Electrochemical repair method of lithium battery

Abstract

The present application discloses an electrochemical repair method for a lithium battery, which monitors the change of battery temperature in real time during the charging process of the lithium battery, and introduces a data processing and analysis algorithm at the back end to analyze the temperature time series change during the charging process, so as to take corresponding measures to control the temperature based on the temperature change, such as reducing the charging current or suspending charging. In this way, the charging current value can be adaptively adjusted based on the actual temperature situation, thereby reducing the polarization phenomenon inside the battery and improving the performance and safety of the battery.

Description

Technical Field

[0001] The present application relates to the field of lithium batteries, and more specifically, to an electrochemical repair method for lithium batteries. Background Art

[0002] Power batteries, depending on the positive electrode materials, mainly include two categories: ternary power batteries and iron-lithium power batteries, which use nickel-cobalt-manganese lithium (NCM) and lithium iron phosphate (LFP) as positive electrode materials respectively. From a performance perspective, iron-lithium power batteries have low capacity and short driving range, but good safety and long cycle life. After iron-lithium power batteries are retired, the market tends to enter the cascade utilization industry, which can give full play to their surplus value, maximize the circular economy, and reduce the construction cost of energy storage systems. Ternary power batteries have high specific capacity and long driving range, but short lifespan and inferior safety performance to iron-lithium power batteries. They are not suitable for cascade utilization fields with complex application environments such as energy storage power stations and backup power supplies for communication base stations.

[0003] As the use time of lithium-ion batteries increases, the capacity of lithium-ion batteries will gradually decrease. This capacity loss limits the service life and performance of the battery. However, retired lithium-ion batteries can be repaired. The capacity loss caused by kinetic polarization is a reversible capacity loss, which can be compensated and repaired through the charge and discharge system; the capacity loss caused by material structure fatigue and internal battery structure damage is an irreversible capacity loss, but it can be partially restored through appropriate physical repair. Through battery repair technology, the efficiency of retired power battery cascade utilization and product quality can be greatly improved.

[0004] The charge and discharge system is a commonly used method for repairing lithium batteries. By periodically charging and discharging the battery, the active substances inside the battery can be redistributed, thereby reducing polarization and increasing the battery capacity. During the charging process, temperature is one of the important factors affecting battery performance and safety. Too high a temperature will cause further damage to the battery, while too low a temperature may affect the battery's performance.

[0005] Therefore, an optimized electrochemical repair scheme for lithium batteries is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an electrochemical repair method for a lithium battery, which monitors the change of the battery temperature in real time during the charging process of the lithium battery, and introduces a data processing and analysis algorithm at the back end to analyze the temperature time series change during the charging process, so as to take corresponding measures to control the temperature based on the temperature change, such as reducing the charging current or suspending charging. In this way, the charging current value can be adaptively adjusted based on the actual temperature conditions, thereby reducing the polarization phenomenon inside the battery and improving the performance and safety of the battery.

[0007] According to one aspect of the present application, there is provided a method for electrochemical repair of a lithium battery, comprising:

[0008] Obtain temperature values ​​at multiple predetermined time points within a predetermined time period during the charging process;

[0009] Arranging the temperature values ​​of the plurality of predetermined time points according to the time dimension into a charging process temperature time series input vector;

[0010] Performing up-sampling based on linear interpolation on the charging process temperature timing input vector to obtain an up-sampled charging process temperature timing input vector;

[0011] Performing local time series feature extraction on the upsampled charging process temperature time series input vector to obtain a sequence of charging process temperature local time series feature vectors;

[0012] Performing local temperature time series feature consistency association coding on each charging process temperature local time series feature vector in the sequence of the charging process temperature local time series feature vectors to obtain a local temperature time series feature consistency topological feature matrix;

[0013] Performing graph-structure-based association coding on the sequence of the charging process temperature local time series feature vectors and the local temperature time series feature consistency topology feature matrix to obtain consistent topology charging process temperature local time series features;

[0014] Based on the local time series characteristics of the temperature in the consistent topology charging process, it is determined whether the charging current value at the current time point should be increased, decreased or maintained.

[0015] Compared with the prior art, the present application provides an electrochemical repair method for a lithium battery, which monitors the changes in battery temperature in real time during the charging process of the lithium battery, and introduces data processing and analysis algorithms at the back end to analyze the temperature timing changes during the charging process, so as to take corresponding measures to control the temperature based on the temperature changes, such as reducing the charging current or suspending charging. In this way, the charging current value can be adaptively adjusted based on the actual temperature conditions, thereby reducing the polarization phenomenon inside the battery and improving the performance and safety of the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 A flow chart of an electrochemical repair method for a lithium battery according to an embodiment of the present application;

[0018] Figure 2 A system architecture diagram of an electrochemical repair method for a lithium battery according to an embodiment of the present application;

[0019] Figure 3 This is a flow chart of sub-step S4 of the electrochemical repair method of a lithium battery according to an embodiment of the present application;

[0020] Figure 4 This is a flow chart of sub-step S5 of the electrochemical repair method of a lithium battery according to an embodiment of the present application;

[0021] Figure 5 This is a flow chart of sub-step S7 of the electrochemical repair method of a lithium battery according to an embodiment of the present application.

