Electric vehicle battery fault analysis method, system, equipment and medium

Abstract

The invention belongs to the technical field of battery data processing, and particularly relates to an electric vehicle battery fault analysis method, system and device and a medium, and the method comprises the steps: firstly collecting multi-dimensional time sequence data of a battery pack, and generating a target matrix through a preset processing strategy; thirdly, extracting feature vectors in the battery data by using a convolutional neural network, storing the feature vectors in a vector database, and adding fault / normal labels; and then, based on a vector clustering method and an Euclidean distance, the similarity between incremental data and historical data is analyzed, and the fault state of the battery pack is judged. And finally, performing fault analysis according to a classification result. A time sequence database and an early warning mechanism are further introduced, the change of battery characteristics can be monitored in real time, and fault early warning can be triggered in time. Through deep learning and data analysis technologies, the intelligent level of battery fault diagnosis can be effectively improved, and the safety and reliability of the battery pack are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery data processing, and in particular relates to an electric vehicle battery fault analysis method, system, equipment and medium. Background Art

[0002] With the rapid development of new energy technologies, lithium batteries have been widely used in the field of electric vehicles. In order to ensure the safe operation of electric vehicles and optimize battery maintenance strategies, it is crucial to accurately judge the health status of the battery. At present, a variety of battery fault detection methods have been proposed and applied in practice.

[0003] Although existing battery fault detection methods can monitor and diagnose battery status to a certain extent, most of them are based on traditional rule-based methods and lack effective application of deep learning. For example, the invention patent with patent publication number CN118348444A discloses a battery pack fault intelligent detection system based on data analysis. The battery pack monitoring and acquisition module collects the operating data of the battery pack in real time, the feature analysis and extraction module pre-processes and extracts features from the received data, and the battery pack fault identification module identifies the battery pack fault based on the fault identification model, which can detect the battery pack fault in time.

[0004] However, when processing a large amount of incremental data from batteries, the above scheme often finds it difficult to quickly consider the feature differences between normal batteries and faulty batteries at the same time, and is unable to quickly classify and predict the incremental data. In addition, the above scheme does not have a clear dynamic update mechanism and is unable to update the feature vector library in a timely manner, resulting in the classification results being unable to adapt to changes in battery status. Summary of the invention

[0005] The purpose of the present invention is to provide an electric vehicle battery fault analysis method, system, device and medium, which can efficiently extract the implicit relationship between the multi-dimensional features of the battery, distinguish the characteristic differences between normal batteries and faulty batteries, and realize battery status classification of incremental data based on vector retrieval.

[0006] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0007] In a first aspect, the present invention provides a method for analyzing electric vehicle battery failure, the method comprising:

[0008] Process the multi-dimensional time series data of the battery pack operation collected in advance to obtain a target matrix and mark it as historical data. The target matrix includes voltage V, current I, temperature T, SOC information, and is an s×4 two-dimensional matrix;

[0009] The CNN convolutional neural network is used to build an extraction model, with voltage V, current I, and temperature T as input and SOC as output to extract feature vectors related to normal batteries and faulty batteries;

[0010] Storing the historical data and the feature vector in a vector database and adding a fault / normal label;

[0011] Acquire the target matrix of the target battery pack and record it as incremental data, and use the feature vector extracted by the extraction model;

[0012] Based on the vector clustering method and the Euclidean distance between vectors, the candidate feature vectors of the incremental data are determined, and similarity retrieval is performed with the feature vectors in the vector database, and the classification status of the incremental data is judged according to the similarity ratio;

[0013] A failure analysis result of the target battery pack is determined according to the classification status.

[0014] Furthermore, the method further includes: setting a judgment index of a normal battery and a faulty battery based on the attribute information of the target category battery pack, and when the judgment index meets a preset condition, judging the battery pack as a faulty battery, otherwise judging it as a normal battery;

[0015] Among them, the preset conditions include at least one or more of the presence of lithium plating, a range reduction exceeding a preset threshold, a battery temperature increase exceeding a preset threshold, and a voltage increase exceeding a preset threshold.

