Battery fault diagnosis method based on graph structure and lstm

Abstract

The application discloses a battery fault diagnosis method based on a graph structure and an LSTM, belongs to the field of battery fault diagnosis, and comprises a battery pack space connection information representation method based on a graph structure, which converts the battery blocks of a battery module and the connection relationship thereof into a graph model, so that the spatial dependency between the battery blocks can be effectively captured; meanwhile, by combining a graph neural network with a long short-term memory network, the application can accurately extract the space-time features of the battery state, enhance the processing capacity of time series data, and thus improve the accuracy and comprehensiveness of fault diagnosis; in addition, in view of the problem that there is no confidence evaluation mechanism in the prior art, the application designs an accurate confidence evaluation module, and the reliability and transparency of decision-making are effectively improved by calculating the confidence score of the diagnosis result.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of battery fault diagnosis, and particularly relates to a battery fault diagnosis method based on a graph structure and an LSTM. BACKGROUND

[0002] The existing battery fault diagnosis technology for electric vehicles mainly has the following problems: first, the spatial connection information of the battery pack is underutilized, traditional methods mostly rely on single feature data or simple clustering analysis, and the complex spatial topological structure between battery blocks is not fully considered, resulting in that the spatial dependence relationship cannot be effectively modeled; second, the existing technology has limited ability in spatio-temporal feature extraction, most methods ignore the interaction of battery data in time and space, and cannot comprehensively consider the dynamic changes of battery state over time and the influence between different battery blocks; finally, the confidence evaluation system for the battery fault diagnosis result of the electric vehicle is not perfect, many methods lack a confidence evaluation mechanism for the diagnosis result, and cannot effectively measure the reliability of the diagnosis result, making it difficult to make reliable judgments in the face of uncertain or fuzzy fault conditions. SUMMARY

[0003] To solve the above technical problems, the application provides a battery fault diagnosis method based on a graph structure and an LSTM, which comprises:

[0004] obtaining the time series feature information of each battery block in the battery module and the state of the battery pack based on the battery management system, and constructing a time series data set based on the time series feature information of each battery block and the state of the battery pack;

[0005] constructing a graph structure data based on the hardware connection relationship of the battery blocks in the battery module;

[0006] constructing a spatio-temporal feature extraction module of battery block information based on the graph structure data and a long short-term memory network model;

[0007] constructing a calculation model based on the spatio-temporal feature extraction module and a multi-head attention mechanism, training the calculation model through the time series data set, and obtaining an electric vehicle battery fault diagnosis model;

[0008] deploying the electric vehicle battery fault diagnosis model to the BMS platform of the electric vehicle to diagnose the fault of the battery.

[0009] Preferably, the time series feature information of each battery block comprises battery voltage, current, temperature, battery health state, battery state of charge, internal resistance, and charging cycle.

[0010] Preferably, the process of constructing a time series data set based on the time series feature information of each battery block and the state of the battery pack comprises:

[0011] Predefine the feature information of each battery block in the battery module, collect the feature information of each battery block, and obtain complete battery pack input feature data;

[0012] Label the battery pack state of the complete battery pack input feature data, integrate the complete battery pack input feature data and the labeled battery pack state, and obtain the time series data set of battery fault diagnosis.

[0013] Preferably, the process of constructing the graph structure data based on the hardware connection relationship of the battery block in the battery module comprises:

[0014] Each battery block is taken as a node in the graph, and the connection relationship between the battery blocks is taken as an edge between the nodes to construct a graph data structure including spatial information;

[0015] The feature information of each battery block at a single time is taken as the feature of the corresponding node to construct the graph structure data;

[0016] Each set of battery pack input feature data should be 12 time point graph data, and the expression is:

[0017] G m =[G1,G2,...,G 12 ]=Graph(X m );

[0018] Wherein, G m represents the graph data corresponding to the input feature data obtained based on the graph structure representation method embedding, X m is the input feature data, G1 is the graph data of the first time point, G2 is the graph data of the second time point, G 12 is the graph data of the twelfth time point, and Graph(·) is the graph representation and feature embedding operation.