[0022] Figure 6 This is a diagram showing the effect of electrochemical repair according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0025] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0026] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] The charge-discharge system is a commonly used lithium battery repair method. By periodically charging and discharging the battery, the active materials inside the battery can be redistributed, thereby reducing polarization and increasing the battery capacity. During the charging process, temperature is one of the important factors affecting battery performance and safety. Too high a temperature will cause further damage to the battery, while too low a temperature may affect the battery performance. Therefore, an optimized electrochemical repair scheme for lithium batteries is desired.

[0029] In the technical solution of the present application, a method for electrochemical repair of a lithium battery is proposed. Figure 1 The present invention is a flowchart of an electrochemical repair method for a lithium battery according to an embodiment of the present application. Figure 2 1 is a system architecture diagram of the electrochemical repair method of a lithium battery according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the electrochemical repair method of the lithium battery includes the following steps: S1, obtaining the temperature values ​​of multiple predetermined time points within a predetermined time period during the charging process; S2, arranging the temperature values ​​of the multiple predetermined time points according to the time dimension as a charging process temperature timing input vector; S3, upsampling the charging process temperature timing input vector based on linear interpolation to obtain an upsampled charging process temperature timing input vector; S4, extracting local timing features of the upsampled charging process temperature timing input vector to obtain a sequence of charging process temperature local timing feature vectors; S5, performing local temperature timing feature consistency association coding on each charging process temperature local timing feature vector in the sequence of the charging process temperature local timing feature vectors to obtain a local temperature timing feature consistency topological feature matrix; S6, performing graph structure-based association coding on the sequence of the charging process temperature local timing feature vectors and the local temperature timing feature consistency topological feature matrix to obtain a consistent topological charging process temperature local timing feature; S7, based on the consistent topological charging process temperature local timing feature, determining whether the charging current value at the current time point should be increased, decreased or maintained.

[0030] In particular, the S1 obtains the temperature values ​​at multiple predetermined time points in a predetermined time period during the charging process. In one example, the temperature values ​​at multiple predetermined time points in a predetermined time period during the charging process can be obtained by a temperature sensor. It is worth mentioning that a temperature sensor is a device for measuring the temperature of an environment or an object. It can convert temperature into an electrical signal or a digital signal for monitoring, control or recording.

[0031] In particular, S2 arranges the temperature values ​​at the multiple predetermined time points according to the time dimension as a temperature time series input vector of the charging process. Considering that the temperature value in the charging process has a dynamic change law of time series in the time dimension, that is, the temperature value is constantly changing during the charging process, and there is a time series correlation relationship between the temperature values ​​at the multiple predetermined time points. Therefore, in the technical solution of the present application, in order to capture and characterize the time series dynamic change characteristics of the temperature value in the time dimension, it is necessary to arrange the temperature values ​​at the multiple predetermined time points according to the time dimension as a temperature time series input vector of the charging process, so as to integrate the temperature time series distribution information of the charging process.

[0032] Accordingly, in a possible implementation, the temperature values ​​of the multiple predetermined time points can be arranged according to the time dimension into a charging process temperature timing input vector through the following steps, for example: collect temperature value data at predetermined time points to ensure that each time point has a corresponding temperature value; determine the order of the time points and arrange these time points in chronological order during the charging process; create an empty vector for storing the temperature timing input; starting from the earliest time point, traverse the sorted time point list one by one; for each time point, find the corresponding temperature value; add the temperature value to the temperature timing input vector; continue to traverse the next time point, repeat steps 5 and 6 until all time points are traversed; after the traversal is completed, the obtained vector is the charging process temperature timing input vector arranged according to the time dimension.

[0033] In particular, the S3 performs upsampling based on linear interpolation on the charging process temperature time series input vector to obtain an upsampled charging process temperature time series input vector. It should be understood that in order to improve the ability to capture subtle temperature change characteristics during the charging process, in the technical solution of the present application, the charging process temperature time series input vector is further upsampled based on linear interpolation to obtain an upsampled charging process temperature time series input vector, so as to increase the density and smoothness of the temperature data, thereby facilitating the subsequent better representation of the temperature time series characteristics. It should be understood that through the upsampling of linear interpolation, the data points in the original charging process temperature time series input vector can be interpolated to generate more data points. In this way, it helps to increase the resolution in the time dimension, making the change of the charging process temperature more detailed and visible. At the same time, linear interpolation can perform smooth interpolation between sampling points, reduce the impact of noise and mutation, and improve the continuity and stability of the data. In other words, the upsampled charging process temperature time series input vector after upsampling can provide more detailed and accurate charging process temperature time series change information, and provide richer data for subsequent feature extraction and modeling processes. This helps to enhance the model's ability to perceive the temperature timing changes during the charging process, improve the control accuracy of the charging current value, and reduce the polarization phenomenon inside the battery.