[0016] Furthermore, the multi-dimensional time series data of the battery pack operation collected in advance is processed to obtain a target matrix, including:

[0017] Collect multi-dimensional data of battery pack operation through vehicle BMS or cloud platform, including voltage V, current I, temperature T, and SOC information in time series;

[0018] The multi-dimensional data is divided into multiple subsequences X by sliding according to the set window length s j , X j =(x1,x2,...,x s ), where X j belongs to {V,I,T,SOC};

[0019] Normalizing the plurality of subsequences;

[0020] Perform data expansion on the normalized subsequences to generate an input matrix X jx And the output matrix X jy : X jy =(soc s ).

[0021] Furthermore, the extraction model is constructed using CNN convolutional neural network, and its model structure is as follows:

[0022] Input layer, used to accept the feature matrix of shape s×3;

[0023] One-dimensional convolutional layer, used to capture the feature relationship between V, I, and T in the time dimension, and output a multi-channel feature map with a shape of s×c, where c is the number of channels;

[0024] Pooling layer, used to downsample the time dimension, with an output shape of s′×c;

[0025] The fully connected layer is used to expand the output of the pooling layer into a one-dimensional vector, map it to a high-dimensional feature space through several hidden layers, and extract the implicit feature vector h, where the feature vector of a normal battery is h normal , the characteristic vector of the faulty battery is h faulty ;

[0026] Output layer: Output SOC prediction value.

[0027] Furthermore, the method of determining the candidate feature vectors corresponding to the incremental data based on the vector clustering method and the Euclidean distance between vectors, performing similarity retrieval with the feature vectors in the vector database, and judging the classification status of the incremental data according to the similarity ratio includes:

[0028] The feature vector h in the incremental data is clustered by a fast clustering algorithm. new_normal and h new_faulty Perform preliminary clustering to form clusters;

[0029] Based on the Euclidean distance between vectors, a set of candidate feature vectors in the cluster is selected, specifically {h' normal} and {h' faulty};

[0030] h new_normal In the normal battery feature vector set {h' normal} to calculate the cosine similarity:

[0031]

[0032] h new_faulty In the faulty battery feature vector set {h' faulty} to calculate the cosine similarity:

[0033]

[0034] Statistics new_normal In {h' normalThe number of vectors N whose similarity is greater than the threshold value 0.9 in normal , and calculate the proportion:

[0035]

[0036] Statistics new_faulty In {h' faulty The number of vectors N whose similarity is greater than the threshold value 0.9 in faulty , and calculate the proportion:

[0037]

[0038] If P normal >P faulty , the sample is classified as a “normal battery”, otherwise it is a “faulty battery”.

[0039] Furthermore, the method further includes: storing the incremental data and the corresponding classification status in the vector database to update the feature vector library of normal batteries and faulty batteries in the vector database.

[0040] Furthermore, the method further comprises:

[0041] The extracted feature vectors are stored in time series to establish a time series database;

[0042] Monitoring the time series changes of feature vectors in the time series database based on machine learning or deep learning algorithms;

[0043] When the time series change of the feature vector exceeds the warning threshold, the warning mechanism is triggered and a warning model is issued.

[0044] In a second aspect, the present invention proposes an electric vehicle battery fault analysis system, which is used to execute any of the above battery fault analysis methods, and the system includes:

[0045] Data acquisition module, used to collect multi-dimensional time series data of battery pack operation, including voltage V, current I, temperature T, and SOC information;

[0046] A data processing module is used to process the dimensional time series data to obtain a target matrix, specifically an s×4 two-dimensional matrix;

[0047] The first extraction module is used to construct an extraction model using a CNN convolutional neural network, taking voltage V, current I, and temperature T as inputs and SOC as output to extract feature vectors related to normal batteries and faulty batteries;

[0048] A vector database, used to store the historical data and the feature vector and its label information, where the label information is a fault / normal label;

[0049] A second extraction module, used for acquiring the target matrix of the target battery pack and recording it as incremental data, and extracting the feature vector using the extraction model;