[0019] Preferably, the process of constructing the spatio-temporal feature extraction module of the battery block information based on the graph structure data and the long short-term memory network model comprises:

[0020] Construct the spatio-temporal feature extraction module of the battery block information based on the graph neural network and the long short-term memory network model;

[0021] Wherein, the graph neural network performs spatio-temporal feature extraction on the graph structure data to obtain global embedding features;

[0022] The long short-term memory network performs time feature extraction on the global embedding features to obtain time feature embedded intermediate features.

[0023] Preferably, in the graph neural network, the first layer graph convolution layer performs feature extraction, and the expression is:

[0024]

[0025] wherein, W1 represents a learnable weight matrix of the first graph convolutional layer; σ(·) is an activation function, G1 is graph data, A is an adjacency matrix corresponding to the graph data, is a normalized adjacency matrix; D is a degree matrix, and D is a diagonal matrix, the value of the diagonal line element is the number of adjacent nodes of the corresponding node, Emb1 is a global embedding feature obtained through the graph convolutional layer;

[0026] The expression of the intermediate feature after obtaining the time feature embedding is:

[0027] H1=LSTM(Emb1,Q1);

[0028] wherein, Q1 represents a learnable weight matrix of the first LSTM layer, H1 represents an intermediate feature after time feature embedding, and LSTM(·) represents an LSTM layer processing function.

[0029] Preferably, the second layer of graph convolutional layers in the graph neural network performs feature extraction, and the expression is:

[0030]

[0031] wherein, W2 represents a learnable weight matrix of the second graph convolutional layer, and Full(·) represents a full connection layer processing function.

[0032] Preferably, the process of training the computing model through the time series data set to obtain the electric vehicle battery fault diagnosis model comprises:

[0033] The Adam optimizer is used to optimize the cross-entropy loss function, and the network parameters are updated through back propagation multiple times to optimize the computing model, until the cross-entropy loss function tends to be minimum and the model converges, and then the electric vehicle battery fault diagnosis model is output.

[0034] Preferably, the process of diagnosing the fault of the battery based on the electric vehicle battery fault diagnosis model comprises:

[0035] Based on the spatio-temporal feature extraction module, a global embedding feature containing spatio-temporal information is obtained.

[0036] Based on the multi-head attention mechanism, the time series dependence and spatial relationship in the global embedding feature of the spatio-temporal information are captured to obtain an attention output.

[0037] The attention output is weighted and fused to obtain a final multi-head attention output.

[0038] The final multi-head attention output is calculated through a full connection layer and a residual connection to obtain processed feature data;

[0039] The processed feature data is calculated through a Dropout layer to obtain output features.

[0040] The output features are converted into probability values of classification decisions by taking a Softmax layer as an output layer to obtain a final prediction result.

[0041] Preferably, after the final prediction result is obtained, the method further comprises:

[0042] A confidence evaluation method for battery fault diagnosis model output results is constructed, the final prediction result is evaluated based on the confidence evaluation method to obtain an accuracy rate of each diagnosis type;

[0043] The mean and variance of all prediction probabilities under the fault type are calculated based on the accuracy rate of each diagnosis type, and a confidence score and a confidence threshold are obtained based on the mean and variance.

[0044] If the confidence score is greater than the confidence threshold, a diagnosis result and a corresponding confidence score are output, and if the confidence score is less than the confidence threshold, artificial intervention is performed.

[0045] Compared with the prior art, the present application has the following advantages and technical effects:

[0046] The combination of the graph neural network and the LSTM-based spatio-temporal feature extraction method can accurately model the spatial and time sequence dependencies in the battery system, and improve the accuracy of fault diagnosis.

[0047] Through modeling of the spatial relationship between battery blocks and processing of time sequence information, the present application can accurately identify the fault state of the battery under different working conditions, making the battery fault diagnosis more reliable.

[0048] The present application designs a confidence evaluation mechanism to evaluate the reliability of the battery fault diagnosis result. BRIEF DESCRIPTION OF DRAWINGS

[0049] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrations, together with their description, serve to explain the application without unduly limiting it.