[0034] It is worth noting that upsampling based on linear interpolation is a data processing method used to increase the number of data points in a time series or continuous data to obtain higher temporal resolution or a smoother data curve. This method uses linear interpolation to estimate the value of the newly added data points, and performs interpolation calculations based on the known original data points. The basic principle of linear interpolation is to assume that the numerical change between two known data points is linear, and then perform a proportional calculation between the two known data points based on the interpolation position. For example, if there are two known data points A and B, the time interval is t, and the interpolation multiple is m, then m data points will be inserted between A and B. For each inserted point, its ratio between A and B is determined according to its relative position, and then the ratio is applied to the numerical difference between A and B to calculate the value of the interpolation point. By upsampling based on linear interpolation, the density of the data can be increased, the data can be made smoother, and a more detailed temporal resolution can be provided. This is often used in the fields of processing time series data, signal processing, image processing, etc.

[0035] Accordingly, in a possible implementation, the charging process temperature timing input vector can be upsampled based on linear interpolation to obtain an upsampled charging process temperature timing input vector through the following steps, for example: obtaining the original charging process temperature timing input vector. This vector contains temperature data points in the charging process, which are usually collected at a certain time interval; calculating the time interval between the original data points. By calculating the time difference between adjacent data points, the time interval between the original data points can be obtained; determining the upsampling multiple. The upsampling multiple indicates how many times the new sampling rate is the original sampling rate. For example, if the upsampling multiple is 2, it means that the new sampling rate is twice the original sampling rate; calculating the new time interval. Divide the time interval between the original data points by the upsampling multiple to obtain a new time interval; calculating the number of new data points. The number of new data points is equal to the number of original data points multiplied by the upsampling multiple; creating a new time vector. According to the new time interval and the number of new data points, create a new time vector for storing the timestamps of the interpolated data points; creating a new temperature vector. According to the number of new data points, create a new temperature vector for storing the interpolated temperature data points; performing linear interpolation. For each new data point, find two adjacent data points in the original data point according to its timestamp. Use the linear interpolation formula to calculate the temperature value of the new data point according to the timestamp and the temperature values ​​of the adjacent data points; output the new time vector and temperature vector as the upsampled charging process temperature timing input vector.

[0036] In particular, the step S4 extracts local time series features from the upsampled charging process temperature time series input vector to obtain a sequence of local time series feature vectors of the charging process temperature. In particular, in a specific example of the present application, Figure 3As shown, the S4 includes: S41, vector segmentation of the upsampled charging process temperature timing input vector to obtain a sequence of charging process temperature local timing input vectors; S42, feature extraction of the sequence of charging process temperature local timing input vectors by a temperature timing feature extractor based on a deep neural network model to obtain a sequence of charging process temperature local timing feature vectors.

[0037] Specifically, in S41, the upsampled charging process temperature time series input vector is vector segmented to obtain a sequence of charging process temperature local time series input vectors. It should be understood that when extracting the time series change characteristics of the charging process temperature, in order to better capture the time series change characteristic information of the temperature value in the time dimension, it is further necessary to vector segment the upsampled charging process temperature time series input vector to obtain a sequence of charging process temperature local time series input vectors, so as to better extract the local time series detail change characteristic information of the charging process temperature in different time periods in the subsequent step.

[0038] It is worth noting that vector segmentation is the process of splitting a vector into multiple sub-vectors according to certain rules or conditions. In vector segmentation, the original vector is divided into multiple smaller vectors, and each sub-vector contains a part of the elements of the original vector. Vector segmentation can be used in many fields such as data processing, feature extraction, and sequence analysis. It can decompose a larger vector into smaller parts, making processing and analysis more flexible and efficient. The segmented sub-vectors can be operated independently or used as input for further calculations or model building.

[0039] Specifically, the S42 performs feature extraction on the sequence of the local time series input vectors of the temperature of the charging process through a temperature time series feature extractor based on a deep neural network model to obtain a sequence of local time series feature vectors of the temperature of the charging process. In particular, in a specific example of the present application, the temperature time series feature extractor based on a deep neural network model is a temperature time series feature extractor based on a one-dimensional convolution layer. That is, the sequence of local time series input vectors of the temperature of the charging process is subjected to feature extraction in a temperature time series feature extractor based on a one-dimensional convolution layer to extract the local time series detail change feature information of the temperature value of the charging process in each local time period on the time dimension, thereby obtaining a sequence of local time series feature vectors of the temperature of the charging process. In this way, it is conducive to the time series change trend analysis of the document value of the charging process and the control of the charging current value. Specifically, each layer of the temperature timing feature extractor based on the one-dimensional convolution layer is used to perform the following operations on the input data in the forward transfer of the layer: convolution processing is performed on the input data to obtain a convolution feature map; pooling is performed on the convolution feature map based on a feature matrix to obtain a pooling feature map; and nonlinear activation is performed on the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the temperature timing feature extractor based on the one-dimensional convolution layer is a sequence of local timing feature vectors of the temperature of the charging process, and the input of the first layer of the temperature timing feature extractor based on the one-dimensional convolution layer is a sequence of local timing input vectors of the temperature of the charging process.