[0050] A fault diagnosis module, used to determine the candidate feature vectors corresponding to the incremental data based on the vector clustering method and the Euclidean distance between vectors, and perform similarity search with the feature vectors in the vector database, and judge the classification status of the incremental data according to the similarity ratio;

[0051] A fault analysis module, used to determine a fault analysis result of the target battery pack according to the classification state;

[0052] The vector database is also used to store the incremental data and the corresponding classification status;

[0053] The system also includes a time series database and an early warning module. The time series database is used to store the extracted feature vectors in time series; the early warning module is used to monitor the time series changes of the feature vectors in the time series database based on machine learning or deep learning algorithms; when the time series changes of the feature vectors exceed the early warning threshold, the early warning mechanism is triggered and an early warning model is issued.

[0054] In a third aspect, the present invention provides an electronic device, comprising:

[0055] A processor; and a memory for storing instructions executable by the processor;

[0056] The processor is configured to execute the instructions to implement the battery failure analysis method as described in any one of the above items.

[0057] In a fourth aspect, the present invention proposes a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute any of the battery failure analysis methods described above.

[0058] The beneficial effects of the present invention are:

[0059] 1. The present invention can accurately capture the difference between normal and faulty states by constructing a model based on deep learning convolutional neural networks; it uses a vector database for similarity retrieval and classification, and realizes rapid classification and fault diagnosis of incremental data, significantly improving the accuracy and efficiency of fault diagnosis. This method not only takes into account the multi-dimensional time series data of the battery, but also further optimizes the model parameters by monitoring the time series changes of the feature vectors, and enhances the model's adaptability to different operating conditions.

[0060] 2. The present invention can monitor the health status of the battery pack in real time and detect potential faults in a timely manner by comparing and analyzing historical data and incremental data, thereby reducing the safety risks caused by battery failures and extending the service life of the battery.

[0061] 3. The present invention dynamically classifies incremental data and automatically updates the battery feature vector library in the vector database, so that the system can continuously optimize itself and adapt to the feature changes of different battery packs. Through the early warning mechanism, the system can issue an alarm in time when the battery features change abnormally, providing effective support for fault prevention and maintenance of electric vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a flow chart of a method for analyzing a battery failure in an electric vehicle provided in an embodiment of the present application;

[0063] Figure 2 A schematic diagram of another process of the electric vehicle battery failure analysis method provided in the embodiment of the present application;

[0064] Figure 3 A system architecture diagram of an electric vehicle battery fault analysis system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0066] Example 1

[0067] like Figure 1-2 As shown, this embodiment proposes a method for analyzing electric vehicle battery failure, the method comprising the following steps:

[0068] S1. Based on the attribute information of the target category battery pack, the judgment index of the normal battery and the faulty battery is set. When the judgment index meets the preset conditions, the battery pack is judged to be a faulty battery, otherwise it is judged to be a normal battery; wherein the preset conditions include at least one or more of the presence of lithium plating, the range reduction amplitude exceeding the preset threshold, the battery temperature increase amplitude exceeding the preset threshold and the voltage increase amplitude exceeding the preset threshold.

[0069] S2. Process the multi-dimensional time series data of the battery pack operation collected in advance to obtain a target matrix and mark it as historical data. The target matrix includes voltage V, current I, temperature T, SOC (State of Charge) information, and is an s×4 two-dimensional matrix.

[0070] S3. Use CNN convolutional neural network to build an extraction model, with voltage V, current I, temperature T as input and SOC as output to extract feature vectors related to normal batteries and faulty batteries.

[0071] S4. Store the historical data and feature vectors into a vector database and attach fault / normal labels;

[0072] S5. Obtain the target matrix of the target battery pack and record it as incremental data, and use the feature vector extracted by the extraction model.

[0073] S6. Based on the vector clustering method and the Euclidean distance between vectors, determine the candidate feature vectors of the incremental data, perform similarity search with the feature vectors in the vector database, and judge the classification status of the incremental data based on the similarity ratio.

[0074] S7. Determine the fault analysis result of the target battery pack according to the classification status.

