Lithium battery life prediction method and system based on graph neural network

Abstract

The invention discloses a lithium battery life prediction method and system based on a graph neural network, and the method comprises the steps: obtaining the charge-discharge cycle data of a plurality of historical periods of a lithium battery, and extracting the capacity data of the battery according to the charge-discharge cycle data; segmenting the battery capacity data to obtain a plurality of sequence samples corresponding to the time sequence, and dividing the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model; self-adaptively adding white noise into the battery capacity data, and amplifying the non-correlation degree among all modes so as to decompose the modes of the battery capacity data; and constructing a graph neural network, combining the adjacent matrix, graph convolution and an attention mechanism to obtain a lithium battery life prediction model, and predicting the residual life of the lithium battery through the lithium battery life prediction model. The problems that the lithium battery life prediction precision is insufficient and the prediction precision improvement is limited in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery life prediction, and in particular to a lithium battery life prediction method and system based on graph neural network. Background Art

[0002] Although the existing battery life prediction technology has improved the prediction accuracy to a certain extent, it still has some shortcomings. First, many methods do not adequately model the complex dependencies between modal components and cannot fully explore the synergy between the components. Secondly, the feature fusion and information integration methods are relatively simple and cannot effectively deal with the problem of feature dispersion and noise. Finally, some methods rely on artificially set parameters or rules, lack adaptability, and are difficult to adapt to the changes and complexity of different data.

[0003] Existing methods for predicting the remaining life of lithium batteries usually use modal decomposition technology, such as CEEMDAN, to decompose battery capacity data into multiple modal components to extract different feature information. Although existing methods have improved the prediction accuracy to a certain extent, there are still several key problems. First, many methods do not adequately model the complex dependencies between modal components and cannot fully explore the synergy between components, resulting in limited improvement in prediction accuracy. Secondly, the feature fusion and information integration methods are relatively single, and they fail to effectively deal with the problem of feature dispersion, affecting the comprehensive understanding of the battery degradation process. Finally, some methods rely on manually set parameters or rules, lack adaptability, and cannot flexibly adapt to the changes and complexity of different data sets.

[0004] Although the existing battery life prediction technology has improved the prediction accuracy to a certain extent, it still has some shortcomings. First, many methods do not adequately model the complex dependencies between modal components and cannot fully explore the synergy between the components. Secondly, the feature fusion and information integration methods are relatively simple and cannot effectively deal with the problem of feature dispersion and noise. Finally, some methods rely on artificially set parameters or rules, lack adaptability, and are difficult to adapt to the changes and complexity of different data. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a lithium battery life prediction method and system based on graph neural network, aiming to solve the problem of insufficient accuracy of lithium battery life prediction and limited improvement of prediction accuracy in the prior art.

[0006] A first aspect of the present invention is to provide a lithium battery life prediction method based on a graph neural network, the method comprising: Acquire charge and discharge cycle data of several historical periods of the lithium battery, and extract battery capacity data based on the charge and discharge cycle data; Segmenting the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and dividing the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model; Adaptively adding white noise to the battery capacity data to amplify the degree of non-correlation between various modes, so as to decompose the modes of the battery capacity data; A graph neural network is constructed, and the adjacency matrix, graph convolution and attention mechanism are combined to obtain a lithium battery life prediction model, and the remaining life of the lithium battery is predicted by the lithium battery life prediction model.

[0007] According to one aspect of the above technical solution, the battery capacity data is segmented to obtain a plurality of sequence samples corresponding to the time series, and the sequence samples are divided into a training set and a test set according to preset time nodes to train a preset neural network model, including: The continuous battery capacity data is divided into sliding windows of fixed length. Each fixed-length data segment is used as a sequence sample to capture the dynamic change characteristics in the time series. The target value of each sequence sample is a time step after the time series, which is used to predict the future capacity value. Dividing the plurality of sequence samples according to a preset ratio in chronological order to obtain a training set for training a neural network model and a test set for testing the neural network model; The neural network model is trained using the training set and the test set.

[0008] According to one aspect of the above technical solution, the step of adaptively adding white noise to the battery capacity data to amplify the degree of non-correlation between the modes to decompose the modes of the battery capacity data includes: Adding white noise to the original battery capacity data in a dispersed manner using an integrated empirical mode decomposition method based on adaptive white noise; The non-correlation degree between each mode in the battery capacity data is amplified by adding white noise, so as to perform modal decomposition on the capacity sequence in the battery capacity data.