[0050] Figure 1 The overall technical flowchart of the embodiment of the application;

[0051] Figure 2 The schematic diagram of the graph construction method based on hardware electrical connection of the embodiment of the application;

[0052] Figure 3 The schematic diagram of the graph representation and feature embedding operation of the embodiment of the application;

[0053] Figure 4 The structure block diagram of the spatio-temporal feature extraction module of the embodiment of the application;

[0054] Figure 5 The structure block diagram of the electric vehicle battery fault diagnosis model of the embodiment of the application. DETAILED DESCRIPTION

[0055] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0056] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0057] Embodiment one

[0058] As shown in the figure, the embodiment provides a battery fault diagnosis method based on graph structure and LSTM, which comprises: Figure 1

[0059] Obtaining the time sequence feature information of each battery block in the battery module of the electric vehicle and the state of the battery pack based on the battery management system, and constructing a time sequence data set based on the time sequence feature information of each battery block and the state of the battery pack;

[0060] Constructing a graph structure data based on the hardware connection relationship of the battery blocks in the battery module;

[0061] Constructing a spatio-temporal feature extraction module of battery block information based on the graph structure data and the long short-term memory network model;

[0062] ​A computing model is constructed based on the spatio-temporal feature extraction module and the multi-head attention mechanism, the computing model is trained through the time series data set, and an electric vehicle battery fault diagnosis model is obtained;

[0063] The electric vehicle battery fault diagnosis model is deployed on an electric vehicle BMS platform to diagnose the fault of the battery.

[0064] Specifically includes:

[0065] S1, a time series data set for battery fault diagnosis is constructed, the time series feature information (including battery voltage, current, temperature, battery state SOC and SOH, internal resistance, and charging period) of each battery block in the battery management system (BMS) of the electric vehicle battery module is collected, and the state of the battery pack (including battery normal and overcharge, overdischarge, short circuit, capacity recession and other faults) is constructed together to construct a data set;

[0066] The target of the embodiment is to diagnose the state of the whole battery pack in real time based on the feature information of each battery block in the electric vehicle battery module. To achieve this goal, it is necessary to first collect the time series data set for training the battery fault diagnosis model. Since the battery management system (BMS) can continuously provide feature measurement data of the electric vehicle battery, the embodiment relies on the BMS to construct the time series data set. The specific steps include:

[0067] S1-1, the feature information of each battery block in the battery module is pre-defined, including battery voltage V, current I, temperature T, battery state SOC and SOH, R, and charging period C, that is, the once sampling feature information x of the battery block is:

[0068] x = [V, I, T, SOC, SOH, R, C];

[0069] 12 times are collected at a time interval of 5s, and a group of time series data x of the i-th battery block is obtained i :

[0070]

[0071] Wherein, represents the feature information (including voltage V, current I, temperature T, battery state SOC and SOH, internal resistance R, and charging period C) of the i-th battery block at j time point; and the total number of battery blocks in the battery module is N, then i∈[1,N], j∈[1,12];

[0072] S1-2, the data collection operation in S1-1 is performed for each battery block, and a group of complete battery pack input feature data X is obtained m :

[0073] X m= [x1, x2,..., x N ] ;

[0074] wherein x N represents the timing feature data of the Nth battery block;

[0075] S1-3, labeling the battery pack state Y m based on the input feature data X, including battery normal and fault categories such as overcharge, overdischarge, short circuit, capacity decay, etc., and inputting the battery pack feature data X m and the battery pack state Y m together constitute a set of data sets:

[0076] data m = [X m , Y m ] ;

[0077] S1-4, repeating steps S1-1 to S1-3, a total of M sets of battery fault diagnosis timing data sets D:

[0078] D = [data1, data2,..., data M ] ;

[0079] wherein any group of data data m contains the feature information of the battery pack and the corresponding battery state, and m e [1, M]; and divide the training set and the test set, the training set is used as the basis for subsequent model training, and the test set is used for model confidence evaluation and fine-tuning.