[0040] It is worth noting that the one-dimensional convolution layer is a neural network layer commonly used in deep learning to process data with temporal structure. It is one of the core components of convolutional neural networks (CNNs). The one-dimensional convolution layer extracts features by performing a one-dimensional convolution operation on the input data. It uses a learnable filter (also called a convolution kernel or kernel) to perform an element-by-element sliding window convolution operation on the input data. The weights of the filter are learned through backpropagation during the training process to automatically capture the features in the input data. The working principle of the one-dimensional convolution layer is as follows: Input data: The one-dimensional convolution layer receives a one-dimensional input data, such as an encoded representation of time series data or text data; Filter: The one-dimensional convolution layer contains multiple filters, each of which is a one-dimensional weight vector. The number of these filters can be set as needed; Convolution operation: For each filter, the one-dimensional convolution layer performs an element-by-element convolution operation on the filter with the input data. This is equivalent to sliding the filter on the input data and performing a dot product operation on each position; Feature map: The result of the convolution operation generates a feature map, which contains the representation of the local features in the input data. The size of the feature map depends on the number of filters and the stride of the convolution operation; Activation function: For each feature map, an activation function can be applied to introduce nonlinearity. Common activation functions include ReLU, sigmoid, and tanh.

[0041] It is worth mentioning that in other specific examples of the present application, the sequence of the local time series input vector of the temperature of the charging process can also be feature extracted by a temperature time series feature extractor based on a deep neural network model to obtain the sequence of the local time series feature vector of the temperature of the charging process, for example: creating a temperature time series feature extractor model based on a deep neural network. The model can be a neural network model composed of stacking multiple convolutional layers, pooling layers and fully connected layers, which is used to extract features from the time series input vector; the sequence of the local time series input vector of the temperature of the charging process is provided as input to the temperature time series feature extractor model; the temperature time series feature extractor model converts each local time series input vector into a corresponding local time series feature vector through forward propagation calculation. These feature vectors can be extracted high-level abstract features with better characterization capabilities; the sequence of local time series feature vectors of the temperature of the charging process extracted can be used for further analysis, modeling or other tasks. For example, they can be input into the subsequent fully connected layer for classification, regression or other prediction tasks; according to specific needs, appropriate regularization techniques such as batch normalization, dropout, etc. can be added to the temperature time series feature extractor model to improve the generalization ability of the model and reduce overfitting.

[0042] It is worth mentioning that in other specific examples of the present application, the upsampled charging process temperature timing input vector can also be subjected to local timing feature extraction to obtain a sequence of local timing feature vectors of the charging process temperature by other means, for example: determining the window size. The window size refers to the number of data points used to extract local timing features. According to specific needs and the characteristics of the charging process, select a suitable window size; determine the sliding step size. The sliding step size refers to the moving distance of each sliding window. A portion of the window size can be selected as the sliding step size, or other suitable values ​​can be selected; create an empty sequence of local timing feature vectors. This sequence is used to store the extracted local timing feature vectors; traverse the upsampled charging process temperature timing input vector. Starting from the beginning, traverse the upsampled charging process temperature timing input vector with the sliding step size as the interval; for each window, extract the local timing features. In the current window, extract the local timing features from the upsampled charging process temperature timing input vector. Various statistics, such as mean, variance, maximum value, minimum value, etc., can be used as local timing features; the extracted local timing features are added to the local timing feature vector sequence; the sliding window is continued until the entire upsampled charging process temperature timing input vector is traversed; the local timing feature vector sequence is used as the output of the charging process temperature local timing feature vector.

[0043] In particular, the S5 performs local temperature time series feature consistency association coding on each charging process temperature local time series feature vector in the sequence of the charging process temperature local time series feature vector to obtain a local temperature time series feature consistency topological feature matrix. In particular, in a specific example of the present application, Figure 4 As shown, the S5 includes: S51, calculating the cosine similarity between any two local timing feature vectors of the charging process temperature in the sequence of the local timing feature vectors of the charging process temperature to obtain a local temperature timing feature consistency topological matrix; S52, passing the local temperature timing feature consistency topological matrix through a topological feature extractor based on a convolutional neural network model to obtain the local temperature timing feature consistency topological feature matrix.