[0075] In the above embodiment, constructing the vector database is specifically as follows:

[0076] Pass historical data through CNN normal and CNN faulty (Two extraction models) The extracted feature vector h normal and h faulty , stored in the vector database respectively, and labeled (0 for normal, 1 for fault).

[0077] Preferably, in step S2, the multi-dimensional time series data of the battery pack operation collected in advance is processed to obtain a target matrix, including:

[0078] Data collection: Collect multi-dimensional data of battery pack operation through vehicle BMS or cloud platform, including voltage V, current I, temperature T, and SOC information in time series;

[0079] Time dimension separation: Slide and split the multi-dimensional data into multiple subsequences X according to the set window length s j , X j =(x1,x2,...,x s ), where X j belongs to {V,I,T,SOC};

[0080] Feature normalization: normalize multiple subsequences;

[0081] More specifically, the voltage, current, temperature, and SOC data are normalized and each indicator is mapped to the [0,1] interval to ensure that the dimensional differences between different features do not affect model training: Normalization formula: Where X is the original data, Xmax and X min are the maximum and minimum values ​​of the feature respectively.

[0082] Sample enhancement: The data is expanded through random perturbations, small adjustments to voltage and current, etc. to enhance the generalization ability of the model.

[0083] More specifically, the input matrix X of each sample after processing j for:

[0084]

[0085] Each sample contains s time points, each of which includes V, I, T, and SOC data, that is, the input matrix X j It is a two-dimensional matrix of s×4, each row represents the information of a time point, and each column is: voltage, current, temperature, SOC.

[0086] Perform data expansion on multiple subsequences after normalization to generate the input matrix X jx And the output matrix X jy :

[0087] X jy =(soc s ).

[0088] Preferably, in step S3, a CNN convolutional neural network is used to construct an extraction model, and its model structure is as follows:

[0089] Input layer, used to accept the feature matrix of shape s×3;

[0090] One-dimensional convolutional layer, used to capture the feature relationship between V, I, and T in the time dimension, and output a multi-channel feature map with a shape of s×c, where c is the number of channels;

[0091] Pooling layer, used to downsample the time dimension, with an output shape of s′×c;

[0092] The fully connected layer is used to expand the output of the pooling layer into a one-dimensional vector, map it to a high-dimensional feature space through several hidden layers, and extract the implicit feature vector h, where the feature vector of a normal battery is h normal , the characteristic vector of the faulty battery is h faulty ;

[0093] Output layer: Output SOC prediction value.

[0094] It is understandable that the present application also includes extraction model training, and the training process is to train two model CNNs for normal battery data and faulty battery data respectively. normal and CNN fau lty , whose loss function is the mean square error (MSE): Where N represents the total number of training samples, f CNN Represents the mapping function of the model. Optimization method: Use adaptive optimization algorithms (such as Adam) to adjust model parameters, set learning rate, regularization parameters, etc.

[0095] The implementation also includes verification of the two extraction model outputs to ensure that the feature vectors are discriminative:

[0096] Visual verification: Use dimensionality reduction techniques (such as PCA and t-SNE) to visualize the feature vectors and check the distribution differences between normal and fault features.

[0097] Vector consistency check: Calculate the distance distribution within the feature vector to ensure the stability of the model output.

[0098] Further preferably, in step S6, based on the vector clustering method and the Euclidean distance between vectors, the candidate feature vectors corresponding to the incremental data are determined, and similarity retrieval is performed with the feature vectors in the vector database, and the classification status of the incremental data is judged according to the similarity ratio, including:

[0099] The feature vector h in the incremental data is clustered by a fast clustering algorithm (such as K-means). new_normal and h new_faulty Perform preliminary clustering to form clusters;

[0100] Based on the Euclidean distance between vectors, a set of candidate feature vectors in the cluster is selected, specifically {h' normal} and {h' faulty};

[0101] h new_normal In the normal battery feature vector set {h' normal} to calculate the cosine similarity:

[0102]

[0103] h new_faulty In the faulty battery feature vector set {h' faulty} to calculate the cosine similarity:

[0104]

[0105] Statistics new_normal In {h' normal The number of vectors N whose similarity is greater than the threshold value 0.9 in normal , and calculate the proportion:

[0106]

[0107] Statistics new_faulty In {h' faulty The number of vectors N whose similarity is greater than the threshold value 0.9 in faulty , and calculate the proportion:

[0108]

[0109] If P normal >P faulty , the sample is classified as a “normal battery”, otherwise it is a “faulty battery”. That is, if the similarity ratio between the incremental data and the normal battery is greater than the similarity ratio between the incremental data and the faulty battery, it is judged as a “normal battery”; otherwise, it is judged as a “faulty battery”.