[0080] S2, design a graph structure representation method, construct a graph representation through the hardware connection relationship between the battery blocks in the battery module, and characterize the spatial characteristics of the battery module through the graph data, so as to obtain specific spatial information for different battery connection methods;

[0081] The embodiment is based on the hardware connection relationship between the battery blocks in the battery module, and captures the spatial characteristics of the battery module by constructing a graph representation method. The battery blocks in the battery module are usually connected in series and parallel, so in the fault diagnosis, the embodiment designs a graph structure representation method of the battery block based on this characteristic, so that the battery fault diagnosis model can use the spatial relationship between the battery blocks to improve the accuracy of fault diagnosis. The specific steps include:

[0082] S2-1, construct a graph structure, as the battery modules usually adopt series and parallel connection to balance the voltage and capacity requirements of the battery; among them, series connection increases the total voltage of the battery pack, while parallel connection increases the capacity and current carrying capacity of the battery pack. Therefore, based on this principle, each battery block in the battery module is regarded as a node (Node) in the graph, and the hardware connection relationship between the battery blocks (including series and parallel connection) is regarded as the edge (Edge) between the nodes, as shown in Figure 2 . In this way, the spatial topological relationship of the battery module is structured into a spatial information graph data structure through the graph structure.

[0083] S2-2, design the node feature representation in the graph, through the graph structure representation method of S2-1, each battery block can be embedded into the overall graph structure, so the feature information of each battery block at a single moment is regarded as the feature of the corresponding node, and the design of the node feature representation in the graph is realized; in addition, since the time series data of the battery pack collected by S1 contains feature information of 12 time points, each group of battery pack input feature data X m corresponds to 12 time point graph data:

[0084] G m =[G1,G2,...,G 12 ]=Graph(X m );

[0085] Among them, G m represents the graph data corresponding to the input feature data X m embedded based on the graph structure representation method, including 12 time point graph data; Graph(·) is the graph representation and feature embedding operation of this embodiment, and the process is as shown in Figure 3 .

[0086] S2-3, based on the steps of S2-1 and S2-2, embed all the input feature data collected in S1 into graph data, and form a group of data with the corresponding battery state category.

[0087] S3, based on the graph structure data and LSTM, construct a spatio-temporal feature extraction module of battery block information, which can fuse the spatio-temporal information of battery feature data to provide global embedding features with stronger expression ability;

[0088] This embodiment constructs a spatio-temporal feature extraction module based on graph structure data and LSTM (Long Short Term Memory Network) to extract the spatio-temporal features of the battery block information. This module aims to fuse the spatio-temporal information of battery feature data, so as to provide stronger expression ability and generate global embedding features, so that the subsequent model can accurately predict the overall failure state of the battery pack. The specific steps include:

[0089] S3-1 designs the basic structure of the spatio-temporal feature extraction module, extracts the spatio-temporal features of the battery block through the combination of the graph neural network (GNN) and the LSTM, and the module structure is as shown in Figure 4 The number of graph convolution layers is consistent with the number of sampling time points, and if the number of sampling time points is changed, the structure of the spatio-temporal feature extraction module can also be flexibly adjusted to adapt to different situations.

[0090] The graph convolution layer is used to capture the mutual influence relationship between the battery blocks in the battery module. The feature information of each battery block is propagated through the graph convolution layer, and in this way, the graph convolution can effectively represent the spatial information of the battery blocks in the battery module; and the battery block feature representation at the current time is fused into the graph convolution layer at the subsequent time through the LSTM layer, so as to ensure that the model pays attention to the influence relationship in the time dimension, and finally the global embedding feature at each time point is obtained.

[0091] S3-2 performs spatio-temporal feature extraction on the graph data G m containing 12 time points; first, the adjacency matrix A corresponding to the graph data G1 is obtained, and then the global embedding feature Emb1 obtained through the graph convolution layer is:

[0092]

[0093] wherein W1 represents the learnable weight matrix of the first graph convolution layer; σ(·) is an activation function, and ReLU function is adopted; G1 is the graph data, A is the adjacency matrix corresponding to the graph data, is the normalized adjacency matrix; D is a degree matrix, and D is a diagonal matrix, and the value of the diagonal line element is the number of adjacent nodes of the corresponding node; then the time feature is extracted through the LSTM layer:

[0094] H1=LSTM(Emb1,Q1);

[0095] wherein Q1 represents the learnable weight matrix of the first LSTM layer, H1 represents the intermediate feature after the time feature embedding, and LSTM(·) represents the LSTM layer processing function;

[0096] For the second graph convolution layer, the input is the graph data G2 at the second time point and the output feature of the first LSTM layer, and the process is:

[0097]

[0098] wherein W2 represents the learnable weight matrix of the second graph convolution layer, and Full(·) represents the full connection layer processing function, which aims to improve the time feature expression ability and standardize the data length of H1;

[0099] Therefore, based on the flow, all global embedding features with fusion time information can be obtained, i.e., Emb1, Emb2,..., Emb 12 .