[0044] Specifically, the S51 calculates the cosine similarity between any two local time series feature vectors of the charging process temperature in the sequence of the local time series feature vectors of the charging process temperature to obtain a local temperature time series feature consistency topological matrix. Considering that in the lithium battery charging process, the temperature time series features in each local time period have a correlation relationship based on the overall time series. In order to fully capture the change pattern and law of the temperature during the charging process, so as to understand the dynamic changes inside the battery, in the technical solution of the present application, the cosine similarity between any two local time series feature vectors of the charging process temperature in the sequence of the local time series feature vectors of the charging process temperature is further calculated to obtain a local temperature time series feature consistency topological matrix. It should be understood that by comparing the similarity between the local time series feature vectors of the charging process temperature in different time segments, the temperature correlation and change trend in each local time period can be captured and depicted. In addition, by analyzing the local temperature time series feature consistency topological matrix, the pattern and law of temperature change during the battery charging process can be revealed, which helps to understand the dynamic changes inside the battery and provide guidance for subsequent charging strategy adjustments. For example, if it is found that the temperature change characteristics within a certain local time period are inconsistent with the temperature change characteristics within other local time periods, the charging current can be adjusted according to this characteristic information to achieve a better repair effect.

[0045] Specifically, in S52, the local temperature timing feature consistency topological matrix is ​​passed through a topological feature extractor based on a convolutional neural network model to obtain the local temperature timing feature consistency topological feature matrix. That is, the local temperature timing feature consistency topological matrix is ​​passed through a topological feature extractor based on a convolutional neural network model to perform feature mining to extract the consistent topological association feature information between the local timing features related to the temperature in each local time period during the charging process, thereby obtaining the local temperature timing feature consistency topological feature matrix. Specifically,

[0046] It is worth noting that convolutional neural network (CNN) is a deep learning model that is mainly used to process tasks with grid-structured data, such as image and speech recognition. The core idea of ​​CNN is to achieve feature extraction and classification of input data through components such as convolutional layers, pooling layers, and fully connected layers. The following are the main components and features of CNN: Convolutional layer: The convolutional layer is the core component of CNN. It extracts local features of the input data by performing convolution operations on the input data using a set of learnable filters (also called convolution kernels). The convolution operation can capture the spatial local relationship of the input data, such as edges and textures in an image; Pooling layer: The pooling layer is used to reduce the spatial size of the feature map while retaining important feature information. Commonly used pooling operations include maximum pooling and average pooling, which extract the maximum value or average value of the local area as the pooling result; Fully connected layer: The fully connected layer connects the output of the convolution and pooling layers to one or more fully connected layers for classification or regression of the extracted features. The neurons in the fully connected layer are connected to all the neurons in the previous layer; Activation function: In convolutional neural networks, activation functions are usually applied to the output of convolutional layers and fully connected layers to introduce nonlinear transformations and increase the network's expressive power. Common activation functions include ReLU, Sigmoid, and Tanh; Weight sharing: The convolutional layer in CNN uses a weight sharing strategy, that is, the same convolution kernel is applied at different positions. This can greatly reduce the number of parameters that need to be learned and improve the efficiency and generalization ability of the model. By stacking multiple convolutional layers and pooling layers, CNN can extract abstract features of input data layer by layer, from low-level features (such as edges) to high-level features (such as object shape). This makes CNN perform well in computer vision tasks such as image classification, object detection, and semantic segmentation. At the same time, CNN can also be applied in other fields, such as natural language processing and signal processing.

[0047] It is worth mentioning that in other specific examples of the present application, each local time series feature vector of the charging process temperature in the sequence of the local time series feature vector of the charging process temperature can also be encoded by other methods to obtain a local temperature time series feature consistency topological feature matrix, for example: define a window size (for example, W) and a stride (for example, S). The window size represents the number of local feature vectors considered in the sequence, and the stride represents the sliding interval of the window in the sequence; the window is sliding according to the stride from the sequence of the local time series feature vector of the charging process temperature, and a subsequence of local feature vectors with a window size of W is taken each time; for the local feature vector subsequence in each window, the similarity or consistency of its internal features is calculated. Various similarity metrics can be used, such as Euclidean distance, cosine similarity, etc.; based on the results of the similarity calculation, a local temperature time series feature consistency association coding matrix is ​​constructed. Each element of the matrix represents the degree of consistency between the corresponding local feature vector subsequences; the local temperature timing feature consistency association coding matrix can be further processed, for example, by applying thresholds or weights to screen or adjust the consistency relationship; the final local temperature timing feature consistency topological feature matrix can be used to describe the consistency relationship between the local temperature timing feature vectors of the charging process.

