characteristics of each cell based on the standardized Z-score.

[0196] In some embodiments, the diagnostic unit 240 includes:

[0197] Acquisition subunit 241 is used to acquire a pre-constructed Bayesian Gaussian mixture model based on an adaptive mechanism; wherein, the posterior distribution of the Bayesian Gaussian mixture model is determined using a variational distribution algorithm and a KL divergence minimization algorithm;

[0198] Adjustment subunit 242 is used to adaptively adjust the model parameters of the Bayesian-Gaussian mixture model to obtain the fault identification model;

[0199] Clustering subunit 243 is used to perform clustering processing based on the fault identification model and normalized two-dimensional features to obtain the clustering result corresponding to each cell; wherein, the clustering result includes clustering label and clustering parameters;

[0200] Determine subunit 244, which is used to determine the fault-free cell class and multiple cell classes to be identified based on the clustering results;

[0201] The second calculation subunit 245 is used to calculate the inter-class distance between each cell class to be identified and the cell class without faults based on the clustering results.

[0202] Subunit 244 is also used to identify the cell category to be identified as a faulty cell category when the inter-class distance is greater than a preset distance threshold.

[0203] The determination subunit 244 is also used to determine the faulty cells that have faults based on the type of faulty cells.

[0204] In some embodiments, the battery pack voltage fault diagnosis device further includes:

[0205] The acquisition unit 210 is also used to acquire the normalized anomaly fraction feature and the normalized Z-score feature of the faulty battery cell;

[0206] The construction unit 230 is also used to construct the state space of each faulty cell based on the normalized anomaly score characteristics and the normalized Z score characteristics;

[0207] Classification unit 260 is used to classify faulty cells according to the state space to obtain sudden faulty cells and progressive faulty cells;

[0208] The output unit 250 is also used to output a second warning message for cells with sudden and gradual failures.

[0209] like Figure 8 As shown, this application provides an electronic device 300, which includes a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanism (not shown). The memory 302 stores a computer program that can be executed by the processor 301. When the computing device is running, the processor 301 executes the computer program to perform the method in any of the aforementioned optional implementations.

[0210] This application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the method in any of the aforementioned optional implementations.

[0211] The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0212] This application provides a computer program product, which includes a computer program that, when run by a processor, executes the method in any of the aforementioned optional implementations.

[0213] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application. In particular, as long as there is no conflict, the various technical features mentioned in the embodiments can be combined in any way. This application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for diagnosing battery pack voltage faults, characterized in that, include: Obtain the charging voltage sequence of each cell in the battery pack; Based on the charging voltage sequence, calculate the normalized anomaly fraction feature and normalized Z-score feature for each cell; Based on the normalized anomaly score feature and the normalized Z score feature, a normalized two-dimensional feature for each battery cell is constructed. Fault diagnosis is performed using a pre-constructed Bayesian-Gaussian mixture model based on an adaptive mechanism and the normalized two-dimensional features to identify faulty battery cells. Output the first warning information for the faulty battery cell.

2. The battery pack voltage fault diagnosis method according to claim 1, characterized in that, The step of calculating the normalized anomaly fraction feature and normalized Z-score feature of each cell based on the charging voltage sequence includes: The charging voltage sequence is preprocessed to obtain a preprocessed voltage sequence; The preprocessed voltage sequence is divided by a preset sliding window to obtain multiple division sequences corresponding to each cell; The normalized anomaly score feature corresponding to each cell is calculated based on the multiple partitioning sequences; and the normalized Z score feature corresponding to each cell is calculated based on the multiple partitioning sequences.

3. The battery pack voltage fault diagnosis method according to claim 2, characterized in that, The step of calculating the normalized anomaly score feature corresponding to each cell based on the multiple partitioning sequences includes: Kernel density estimation is performed on the multiple partition sequences to obtain the voltage value of each cell at each time point; Based on the voltage value of each cell at each time point, calculate the anomaly score of each cell at each time point to obtain a time-based anomaly score sequence for each cell; The maximum value on the anomaly score sequence is determined as the normalized anomaly score feature for each cell.

4. The battery pack voltage fault diagnosis method according to claim 2, characterized in that, The step of calculating the normalized Z-score feature corresponding to each cell based on the multiple partition sequences includes: Based on the plurality of division sequences, the cell voltage deviation corresponding to each division sequence is calculated; Based on the cell voltage deviation, calculate the average value of the absolute value of the voltage deviation corresponding to each of the division sequences and the standard deviation of the absolute value of the voltage deviation corresponding to each of the division sequences; Calculate the standardized Z-score for each partition sequence based on the mean and the standard deviation; The normalized Z-score characteristics of each cell are determined based on the standardized Z-score.

5. The battery pack voltage fault diagnosis method according to claim 1, characterized in that, The fault diagnosis, performed using a pre-constructed adaptive Bayesian-Gaussian mixture model and the normalized two-dimensional features, identifies faulty battery cells, including: Obtain a pre-constructed Bayesian Gaussian mixture model based on an adaptive mechanism; wherein, the posterior distribution of the Bayesian Gaussian mixture model is determined using a variational distribution algorithm and a KL divergence minimization algorithm; The model parameters of the Bayesian-Gaussian mixture model are adaptively adjusted to obtain a fault identification model; Clustering is performed based on the fault identification model and the normalized two-dimensional features to obtain the clustering result for each battery cell; wherein, the clustering result includes clustering labels and clustering parameters; Based on the clustering results, the categories of fault-free battery cells and multiple categories of battery cells to be identified are determined. Based on the clustering results, calculate the inter-class distance between each of the cell classes to be identified and the class of fault-free cells; The cell categories to be identified that have an inter-class distance greater than a preset distance threshold are classified as faulty cells. The faulty cells are identified based on the aforementioned faulty cell type.

6. The battery pack voltage fault diagnosis method according to claim 1, characterized in that, The method further includes: Obtain the normalized anomaly score feature and normalized Z score feature of the faulty battery cell; The state space of each faulty cell is constructed based on the normalized anomaly score feature and the normalized Z score feature; The faulty cells are classified according to the state space to obtain sudden faulty cells and gradual faulty cells; Output a second early warning message for the sudden failure battery cell and the gradual failure battery cell.

7. A battery pack voltage fault diagnosis device, characterized in that, The battery pack voltage fault diagnosis device includes: The acquisition unit is used to acquire the charging voltage sequence of each cell in the battery pack; The calculation unit is used to calculate the normalized anomaly fraction feature and the normalized Z-score feature of each cell based on the charging voltage sequence. A construction unit is used to construct normalized two-dimensional features for each cell based on the normalized anomaly score features and the normalized Z score features. The diagnostic unit is used to perform fault diagnosis using a pre-built Bayesian-Gaussian mixture model based on an adaptive mechanism and the normalized two-dimensional features to identify faulty battery cells. The output unit is used to output the first warning information for the faulty battery cell.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to cause the electronic device to perform the battery pack voltage fault diagnosis method according to any one of claims 1 to 6.

9. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, performs the battery pack voltage fault diagnosis method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, performs the battery pack voltage fault diagnosis method according to any one of claims 1 to 6.

Images