[0100] S4, based on the spatio-temporal feature extraction module constructed in S3, a complete electric vehicle battery fault diagnosis model is designed in combination with a multi-head attention mechanism. The model takes the time series feature information of each battery block in the battery module as input, and predicts the state type of the entire battery pack;

[0101] The battery fault diagnosis model of the embodiment is based on the spatio-temporal feature extraction module in S3, and a complete electric vehicle battery fault diagnosis model is constructed in combination with a multi-head attention mechanism (Multi-Head Attention). The core task of the model is to use the time series feature information of each battery block in the battery module to predict and identify the fault state of the entire battery pack. The overall model structure framework is as shown in Figure 5 . Specifically, it includes:

[0102] S4-1, the model first processes the time series input feature data X m to obtain the graph data G m through the feature embedding method based on the graph structure representation designed in S2; then the global embedding features Emb containing spatio-temporal information Emb1, Emb2,..., Emb 12 are obtained through the spatio-temporal feature extraction module designed in S3;

[0103] S4-2, then the multi-head attention mechanism is used to capture the time series dependence and spatial relationship in the battery block spatio-temporal feature data in parallel, to model the global state of the battery pack from multiple angles. And through the weighted fusion of the relationship between different time points and battery blocks, the expression ability and accuracy of the battery fault diagnosis model are improved.

[0104] Specifically, it includes:

[0105] The multi-head Query, Key and Value matrices are constructed by linear transformation of Emb:

[0106]

[0107] Where Q l , K l and V l represent the Query, Key and Value matrices of the lth attention head, respectively; and represent the weight matrix of the lth attention head; there are L attention heads, so l∈[1,L]; then the weighted sum of the attention output of each attention head:

[0108]

[0109] where d is a normalization factor; and finally calculate the final multi-head attention output MultiHd:

[0110] MultiHd = Concat(Hd1, Hd2,..., Hd L )·W O ;

[0111] where Concat(·) is a channel concatenation operation, and W O is a linear transformation weight matrix;

[0112] S4-3, extract the complex spatio-temporal relationship between battery blocks and reduce the risk of gradient vanishing through full connection layer and residual connection, so as to improve the accuracy and training efficiency of battery fault diagnosis; that is:

[0113] FM = Concat(Full(MultiHd), MultiHd);

[0114] where Full(·) represents a full connection layer processing function; FM represents the feature data processed by the full connection layer and the residual connection; and the output feature DF is obtained by preventing model training overfitting through the Dropout layer:

[0115] DF = Dropout(FM);

[0116] where Dropout(·) represents a Dropout layer function, and the Dropout rate is set to 0.3.

[0117] S4-4, finally, use the Softmax layer as the output layer, which is responsible for converting the final output feature DF of the model into probability values that can be classified and decided, that is:

[0118] Y' m = Softmax(FM);

[0119] where Y' m is the final prediction result vector; and the class with the highest corresponding probability is the final result of the model fault diagnosis.

[0120] S5, design the loss function and training method of the battery fault diagnosis model, and design the fault diagnosis confidence evaluation method to evaluate the reliability of the diagnosis result; finally, deploy the trained battery fault diagnosis model to the electric vehicle BMS platform to realize real-time battery fault diagnosis and confidence evaluation.