[0048] In particular, the S6 performs graph-structure-based association coding on the sequence of the local time series feature vectors of the charging process temperature and the local temperature time series feature consistency topological feature matrix to obtain the consistent topological charging process temperature local time series feature. That is, the waveform feature vectors of the sound signal sampling window in the sequence of the local time series feature vectors of the charging process temperature are used as the feature representation of the node, and the local temperature time series feature consistency topological feature matrix is ​​used as the feature representation of the edge between the nodes. The charging process temperature time series feature matrix obtained by two-dimensional arrangement of the sequence of the local time series feature vectors of the charging process temperature and the local temperature time series feature consistency topological feature matrix are passed through a graph neural network model to obtain the consistent topological charging process temperature local time series feature matrix. Specifically, the graph neural network model performs graph structure data encoding on the charging process temperature time series feature matrix and the local temperature time series feature consistency topological feature matrix through learnable neural network parameters to obtain the consistent topological charging process temperature local time series feature matrix containing irregular local temperature time series consistency topological association features and the temperature time series feature information in each local time series segment.

[0049] It is worth noting that graph neural network (GNN) is a deep learning model for processing graph structured data. Unlike traditional convolutional neural networks (CNN) which are suitable for processing grid structured data (such as images), GNN is specially designed to process non-grid structured data, such as social networks, knowledge graphs, molecular structures, etc. The core idea of ​​graph neural networks is to learn the representation of nodes through the propagation and aggregation of local neighbor information between nodes. The following are the main components and characteristics of GNN: Node representation: GNN captures the characteristics of nodes by representing each node as a vector or matrix. Initially, each node has an initial feature vector. During the training process of GNN, the node representation is updated with the iteration of information propagation and aggregation; Graph convolution layer: The graph convolution layer is the core component of GNN, which is used to propagate and aggregate the information of nodes in the graph structure. The graph convolution layer updates the representation of each node by considering the neighbor information of the node. Typical graph convolution layers use weighted average or splicing of neighbor nodes to aggregate information; Graph pooling layer: Similar to the pooling layer in CNN, the graph pooling layer is used to reduce the scale of the graph and retain important node information. Common graph pooling methods include graph subgraph sampling and graph structure aggregation; Graph attention mechanism: In order to better handle the node relationships in the graph, the graph attention mechanism is introduced in GNN. The graph attention mechanism pays more attention to important neighbor nodes by learning the attention weights between each node and its neighbor nodes; Graph-level tasks and node-level tasks: GNN can be used to solve graph-level tasks and node-level tasks. Graph-level tasks involve predicting or classifying the entire graph, such as graph classification and graph generation. Node-level tasks involve predicting or classifying each node, such as node classification and node embedding. The advantage of graph neural networks is that they can process unstructured graph data and extract useful features from the global structure and local neighbor information of the graph. It has a wide range of applications in social network analysis, recommendation systems, molecular design and other fields.

[0050] In particular, the S7 determines whether the charging current value at the current time point should be increased, decreased or maintained based on the local timing characteristics of the temperature during the consistent topology charging process. In particular, in a specific example of the present application, Figure 5 As shown, the S7 includes: S71, performing soft matching of the rank arrangement distribution of the consistent topology charging process temperature local timing feature matrix using the feature scale as an imitation mask to obtain an optimized consistent topology charging process temperature local timing feature matrix; S72, passing the optimized consistent topology charging process temperature local timing feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the charging current value at the current time point should be increased, decreased or maintained.

[0051] Specifically, in S71, the consistent topology charging process temperature local time series feature matrix is ​​soft matched with the feature scale as the rank arrangement distribution of the imitation mask to obtain the optimized consistent topology charging process temperature local time series feature matrix. In particular, in the technical solution of the present application, when the sequence of the local time series feature vectors of the charging process temperature and the consistent topological feature matrix of the local temperature time series features are obtained through a graph neural network model, each row feature vector of the consistent topological charging process temperature local time series feature matrix expresses the topological association representation of the temperature time series association features in the local time domain based on the similarity topology of the temperature time series features between each local time domain in the global time domain. Therefore, if the temperature time series association features in each local time domain are used as foreground object features, the graph neural network model will also introduce background distribution noise when performing feature topological association expression, and while performing high-rank distribution representation between vectors and matrices, it will also cause the probability density mapping error of the consistent topological charging process temperature local time series feature matrix relative to the association feature time series distribution in each local time domain due to the heterogeneous distribution of the high-dimensional features of the local time series association features of the temperature values ​​in the local time series, thereby affecting the accuracy of the classification result obtained by the classifier of the classification feature matrix. Based on this, the applicant of the present application performs soft matching of the rank arrangement distribution of the consistent topology charging process temperature local time series feature matrix, for example, denoted as M, with the feature scale as the mimicking mask, which is specifically expressed as:

[0052]