[0009] According to one aspect of the above technical solution, a graph neural network is constructed, and the complex relationship between modal components is characterized by combining the adjacency matrix, graph convolution and attention mechanism, and a lithium battery life prediction model is obtained to predict the remaining life of the lithium battery through the lithium battery life prediction model, including: Obtaining the time series; Converting multiple modal components obtained based on capacity sequence decomposition into multiple graph nodes in a modal component graph, and using an adjacency matrix to represent the relationship between the multiple graph nodes in the modal component graph; Obtain the structure of the modal component graph at each moment, use the graph attention layer to capture the relationship between the modal components using the attention mechanism, calculate the attention weight between the modal component and its associated modal component, update the current modal component state according to the attention weight, so as to extract the node spatial features between multiple modal components, and aggregate the node spatial features by average pooling to obtain the global spatial features; According to the global spatial features, a gated recurrent unit is used to construct a feature matrix of the time series, and the input time series is modeled through a gating mechanism; The features processed by the graph attention layer and the gated recurrent unit are input into the fully connected layer to predict and output the remaining life of the lithium battery.

[0010] According to one aspect of the above technical solution, according to the global spatial features, a gated recurrent unit is used to construct a feature matrix of a time series, and in the step of modeling the input time series through a gating mechanism, the step of modeling the input time series through a gating mechanism using a gated recurrent unit includes: a. Input: Controls the global spatial characteristics of the current input Current status The influence of is expressed as: ; In the formula, represents the reset gate vector, represents the activation function, and Respectively represent the weight and bias of the gated recurrent unit; b. Update gate: control the previous step state and the global space characteristics of the current input The degree of fusion is expressed as: ; In the formula, To update the gate vector, To update the gate weights, To update the gate bias; c. Candidate state calculation: Generate the candidate state at the current moment , the expression is: ; In the formula, is the candidate state weight, Bias for candidate states; d. State update: Combined update gate vector With candidate status , calculate the final state at the current moment , the expression is: ; e. Every moment , the gated recurrent unit will output a hidden state vector, namely , represents the current moment feature, the last hidden state of the gated recurrent unit It is used as the final feature representation for lithium battery life prediction.

[0011] According to one aspect of the above technical solution, the features processed by the graph attention layer and the gated recurrent unit are input into the fully connected layer, and the remaining life of the lithium battery is predicted and output, and the expression is: ; In the formula, is the predicted value of the remaining life of the lithium battery, is the weight of the fully connected layer, is the bias of the fully connected layer.

[0012] According to one aspect of the above technical solution, the method further includes: The mean absolute percentage error is used as an evaluation index to evaluate the remaining life prediction value output by the lithium battery life prediction model. The expression is: ; In the formula, is the actual value of the remaining life of the lithium battery, is the predicted value of the remaining life of the lithium battery, is the sample size.

[0013] A second aspect of the present invention is to provide a lithium battery life prediction system based on a graph neural network, which is applied to the method described in the above technical solution, and the system comprises: A data extraction module, used to obtain charge and discharge cycle data of several historical periods of the lithium battery, and extract battery capacity data based on the charge and discharge cycle data; A sample division module, used to divide the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and divide the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model; A modal decomposition module, used for adaptively adding white noise to the battery capacity data, amplifying the degree of non-correlation between various modes, so as to decompose the modes of the battery capacity data; The life prediction module is used to construct a graph neural network and combine the adjacency matrix, graph convolution and attention mechanism to obtain a lithium battery life prediction model, and the remaining life of the lithium battery is predicted by the lithium battery life prediction model.

[0014] A third aspect of the present invention is to provide a readable storage medium having a computer program stored thereon, which implements the method described in the above technical solution when executed by a processor.

[0015] The fourth aspect of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the above technical solution when executing the computer program.