[0121] In order to realize effective training of the model, the loss function of battery fault diagnosis is designed in the embodiment; and the confidence evaluation method is designed for the diagnosis result of the battery fault diagnosis model, so that the user can have more accurate judgment according to the output result of the battery fault diagnosis model; the specific steps include:

[0122] S5-1, design the loss function, since the battery fault diagnosis is a multi-class classification problem, the cross-entropy loss function (Cross-Entropy Loss) is selected to calculate the difference between the model prediction and the true label, that is:

[0123]

[0124] Wherein, Y m and Y' m represent the real class label and the model output result vector respectively; L m represents the loss function calculation result of the mth group of data, and cls is the total number of classes;

[0125] S5-2, the Adam optimizer is used to optimize the cross-entropy loss function to minimize the difference between the prediction result and the true label, and the network parameters are updated through back propagation. The performance of the model is optimized through multiple iterations until the loss function tends to be minimum and the model converges, and finally the trained battery fault diagnosis model is obtained;

[0126] S5-3, design the confidence evaluation method of the output result of the battery fault diagnosis model, and use the test set to diagnose multiple times to obtain its diagnosis result {(x1,y'1),(x2,y'2),...,(x n ,y' n )}, wherein x i is the input feature information, y' i is the model output result, the corresponding label is y i , and the accuracy of each diagnosis type is calculated:

[0127]

[0128] Wherein, Acc c represents the diagnosis accuracy of class c, and the total number of classes is cls, that is: c∈[1, cls]; Funz(·) is an indicator function, which indicates whether the output is consistent with the actual label;

[0129] Then, the mean μ c and variance of all prediction probabilities under the fault type c are calculated to reflect the confidence distribution:

[0130]

[0131] where P(y i = c|x i ) represents the probability of the output result being the class c under the premise of the input being x i , n c is the number of samples of fault type c; then a confidence adjustment strategy is established, and the adjusted confidence score AdCon is obtained:

[0132] AdCon = a P(y i = c|x i ) + (1-a) m c ;

[0133] where a is an adjustment coefficient, the value of which is dynamically adjusted according to the output accuracy of different classes, that is, a is increased when the accuracy is high, and vice versa;

[0134] Finally, for each class, the confidence threshold Th c is dynamically adjusted:

[0135] Th c = m c + k s c ;

[0136] where k is a constant that controls the looseness of the threshold; thus, for the output result of the battery fault diagnosis model, the confidence score AdCon of the result is calculated to reflect the reliability of the diagnosis result this time, and compared with the threshold Th c of the class, if AdCon is greater than Th c , the diagnosis result and the corresponding confidence score are returned, if AdCon is less than Th c , “unable to judge” is output, and further confirmation can be made through manual intervention.

[0137] S5-4, finally, the trained battery fault diagnosis model is deployed on the electric vehicle BMS platform to realize real-time battery fault diagnosis and confidence evaluation, the specific steps including:

[0138] (1) based on the BMS, real-time data information of each battery block in the battery module of the electric vehicle is collected, and the time series data Text in is obtained by using the method of S1;

[0139] (2) using the graph structure representation method in S2, the feature data of each battery block is represented through graph structure, the corresponding graph data G in is obtained, and input into the battery fault diagnosis model, and finally the diagnosis result is output;

[0140] (3) According to the confidence evaluation method in S5-3, the confidence of the battery fault diagnosis result is evaluated, if the confidence of the diagnosis result is higher than the preset threshold, the fault type and its confidence score are output; if it is lower than the threshold, it prompts "unable to judge" and can request manual intervention confirmation, to ensure the accuracy and safety of the system.

[0141] (4) Repeat steps (1), (2), (3) to realize real-time monitoring and fault prediction of battery state information, to ensure that the electric vehicle can continuously monitor the battery pack during driving, and timely issue a warning when an abnormality occurs, to ensure the safety and reliability of the battery pack.