[0053] Wherein, M is the local time series characteristic matrix of temperature in the consistent topology charging process, m i,j is the eigenvalue of the (i, j)th position of the local time series characteristic matrix of the temperature of the consistent topology charging process, S is the scale of the local time series characteristic matrix of the temperature of the consistent topology charging process, represents the square of the Frobenius norm of the temperature local timing characteristic matrix of the consistent topology charging process, ||M||2 represents the bi-norm of the temperature local timing characteristic matrix of the consistent topology charging process, and α is a weighted hyperparameter, exp(·) represents the exponential operation, and m′ i,jIt is the eigenvalue of the (i, j)th position of the local temporal feature matrix of the temperature of the optimized consistent topology charging process. Here, the soft matching of the rank arrangement distribution of the feature scale as an imitation mask can focus on the foreground object features and ignore the background distribution noise when mapping the high-dimensional features to be classified into the probability density space, and the distribution soft matching of the pyramid rank arrangement distribution is performed by different norms of the local temporal feature matrix M of the temperature of the consistent topology charging process, so as to effectively capture the correlation between the central area and the tail area of ​​the probability density distribution, avoid the probability density mapping deviation caused by the spatial heterogeneous distribution of the high-dimensional features of the local temporal feature matrix M of the temperature of the consistent topology charging process, and improve the accuracy of the classification results obtained by the classifier of the local temporal feature matrix of the temperature of the consistent topology charging process. In this way, corresponding measures can be taken to control the temperature based on the change of temperature, such as reducing the charging current or suspending charging. In this way, the charging current value can be adaptively adjusted based on the actual change of temperature, thereby reducing the polarization phenomenon inside the battery and improving the performance and safety of the battery.

[0054] Specifically, in S72, the temperature local timing feature matrix of the optimized consistency topology charging process is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the charging current value at the current time point should be increased, decreased, or maintained. More specifically, the temperature local timing feature matrix of the optimized consistency topology charging process is expanded into a classification feature vector based on a row vector or a column vector; the classification feature vector is fully connected encoded using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and the encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result. Among them, the classification label of the classifier is a control label of the charging current value. After obtaining the classification result, the charging current value is adaptively adjusted based on the temperature change, thereby reducing the polarization phenomenon inside the battery and improving the performance and safety of the battery.

[0055] A classifier is a machine learning model or algorithm that is used to classify input data into different categories or labels. A classifier is part of supervised learning and performs classification tasks by learning the mapping relationship from input data to output categories.

[0056] A fully connected layer is a common type of layer in a neural network. In a fully connected layer, each neuron is connected to all neurons in the previous layer, and each connection has a weight. This means that each neuron in a fully connected layer receives inputs from all neurons in the previous layer, and performs a weighted sum of these inputs through weights, and then passes the result to the next layer.

[0057] The Softmax classification function is a commonly used activation function for multi-classification problems. It converts each element of the input vector into a probability value between 0 and 1, and the sum of these probability values ​​is equal to 1. The Softmax function is often used in the output layer of a neural network, especially for multi-classification problems, because it can map the network output to the probability distribution of each category. During the training process, the output of the Softmax function can be used to calculate the loss function and update the network parameters through the back propagation algorithm. It is worth noting that the output of the Softmax function does not change the relative size relationship between the elements, but only normalizes them. Therefore, the Softmax function does not change the characteristics of the input vector, but only converts it into a probability distribution form.

[0058] It is worth mentioning that in other specific examples of the present application, it is also possible to determine whether the charging current value at the current time point should be increased, decreased or maintained based on the local timing characteristics of the temperature of the consistent topology charging process in other ways, for example: first, it is necessary to collect the temperature and charging current data of the consistent topology charging process; extract the local timing characteristics of the temperature from the collected data; and establish a model based on the extracted characteristics to predict the charging current value at the current time point. Machine learning algorithms, such as regression models or time series models, can be used to build a prediction model; and historical data are used to train the model. The known local timing characteristics of the temperature and the corresponding charging current values ​​are used as input and output, and the relationship between them is learned through model training; the trained model is used to predict the local timing characteristics of the temperature at the current time point to determine whether the charging current value at the current time point should be increased, decreased or maintained. According to the output of the model, the difference between the current current value and the predicted value can be judged, and then the action to be taken can be determined.

[0059] In summary, the electrochemical repair method of the lithium battery according to the embodiment of the present application is explained, which monitors the changes in battery temperature in real time during the charging process of the lithium battery, and introduces data processing and analysis algorithms at the back end to analyze the temperature timing changes during the charging process, so as to take corresponding measures to control the temperature based on the temperature changes, such as reducing the charging current or suspending charging. In this way, the charging current value can be adaptively adjusted based on the actual temperature conditions, thereby reducing the polarization phenomenon inside the battery and improving the battery performance and safety.

[0060] In the embodiments of the present application, Figure 6 As shown, the capacity loss caused by dynamic polarization is a reversible capacity loss and can be compensated and repaired by the charge and discharge system (see Figure 4); The capacity loss caused by material structure fatigue and battery internal structure damage is irreversible capacity loss, but it can be partially restored through appropriate physical repair. Battery repair technology can greatly improve the efficiency of retired power battery cascade utilization and product quality, which is a breakthrough core technology in the retired power battery cascade utilization industry.