[0016] Compared with the prior art, the lithium battery life prediction method and system based on graph neural network shown in the present invention has the following beneficial effects: The method shown in the present invention abandons the reliance on artificially set parameters and rules, and adopts adaptive learning and feature integration mechanisms, so that the lithium battery life prediction model can flexibly respond to the changes and complexity of different data sets, and improve its stability and prediction accuracy in a variety of complex situations. This fusion of spatiotemporal features not only enhances the synergy between modal components, but also significantly improves the accuracy and effectiveness of the prediction results, providing an efficient and highly adaptable solution for the accurate prediction of the remaining life of lithium batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 A schematic diagram of a flow chart of a lithium battery life prediction method based on a graph neural network provided in an embodiment of the present invention; Figure 2 A structural block diagram of a lithium battery life prediction system based on a graph neural network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. Several embodiments of the present invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0020] Embodiment 1 See also Figure 1 The first embodiment of the present invention provides a lithium battery life prediction method based on a graph neural network. The method shown in this embodiment includes steps S10 to S40: Step S10, acquiring charge and discharge cycle data of several historical periods of the lithium battery, and extracting battery capacity data according to the charge and discharge cycle data.

[0021] In this embodiment, the original discharge cycle data of several historical cycles of the lithium battery are obtained, specifically according to the historical cycle charge and discharge data of the lithium battery at 24°C provided by the lithium battery NASA data set, and the capacity index is extracted based on the discharge process as the battery health life feature.

[0022] Step S20, segmenting the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and dividing the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model.

[0023] In this embodiment, the battery capacity data is segmented to obtain a plurality of sequence samples corresponding to the time series, and the sequence samples are divided into a training set and a test set according to preset time nodes to train a preset neural network model, including: The continuous battery capacity data is divided into sliding windows of fixed length. Each fixed-length data segment is used as a sequence sample to capture the dynamic change characteristics in the time series. The target value of each sequence sample is a time step after the time series, which is used to predict the future capacity value. Dividing the plurality of sequence samples according to a preset ratio in chronological order to obtain a training set for training a neural network model and a test set for testing the neural network model; The neural network model is trained using the training set and the test set.

[0024] Specifically, step S20 mainly pre-processes the data, including: (1) Sequence construction: The continuous battery capacity data is divided into sliding windows of fixed length T. Each segment of data with a length of T is used as a sequence sample to capture the dynamic change characteristics in the time series. The target value of each sample is a time step after the sequence, which is used to predict the future capacity value; (2) Training set and test set division: The data is divided into two parts in chronological order. The first 80% of the data is used for training data, and the last 20% is used to test the performance of the model, ensuring the continuity between the training set and the test set. The neural network model is trained using the divided training set and test set.

[0025] Step S30: adaptively adding white noise to the battery capacity data to amplify the degree of non-correlation between the modes, so as to decompose the modes of the battery capacity data.

[0026] In this embodiment, the step of adaptively adding white noise to the battery capacity data to amplify the degree of non-correlation between the modes to decompose the modes of the battery capacity data includes: Adding white noise to the original battery capacity data in a dispersed manner using an integrated empirical mode decomposition method based on adaptive white noise; The non-correlation degree between each mode in the battery capacity data is amplified by adding white noise, so as to perform modal decomposition on the capacity sequence in the battery capacity data.

[0027] Specifically, the method of integrated empirical mode decomposition based on adaptive white noise is used to decompose the battery capacity data, which is often used to process non-stationary and nonlinear data. It adds white noise to the original data, which can amplify the degree of irrelevance between the modes and facilitate the mode decomposition of the original data.

[0028] Step S40, constructing a graph neural network, and combining the adjacency matrix, graph convolution and attention mechanism to obtain a lithium battery life prediction model, and predicting the remaining life of the lithium battery through the lithium battery life prediction model.

[0029] In this embodiment, a graph neural network is constructed, and the complex relationship between modal components is characterized by combining the adjacency matrix, graph convolution and attention mechanism, and a lithium battery life prediction model is obtained to predict the remaining life of the lithium battery through the lithium battery life prediction model, including: Obtaining the time series; Converting multiple modal components obtained based on capacity sequence decomposition into multiple graph nodes in a modal component graph, and using an adjacency matrix to represent the relationship between the multiple graph nodes in the modal component graph; Obtain the structure of the modal component graph at each moment, use the graph attention layer to capture the relationship between the modal components using the attention mechanism, calculate the attention weight between the modal component and its associated modal component, update the current modal component state according to the attention weight, so as to extract the node spatial features between multiple modal components, and aggregate the node spatial features by average pooling to obtain the global spatial features; According to the global spatial features, a gated recurrent unit is used to construct a feature matrix of the time series, and the input time series is modeled through a gating mechanism; The features processed by the graph attention layer and the gated recurrent unit are input into the fully connected layer to predict and output the remaining life of the lithium battery.