[0142] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A battery fault diagnosis method based on graph structure and LSTM, characterized in that, include: The timing characteristics of each battery cell in the battery module of an electric vehicle and the state of the battery pack are obtained based on the battery management system, and a timing dataset is constructed based on the timing characteristics of each battery cell and the state of the battery pack. Graph structure data is constructed based on the hardware connection relationship of battery blocks within the battery module; A spatiotemporal feature extraction module for battery block information is constructed based on the graph structure data and the long short-term memory network model. A computational model is constructed based on the spatiotemporal feature extraction module and the multi-head attention mechanism. The computational model is trained using the time-series dataset to obtain an electric vehicle battery fault diagnosis model. The electric vehicle battery fault diagnosis model is deployed on the electric vehicle BMS platform to diagnose battery faults. The process of constructing a spatiotemporal feature extraction module for battery block information based on the graph structure data and long short-term memory network model includes: A spatiotemporal feature extraction module for battery block information is constructed based on graph neural network and long short-term memory network models; The graph neural network extracts spatiotemporal features from graph structure data to obtain global embedding features; The Long Short-Term Memory Network extracts temporal features from the global embedded features to obtain intermediate features after temporal feature embedding. In the graph neural network, the feature extraction expression for the first graph convolutional layer is: ； in, This represents the learnable weight matrix of the first graph convolutional layer; For activation function, For graph data, This is the adjacency matrix corresponding to the graph data. This is the standardized adjacency matrix; Let be a degree matrix, and This is a diagonal matrix where the values ​​of the diagonal elements are the number of adjacent nodes for the corresponding node. The global embedding features are obtained after graph convolutional layers; The expression for the intermediate features obtained after embedding the time features is: ； in, This represents the learnable weight matrix of the first LSTM layer. This represents the intermediate features after embedding the temporal features. This represents the LSTM layer processing function; It also includes the following: In the graph neural network, the feature extraction expression for the second graph convolutional layer is: ； in, This represents the learnable weight matrix of the second graph convolutional layer. This represents the processing function of the fully connected layer; The process of diagnosing battery faults based on the electric vehicle battery fault diagnosis model includes: Based on the spatiotemporal feature extraction module, global embedded features containing spatiotemporal information are obtained; Based on the multi-head attention mechanism, the temporal dependencies and spatial relationships in the global embedded features of the spatiotemporal information are captured to obtain the attention output; The attention outputs are weighted and fused to obtain the final multi-head attention output; The final multi-head attention output is calculated using fully connected layers and residual connections to obtain processed feature data; The processed feature data is calculated using the Dropout layer to obtain the output features; The Softmax layer is used as the output layer to transform the output features into probability values ​​for classification decisions, thereby obtaining the final prediction result.

2. The method according to claim 1, characterized in that, The timing characteristics of each battery cell include battery voltage, current, temperature, battery health status, battery charging status, internal resistance, and charging cycle.

3. The method according to claim 1, characterized in that, The process of constructing a time-series dataset based on the time-series feature information of each battery block and the state of the battery pack includes: Predefine the feature information of each battery block in the battery module, collect the feature information of each battery block, and obtain complete battery pack input feature data; The complete battery pack input feature data is labeled with battery pack status, and the complete battery pack input feature data and the labeled battery pack status are integrated to obtain a time series dataset for battery fault diagnosis.

4. The method according to claim 1, characterized in that, The process of constructing graph structure data based on the hardware connection relationship of battery blocks within the battery module includes: Each battery block is treated as a node in the graph, and the connections between battery blocks are treated as edges between nodes to construct a graph data structure that includes spatial information. The feature information of each battery block at a single moment is used as the feature of the corresponding node to construct graph structure data; Each battery pack should receive graphical data at 12 time points as input feature data, expressed as follows: ； in, This represents the graph data corresponding to the input feature data, obtained by embedding based on graph structure representation methods. For input feature data, The graph data is for the first time point. For the graph data at the second time point, The graph data is for the twelfth time point. This refers to graph representation and feature embedding operations.

5. The method according to claim 1, characterized in that, The process of training the computational model using the time-series dataset to obtain an electric vehicle battery fault diagnosis model includes: The Adam optimizer is used to optimize the cross-entropy loss function, and the computational model is iteratively optimized multiple times by updating the network parameters through backpropagation until the cross-entropy loss function tends to a minimum and the model converges, and then the model is output to obtain the electric vehicle battery fault diagnosis model.

6. The method according to claim 1, characterized in that, After obtaining the final prediction result, the process also includes: A confidence evaluation method for the output results of a battery fault diagnosis model is constructed. The final prediction results are evaluated based on the confidence evaluation method to obtain the accuracy of each diagnosis type. Based on the accuracy of each diagnostic type, calculate the mean and variance of all predicted probabilities under the fault type, and obtain the confidence score and confidence threshold based on the mean and variance. If the confidence score is greater than the confidence threshold, the diagnosis result and the corresponding confidence score are output; if the confidence score is less than the confidence threshold, manual intervention is required.

Images