[0110] It is understandable that the above-mentioned incremental data classification step of the present application combines the vector database with the efficient cosine similarity retrieval algorithm, which significantly improves the accuracy and timeliness of battery fault diagnosis. By storing historical feature vectors in the vector database and using fast retrieval technology, this method can quickly classify new battery data samples without retraining the model, greatly shortening the classification time. At the same time, combined with the similarity ratio judgment strategy, the robustness of the classification results is effectively enhanced, and the interference of abnormal data on the classification results is reduced. In addition, the dynamic update mechanism of the vector database ensures the real-time nature of the feature library, so that the classification results can keep up with the changes in the battery status, providing a strong guarantee for the safe operation of electric vehicles.

[0111] Further preferably, the method further comprises: storing the incremental data and the corresponding classification status into a vector database to update a feature vector library of normal batteries and faulty batteries in the vector database.

[0112] In specific implementation, whenever new battery data (incremental data) is collected and classified, these data and their corresponding classification status (normal or faulty) will be stored in the vector database. This means that the vector database not only contains historical data, but also continuously absorbs the latest data, thus maintaining its timeliness and comprehensiveness.

[0113] In some implementations, this function can be implemented by a database management system (DBMS) or dedicated vector database software. When stored, each incremental data is assigned a unique identifier (such as a timestamp or ID) for subsequent retrieval and update.

[0114] Further preferably, the method further comprises:

[0115] The extracted feature vectors are stored in time series to establish a time series database;

[0116] Monitor the time series changes of feature vectors in the time series database based on machine learning or deep learning algorithms;

[0117] When the time series change of the feature vector exceeds the warning threshold, the warning mechanism is triggered and a warning model is issued.

[0118] In specific implementation, the extracted feature vectors will be stored in the order of the time series to form a time series database. This database can record the changes of feature vectors over time, providing a basis for subsequent monitoring and early warning. In addition, in the early warning mechanism, based on machine learning or deep learning algorithms, the system will monitor the feature vectors in the time series database in real time. These algorithms can learn the normal change trend of feature vectors and set a warning threshold. When the time series change of the feature vector exceeds this threshold, the system will trigger the early warning mechanism and send out a warning signal, indicating that the battery may have a potential fault.

[0119] It should be noted that the functions of the vector database and the time series database in the above embodiment are defined as follows:

[0120] Vector database: stores feature vectors and their classification status for similarity calculation and classification.

[0121] Time series database: stores time series data of feature vectors for time series trend monitoring and early warning.

[0122] In another embodiment provided by the present invention, in combination with 3, based on the same inventive concept, a battery failure analysis system for an electric vehicle is also provided, which is applied to execute the above-mentioned battery failure analysis method, and the system includes:

[0123] Data acquisition module, used to collect multi-dimensional time series data of battery pack operation, including voltage V, current I, temperature T, and SOC information;

[0124] The data processing module is used to process the dimensional time series data to obtain the target matrix, specifically an s×4 two-dimensional matrix;

[0125] The first extraction module is used to construct an extraction model using a CNN convolutional neural network, taking voltage V, current I, and temperature T as inputs and SOC as output to extract feature vectors related to normal batteries and faulty batteries;

[0126] Vector database, used to store historical data and feature vectors and their label information, where the label information is fault / normal label;

[0127] A second extraction module is used to obtain a target matrix of a target battery pack and record it as incremental data, using a feature vector extracted by the extraction model;