[0030] Specifically, the core innovation of this embodiment is to introduce a dynamic graph neural network to construct dependencies between modal components, effectively integrate scattered information into feature representations with higher expressive power, and significantly enhance the ability of the lithium battery life prediction model to capture the laws of battery degradation.

[0031] The lithium battery life prediction model achieves the prediction task by combining the two dimensions of time and space. The graph attention layer processes the spatial relationship of each time step to capture the dependency between modal components. At the same time, the gated recurrent unit is used to model the time series features, and finally the integration and prediction of spatiotemporal features are achieved. The specific steps are as follows: After modal decomposition of the capacity sequence, M modal component sequences are obtained: As features of graph nodes.

[0032] 1. Input time series , , where T is the size of the time window.

[0033] 2. Build graph structure (1) Convert the modal components into nodes in the graph Every moment Construct the modal component diagram, and the node feature matrix is: ; in, .

[0034] (2) Constructing the adjacency matrix The adjacency matrix is ​​used to represent the relationship between nodes in a graph. In battery life prediction, the Pearson correlation coefficient between modal components is calculated as the element of the adjacency matrix, which represents the similarity and dependency between different modal components.

[0035] M modal components , each component has TThe values ​​at each time point form a matrix A : ; in, Represents the modal component i At the moment t The correlation coefficient was then obtained using the Pearson formula.

[0036] 3. Graph Attention Layer (GAT) Enter each moment t The modal component diagram structure of ,GAT uses the attention mechanism to capture the relationship between each modal component. By calculating the attention weight between a modal component and its associated modal component, the current modal component state is updated according to the weight, and then the spatial features between multiple modal components are extracted.

[0037] The formula for updating node features is: ; in, yes l The node feature matrix of the layer, is a normalized adjacency matrix, which represents the connection relationship between nodes in the graph. is the weight matrix of the current layer, is an activation function, such as Activation function. The node features are then aggregated using average pooling to obtain a global feature representation, providing a unified input for GRU.

[0038] ; in, N is the number of nodes, It is i The characteristics of a node.

[0039] 4. GRU module (Gated Recurrent Unit) GRU has a simpler structure and fewer parameters, which enables it to converge quickly when processing sequences. It can quickly learn short-term patterns and trends in time series, and is particularly sensitive to short-term fluctuations that change frequently, making it very suitable for capturing battery regeneration in a short period of time.

[0040] (1) Every moment t The graph passes through the GAT module and outputs a global feature vector , the vector can be expressed as: ; in, Nrepresents the number of modal components, It is t Moment i The value of the node feature after being updated and pooled by GAT.

[0041] (2) Time series construction Every moment t The global eigenvector of Considered as an input feature of the time series, for the input of T time steps, a feature matrix can be constructed: ; in, , T is the number of time steps, d is the dimension of the feature vector.

[0042] (3) GRU models the input time series through a gating mechanism. The following are the specific steps of GRU: a. Input: Controls the global spatial characteristics of the current input Current status The influence of is expressed as: ; In the formula, represents the reset gate vector, represents the activation function, and Respectively represent the weight and bias of the gated recurrent unit; b. Update gate: control the previous step state and the global space characteristics of the current input The degree of fusion is expressed as: ; In the formula, To update the gate vector, To update the gate weights, To update the gate bias; c. Candidate state calculation: Generate the candidate state at the current moment , the expression is: ; In the formula, is the candidate state weight, Bias for candidate states; d. State update: Combined update gate vector With candidate status , calculate the final state at the current moment , the expression is: ; e. Every moment , the gated recurrent unit will output a hidden state vector, namely , represents the current moment feature, the last hidden state of the gated recurrent unit It is used as the final feature representation for lithium battery life prediction.

[0043] In this embodiment, the features processed by the graph attention layer and the gated recurrent unit are input into the fully connected layer to predict and output the remaining life of the lithium battery. The expression is: ; In the formula, is the predicted value of the remaining life of the lithium battery, is the weight of the fully connected layer, is the bias of the fully connected layer.