[0061] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for electrochemical repair of a lithium battery, characterized in that: include: Obtain temperature values ​​at multiple predetermined time points within a predetermined time period during the charging process; Arranging the temperature values ​​of the plurality of predetermined time points according to the time dimension into a charging process temperature time series input vector; Performing up-sampling based on linear interpolation on the charging process temperature timing input vector to obtain an up-sampled charging process temperature timing input vector; Performing local time series feature extraction on the upsampled charging process temperature time series input vector to obtain a sequence of charging process temperature local time series feature vectors; Performing local temperature time series feature consistency association coding on each charging process temperature local time series feature vector in the sequence of the charging process temperature local time series feature vectors to obtain a local temperature time series feature consistency topological feature matrix; Performing graph-structure-based association coding on the sequence of the charging process temperature local time series feature vectors and the local temperature time series feature consistency topology feature matrix to obtain consistent topology charging process temperature local time series features; Based on the local time series characteristics of the temperature in the consistent topology charging process, it is determined whether the charging current value at the current time point should be increased, decreased or maintained.

2. The electrochemical repair method of a lithium battery according to claim 1, characterized in that: Extracting local time series features from the upsampled charging process temperature time series input vector to obtain a sequence of local time series feature vectors of the charging process temperature includes: Performing vector segmentation on the up-sampled charging process temperature timing input vector to obtain a sequence of charging process temperature local timing input vectors; The sequence of the local time series input vectors of the temperature of the charging process is subjected to feature extraction by a temperature time series feature extractor based on a deep neural network model to obtain a sequence of the local time series feature vectors of the temperature of the charging process.

3. The electrochemical repair method of a lithium battery according to claim 2, characterized in that: The temperature time series feature extractor based on the deep neural network model is a temperature time series feature extractor based on a one-dimensional convolutional layer.

4. The electrochemical repair method of a lithium battery according to claim 3, characterized in that: Performing local temperature time series feature consistency association coding on each charging process temperature local time series feature vector in the sequence of the charging process temperature local time series feature vectors to obtain a local temperature time series feature consistency topological feature matrix, including: Calculating the cosine similarity between any two local time series feature vectors of the charging process temperature in the sequence of the local time series feature vectors of the charging process temperature to obtain a local temperature time series feature consistency topological matrix; The local temperature timing feature consistency topological matrix is ​​passed through a topological feature extractor based on a convolutional neural network model to obtain the local temperature timing feature consistency topological feature matrix.

5. The electrochemical repair method of a lithium battery according to claim 4, characterized in that: The sequence of the local time series feature vectors of the charging process temperature and the local temperature time series feature consistency topology feature matrix are subjected to graph structure-based association coding to obtain the consistent topology charging process temperature local time series feature, including: The sequence of the local time series feature vectors of the charging process temperature and the consistent topological feature matrix of the local temperature time series features are passed through a graph neural network model to obtain a consistent topological charging process temperature local time series feature matrix as the consistent topological charging process temperature local time series features.

6. The electrochemical repair method of a lithium battery according to claim 5, characterized in that: Based on the local time series characteristics of the temperature during the consistent topology charging process, determining whether the charging current value at the current time point should be increased, decreased, or maintained includes: Performing soft matching of the rank arrangement distribution of the consistent topology charging process temperature local time series feature matrix with the feature scale as the imitation mask to obtain an optimized consistent topology charging process temperature local time series feature matrix; The optimized consistency topology charging process temperature local timing feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the charging current value at the current time point should be increased, decreased or maintained.

7. The electrochemical repair method of a lithium battery according to claim 6, characterized in that: The consistent topology charging process temperature local time series feature matrix is ​​subjected to soft matching of the rank arrangement distribution with the feature scale as the imitation mask to obtain an optimized consistent topology charging process temperature local time series feature matrix, including: The local time series characteristic matrix of the temperature of the consistent topology charging process is soft matched with the rank arrangement distribution of the characteristic scale as the imitation mask by using the following optimization formula to obtain the local time series characteristic matrix of the temperature of the optimized consistent topology charging process; Wherein, the optimization formula is: Wherein, M is the local time series characteristic matrix of temperature in the consistent topology charging process, m i,j is the eigenvalue of the (i, j)th position of the local time series characteristic matrix of the temperature of the consistent topology charging process, S is the scale of the local time series characteristic matrix of the temperature of the consistent topology charging process, represents the square of the Frobenius norm of the temperature local timing characteristic matrix of the consistent topology charging process, ||M||2 represents the bi-norm of the temperature local timing characteristic matrix of the consistent topology charging process, and α is a weighted hyperparameter, exp(·) represents the exponential operation, and m′ i,j It is the eigenvalue of the (i, j)th position of the temperature local timing characteristic matrix of the optimized consistency topology charging process.

8. The electrochemical repair method of a lithium battery according to claim 7, characterized in that: The optimized consistency topology charging process temperature local time series feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the charging current value at the current time point should be increased, decreased or maintained, including: Expanding the local time series characteristic matrix of the temperature of the optimized consistency topology charging process into a classification characteristic vector based on a row vector or a column vector; Performing full connection encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

Images