[0128] A fault diagnosis module is used to determine the candidate feature vectors corresponding to the incremental data based on the vector clustering method and the Euclidean distance between vectors, and to perform similarity retrieval with the feature vectors in the vector database, and to judge the classification status of the incremental data according to the similarity ratio;

[0129] A fault analysis module, used to determine a fault analysis result of a target battery pack according to a classification state;

[0130] The vector database is also used to store incremental data and corresponding classification status;

[0131] The system also includes a time series database and an early warning module. The time series database is used to store the extracted feature vectors in time series; the early warning module is used to monitor the time series changes of the feature vectors in the time series database based on machine learning or deep learning algorithms; when the time series changes of the feature vectors exceed the early warning threshold, the early warning mechanism is triggered and an early warning model is issued.

[0132] It should be noted here that each module in the above-mentioned fault analysis system corresponds to each step in implementing the above-mentioned fault analysis method, and the instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0133] According to the above embodiment of the present invention, the system obtains multi-dimensional time series data of the battery pack through the data acquisition module, and forms a target matrix after data processing. The CNN convolutional neural network is used to extract feature vectors, and similarity retrieval is performed with historical feature vectors in the vector database to achieve accurate diagnosis of battery faults. At the same time, the system establishes a time series database to monitor the time series changes of feature vectors, and triggers an early warning when the change exceeds the early warning threshold, so as to achieve early detection and prevention of faults. The system not only improves the accuracy and timeliness of battery fault analysis, but also reduces the risk of faults through the early warning mechanism, providing a strong guarantee for the safe operation of electric vehicles.

[0134] In another embodiment of the present invention, an electronic device is provided, including:

[0135] a processor; and a memory for storing instructions executable by the processor;

[0136] The processor is configured to execute instructions to implement the battery failure analysis method as described above.

[0137] In another embodiment of the present invention, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute the battery failure analysis method as described above.

[0138] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0139] In addition, each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0140] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for analyzing electric vehicle battery failure, characterized in that: The method comprises: Process the multi-dimensional time series data of the battery pack operation collected in advance to obtain a target matrix and mark it as historical data. The target matrix includes voltage V, current I, temperature T, SOC information, and is an s×4 two-dimensional matrix; The extraction model is constructed using CNN convolutional neural network, with voltage V, current I, temperature T as input and SOC as output to extract feature vectors related to normal batteries and faulty batteries; Storing the historical data and the feature vector in a vector database and adding a fault / normal label; Acquire the target matrix of the target battery pack and record it as incremental data, and use the feature vector extracted by the extraction model; Based on the vector clustering method and the Euclidean distance between vectors, the candidate feature vectors of the incremental data are determined, and similarity retrieval is performed with the feature vectors in the vector database, and the classification status of the incremental data is judged according to the similarity ratio; A failure analysis result of the target battery pack is determined according to the classification status.

2. The electric vehicle battery failure analysis method according to claim 1, characterized in that: The method further includes: setting a judgment index of a normal battery and a faulty battery based on the attribute information of the target category battery pack, and judging the battery pack as a faulty battery when the judgment index meets a preset condition, otherwise judging the battery pack as a normal battery; Among them, the preset conditions include at least one or more of the presence of lithium plating, a range reduction exceeding a preset threshold, a battery temperature increase exceeding a preset threshold, and a voltage increase exceeding a preset threshold.

3. The electric vehicle battery failure analysis method according to claim 2, characterized in that: The multi-dimensional time series data of the battery pack operation collected in advance is processed to obtain a target matrix, including: Collect multi-dimensional data of battery pack operation through vehicle BMS or cloud platform, including voltage V, current I, temperature T, and SOC information in time series; The multi-dimensional data is divided into multiple subsequences X by sliding according to the set window length s j , X j =(x1,x2,...,x s ), where X j belongs to {V,I,T,SOC}; Normalizing the plurality of subsequences; Perform data expansion on the normalized multiple subsequences to generate an input matrix X jx And the output matrix X jy : X jy =(soc s ).