[0044] In addition, in this embodiment, the method further includes: The mean absolute percentage error is used as an evaluation index to evaluate the remaining life prediction value output by the lithium battery life prediction model. The expression is: ; In the formula, is the actual value of the remaining life of the lithium battery, is the predicted value of the remaining life of the lithium battery, is the sample size.

[0045] The present invention successfully captures the spatiotemporal characteristics in the prediction of the remaining life of lithium batteries by combining the graph attention layer and the GRU unit, thereby significantly improving the accuracy and robustness of the prediction. The graph attention layer can accurately capture the dependencies between modal components by adaptively processing the spatial relationship of each time step, solving the problems of scattered modes and inaccurate capture that occur in traditional methods when dealing with capacity regeneration phenomena. In addition, the graph attention layer also enhances the influence of important features by adjusting the weights of features, while reducing the interference of false components on the prediction results, effectively reflecting the changes in the battery state. Finally, the gated recurrent unit effectively captures the changing trend of battery life in different time periods by learning the dynamic laws of time series, further enhancing the understanding of the capacity regeneration phenomenon.

[0046] Compared with the prior art, the method shown in this embodiment abandons the reliance on artificially set parameters and rules, and adopts adaptive learning and feature integration mechanisms, so that the lithium battery life prediction model can flexibly respond to the changes and complexity of different data sets, and improve its stability and prediction accuracy in a variety of complex situations. This fusion of spatiotemporal features not only enhances the synergy between modal components, but also significantly improves the accuracy and effectiveness of the prediction results, providing an efficient and highly adaptable solution for the accurate prediction of the remaining life of lithium batteries.

[0047] Embodiment 2 See also Figure 2 The second embodiment of the present invention provides a lithium battery life prediction system based on a graph neural network, which is applied to the method described in the above embodiment. The system includes: a data extraction module 10, a sample division module 20, a modal decomposition module 30 and a life prediction module 40.

[0048] A data extraction module 10 is used to obtain charge and discharge cycle data of a lithium battery for several historical periods, and extract battery capacity data based on the charge and discharge cycle data; A sample division module 20 is used to divide the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and divide the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model; A modal decomposition module 30, configured to adaptively add white noise to the battery capacity data, amplify the degree of non-correlation between the modes, and decompose the modes of the battery capacity data; The life prediction module 40 is used to construct a graph neural network, and combine the adjacency matrix, graph convolution and attention mechanism to obtain a lithium battery life prediction model, and predict the remaining life of the lithium battery through the lithium battery life prediction model.

[0049] Compared with the prior art, the system shown in this embodiment abandons the reliance on artificially set parameters and rules, and adopts adaptive learning and feature integration mechanisms, so that the lithium battery life prediction model can flexibly respond to the changes and complexity of different data sets, and improve its stability and prediction accuracy in a variety of complex situations. This fusion of spatiotemporal features not only enhances the synergy between modal components, but also significantly improves the accuracy and effectiveness of the prediction results, providing an efficient and highly adaptable solution for the accurate prediction of the remaining life of lithium batteries.

[0050] Embodiment 3 A third embodiment of the present invention provides a readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method described in the above embodiment is implemented.

[0051] Embodiment 4 A fourth embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the above embodiments when executing the computer program.

[0052] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0053] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A lithium battery life prediction method based on graph neural network, characterized in that: The method comprises: Acquire charge and discharge cycle data of several historical periods of the lithium battery, and extract battery capacity data based on the charge and discharge cycle data; Segmenting the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and dividing the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model; Adaptively adding white noise to the battery capacity data to amplify the degree of non-correlation between various modes, so as to decompose the modes of the battery capacity data; A graph neural network is constructed, and the adjacency matrix, graph convolution and attention mechanism are combined to obtain a lithium battery life prediction model, and the remaining life of the lithium battery is predicted by the lithium battery life prediction model.

2. The lithium battery life prediction method based on graph neural network according to claim 1 is characterized in that: The step of segmenting the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and dividing the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model includes: The continuous battery capacity data is divided into sliding windows of fixed length. Each fixed-length data segment is used as a sequence sample to capture the dynamic change characteristics in the time series. The target value of each sequence sample is a time step after the time series, which is used to predict the future capacity value. Dividing the plurality of sequence samples according to a preset ratio in chronological order to obtain a training set for training a neural network model and a test set for testing the neural network model; The neural network model is trained using the training set and the test set.