4. The electric vehicle battery failure analysis method according to claim 3, characterized in that: The extraction model is constructed using a CNN convolutional neural network, and its model structure is as follows: Input layer, used to accept the feature matrix of shape s×3; One-dimensional convolutional layer, used to capture the feature relationship between V, I, and T in the time dimension, and output a multi-channel feature map with a shape of s×c, where c is the number of channels; Pooling layer, used to downsample the time dimension, with an output shape of s′×c; The fully connected layer is used to expand the output of the pooling layer into a one-dimensional vector, map it to a high-dimensional feature space through several hidden layers, and extract the implicit feature vector h, where the feature vector of a normal battery is h normal , the characteristic vector of the faulty battery is h faulty ; Output layer: Output SOC prediction value.

5. The electric vehicle battery failure analysis method according to claim 4, characterized in that: The method of determining the candidate feature vector corresponding to the incremental data based on the vector clustering method and the Euclidean distance between vectors, performing similarity retrieval on the feature vector in the vector database, and judging the classification status of the incremental data according to the similarity ratio includes: The feature vector h in the incremental data is clustered by a fast clustering algorithm. new_normal and h new_faulty Perform preliminary clustering to form clusters; Based on the Euclidean distance between vectors, a set of candidate feature vectors in the cluster is selected, specifically {h' normal } and {h' faulty }; h new_normal In the normal battery feature vector set {h' normal } to calculate the cosine similarity: h new_faulty In the faulty battery feature vector set {h' faulty } to calculate the cosine similarity: Statistics new_normal In {h' normal The number of vectors N whose similarity is greater than the threshold value 0.9 in normal , and calculate the proportion: Statistics new_faulty In {h' faulty The number of vectors N whose similarity is greater than the threshold value 0.9 in faulty , and calculate the proportion: If P normal >P faulty , the sample is classified as "normal battery", otherwise it is "faulty battery".

6. The electric vehicle battery failure analysis method according to claim 5, characterized in that: The method further includes: storing the incremental data and the corresponding classification status in the vector database to update the feature vector library of normal batteries and faulty batteries in the vector database.

7. The electric vehicle battery failure analysis method according to claim 1, characterized in that: The method further comprises: The extracted feature vectors are stored in time series to establish a time series database; Monitoring the time series changes of feature vectors in the time series database based on machine learning or deep learning algorithms; When the time series change of the feature vector exceeds the warning threshold, the warning mechanism is triggered and a warning model is issued.

8. An electric vehicle battery fault analysis system, used to execute the battery fault analysis method according to any one of claims 1 to 7, characterized in that: The system comprises: Data acquisition module, used to collect multi-dimensional time series data of battery pack operation, including voltage V, current I, temperature T, and SOC information; A data processing module is used to process the dimensional time series data to obtain a target matrix, specifically an s×4 two-dimensional matrix; The first extraction module is used to construct an extraction model using a CNN convolutional neural network, taking voltage V, current I, and temperature T as inputs and SOC as output to extract feature vectors related to normal batteries and faulty batteries; A vector database, used to store the historical data and the feature vector and its label information, where the label information is a fault / normal label; A second extraction module, used for acquiring the target matrix of the target battery pack and recording it as incremental data, and extracting the feature vector using the extraction model; A fault diagnosis module, used to determine the candidate feature vectors corresponding to the incremental data based on the vector clustering method and the Euclidean distance between vectors, and perform similarity search with the feature vectors in the vector database, and judge the classification status of the incremental data according to the similarity ratio; A fault analysis module, used to determine a fault analysis result of the target battery pack according to the classification state; The vector database is also used to store the incremental data and the corresponding classification status; The system also includes a time series database and an early warning module. The time series database is used to store the extracted feature vectors in time series; the early warning module is used to monitor the time series changes of the feature vectors in the time series database based on machine learning or deep learning algorithms; when the time series changes of the feature vectors exceed the early warning threshold, the early warning mechanism is triggered and an early warning model is issued.

9. An electronic device, characterized in that: include: processor; and a memory for storing instructions executable by said processor; The processor is configured to execute the instructions to implement the battery failure analysis method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the battery failure analysis method according to any one of claims 1 to 7.

Images