3. The lithium battery life prediction method based on graph neural network according to claim 2 is characterized in that: The step of adaptively adding white noise to the battery capacity data to amplify the degree of non-correlation between the modes to decompose the modes of the battery capacity data includes: Adding white noise to the original battery capacity data in a dispersed manner using an integrated empirical mode decomposition method based on adaptive white noise; The non-correlation degree between each mode in the battery capacity data is amplified by adding white noise, so as to perform modal decomposition on the capacity sequence in the battery capacity data.

4. The lithium battery life prediction method based on graph neural network according to claim 1 is characterized in that: The steps of constructing a graph neural network, combining an adjacency matrix, graph convolution and an attention mechanism, characterizing the complex associations between modal components, and obtaining a lithium battery life prediction model to predict the remaining life of a lithium battery through the lithium battery life prediction model include: Obtaining the time series; Converting multiple modal components obtained based on capacity sequence decomposition into multiple graph nodes in a modal component graph, and using an adjacency matrix to represent the relationship between the multiple graph nodes in the modal component graph; Obtain the structure of the modal component graph at each moment, use the graph attention layer to capture the relationship between the modal components using the attention mechanism, calculate the attention weight between the modal component and its associated modal component, update the current modal component state according to the attention weight, so as to extract the node spatial features between multiple modal components, and aggregate the node spatial features by average pooling to obtain the global spatial features; According to the global spatial features, a gated recurrent unit is used to construct a feature matrix of the time series, and the input time series is modeled through a gating mechanism; The features processed by the graph attention layer and the gated recurrent unit are input into the fully connected layer to predict and output the remaining life of the lithium battery.

5. The lithium battery life prediction method based on graph neural network according to claim 4 is characterized in that: In the step of constructing a feature matrix of a time series using a gated recurrent unit according to the global spatial features, and modeling the input time series through a gating mechanism, the step of modeling the input time series through a gating mechanism using a gated recurrent unit includes: a. Input: Controls the global spatial characteristics of the current input Current status The influence of is expressed as: ； In the formula, represents the reset gate vector, represents the activation function, and Respectively represent the weight and bias of the gated recurrent unit; b. Update gate: control the previous step state and the global space characteristics of the current input The degree of fusion is expressed as: ； In the formula, To update the gate vector, To update the gate weights, To update the gate bias; c. Candidate state calculation: Generate the candidate state at the current moment , the expression is: ； In the formula, is the candidate state weight, Bias for candidate states; d. State update: Combined update gate vector With candidate status , calculate the final state at the current moment , the expression is: ； e. Every moment , the gated recurrent unit will output a hidden state vector, namely , represents the current moment feature, the last hidden state of the gated recurrent unit It is used as the final feature representation for lithium battery life prediction.

6. The lithium battery life prediction method based on graph neural network according to claim 5 is characterized in that: The features processed by the graph attention layer and the gated recurrent unit are input into the fully connected layer to predict and output the remaining life of the lithium battery. The expression is: ； In the formula, is the predicted value of the remaining life of the lithium battery, is the weight of the fully connected layer, is the bias for the fully connected layer.

7. The lithium battery life prediction method based on graph neural network according to any one of claims 1 to 6, characterized in that: The method further comprises: The mean absolute percentage error is used as an evaluation index to evaluate the remaining life prediction value output by the lithium battery life prediction model. The expression is: ； In the formula, is the actual value of the remaining life of the lithium battery, is the predicted value of the remaining life of the lithium battery, is the sample size.

8. A lithium battery life prediction system based on graph neural network, characterized in that: The method applied to any one of claims 1 to 7, wherein the system comprises: A data extraction module, used to obtain charge and discharge cycle data of several historical periods of the lithium battery, and extract battery capacity data based on the charge and discharge cycle data; A sample division module, used to divide the battery capacity data to obtain a plurality of sequence samples corresponding to the time series, and divide the sequence samples into a training set and a test set according to preset time nodes to train a preset neural network model; A modal decomposition module, used for adaptively adding white noise to the battery capacity data, amplifying the degree of non-correlation between various modes, so as to decompose the modes of the battery capacity data; The life prediction module is used to construct a graph neural network and combine the adjacency matrix, graph convolution and attention mechanism to obtain a lithium battery life prediction model, and the remaining life of the lithium battery is predicted by the lithium battery life prediction model.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images