Method and device for predicting health state of retired power battery and electronic equipment

Abstract

The invention relates to a retired power battery health state prediction method and device and electronic equipment, and belongs to the technical field of battery health state assessment, and the method comprises the steps: carrying out the level division of obtained battery parameters, and obtaining a plurality of different levels of battery parameters; constructing a graph network structure with a timestamp by taking the battery numbers and the battery parameter grade values as nodes and taking the names of the battery parameters as edges; according to the graph network structure, constructing a battery state characteristic matrix based on time; and inputting the time-based battery state characteristic matrix into the completely trained battery state-of-health estimation model to obtain the recombined utilization state-of-health of the retired power battery. The technical problem that in the prior art, due to the fact that the change rules of parameters such as capacity, voltage and internal resistance of the retired power battery are affected by the inconsistency of the battery, it is difficult to capture, and the difficulty of accurate estimation of the health state is large is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health status assessment, and in particular to a method, device and electronic equipment for predicting the health status of retired power batteries. Background Art

[0002] There are two ways to reuse retired power batteries: direct utilization and reorganization. Reorganization means that retired power batteries are disassembled and reassembled according to different consistency and capacity levels to form reorganized batteries for reuse. However, since the consistency of retired power batteries is much worse than that of new batteries, the problem of poor consistency is magnified again during the battery reorganization process. The battery's own battery management system and the new battery management system have large deviations and poor accuracy in estimating the parameters related to the health status of the reorganized battery, resulting in overcharge and over-discharge problems in some reorganized batteries, which may cause battery thermal runaway and other safety hazards. Therefore, it is of great significance to estimate the health status of retired power batteries for reorganization, provide data support for the battery management system of the reorganized battery to achieve balanced management, and improve the efficiency of the reuse of retired power batteries.

[0003] In terms of health state estimation for retired power batteries, it is necessary to redesign the health state estimation method in combination with the characteristics of the recombined battery itself. The estimation method is required to solve the problem of poor inconsistency faced by the current recombined battery and be quickly and easily deployed in real applications. The main methods currently used are experimental estimation and model-based methods. The experimental estimation method requires a large number of experiments to obtain data related to the health state to analyze the aging law of the battery. The main disadvantages of this method are high experimental costs, and the experimental conditions are difficult to replicate in actual applications, which affects the accuracy of the estimation and the inability to perform dynamic estimation. The model-based method is to establish the mathematical relationship between the changes in parameters such as current and voltage and SOH attenuation based on the physical and chemical process data during the battery aging process. The disadvantage of this method is that the model parameters are highly dependent, and it is difficult to obtain accurate parameters in actual applications. It usually takes a long time to calculate and complex algorithms to ensure the accuracy of the estimation, resulting in a lag in the estimation results, and real-time estimation cannot be achieved. The changing laws of parameters such as capacity, voltage, and internal resistance of retired power batteries caused by battery inconsistency are difficult to capture, which makes it difficult to accurately estimate the health state. Summary of the invention

[0004] In view of this, it is necessary to provide a method, device and electronic device for predicting the health status of retired power batteries to solve the technical problem in the prior art that the changing patterns of parameters such as capacity, voltage, internal resistance, etc. of retired power batteries caused by battery inconsistency are difficult to capture, resulting in difficulty in accurately estimating the health status.

[0005] In order to solve the above problems, the present invention provides a method for predicting the health status of retired power batteries, comprising: The obtained battery parameters are divided into different levels to obtain battery parameters of multiple different levels; Using battery numbers and battery parameter level values ​​as nodes and battery parameter names as edges, a graph network structure with timestamps is constructed; And according to the graph network structure, a time-based battery status feature matrix is ​​constructed; The time-based battery state feature matrix is ​​input into the well-trained battery health state estimation model to obtain the health state of the retired power battery for reutilization.

[0006] In a possible implementation, the graph network structure with timestamps is constructed by taking battery numbers and battery parameter level values ​​as nodes and battery parameter names as edges, including: Construct the initial graph network structure and define the head entity node and the tail entity node; A static graph network structure at a certain moment is constructed with the battery number as the head entity node, the battery parameter level value as the tail entity node, and the battery parameter name as the edge; Based on the static graph network structure, the correlation between the battery status and the battery parameter level value and the battery parameters is determined.

[0007] In a possible implementation, determining the association between the battery state, the battery parameter level value, and the battery parameter based on the static graph network structure includes: Based on the static graph network structure, a four-tuple model of battery status, battery status value, battery number and time is constructed; Based on the four-tuple model, the time sequence relationship between the battery state value and the battery parameter level value is determined according to the battery number.

[0008] In a possible implementation, constructing a time-based battery state feature matrix according to a graph network structure includes: Using a distance-based translation model TransE, determining a feature coding value of each entity node in the graph network structure, and constructing an initial vector according to the feature coding value; According to the initial vector of each entity node, a battery state feature matrix is ​​constructed.

[0009] In a possible implementation, the distance-based translation model TransE is used to determine the feature coding value of each entity node in the graph network structure, and the initial vector is constructed according to the feature coding value, including: Dividing the graph network structure into a head entity node set, a tail entity node set and an attribute edge set; Based on the equivalence relationship between the sum of the head entity initial vector and the attribute edge initial vector and the tail entity initial vector, the initial distance between the head entity initial vector and the tail entity initial vector is determined according to a preset Euclidean distance scoring function; The initial distance is optimized according to the minimum objective function to determine the initial distance of each entity.

[0010] In a possible implementation, the preset Euclidean distance scoring function is expressed by the following formula:

[0011] in, represents the initial distance, Indicates the head entity node, represents the attribute edge, t represents the tail entity node, and Represent the first distance and the second distance respectively.

[0012] In a possible implementation, the fully trained battery health state estimation model includes at least one of a long short-term memory recurrent network model, a deep learning model, and a support vector machine.

[0013] In a second aspect, the present invention further provides a retired power battery health status prediction device, comprising: A classification module is used to classify the acquired battery parameters to obtain battery parameters of multiple different levels; A graph network structure construction module, which is used to construct a graph network structure with timestamps by using battery numbers and battery parameter level values ​​as nodes and battery parameter names as edges; A feature matrix construction module is used to construct a time-based battery state feature matrix according to a graph network structure; The health state determination module is used to input the time-based battery state feature matrix into the well-trained battery health state estimation model to obtain the health state of the retired power battery for re-utilization.

[0014] In a third aspect, the present invention further provides an electronic device, comprising: a processor and a memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the retired power battery health status prediction method as described above are implemented.

[0015] In a fourth aspect, the present invention also provides a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the retired power battery health status prediction method as described above.

[0016] The beneficial effects of the present invention are as follows: by dividing the acquired battery parameters into levels, multiple battery parameters of different levels are obtained, and a graph network structure is constructed according to the levels of the battery parameters, the battery numbers and the names of the battery parameters, so as to capture the hidden relationship between different battery parameters and their levels on the health status of retired power batteries for reclamation and utilization based on the graph network structure, and construct a battery state feature matrix based on the influence of the battery parameters on consistency represented by the graph network structure, and finally use a well-trained battery health status assessment model to evaluate the health status of retired power batteries for reclamation and utilization according to the battery state feature matrix, and combine the inconsistency and timing information to improve the estimation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A method flow chart of an embodiment of a method for predicting the health status of retired power batteries provided by the present invention; Figure 2 for Figure 1 A method flow chart of an embodiment of step S102; Figure 3 A schematic diagram of a structure of an embodiment of a network structure of a retired power battery health status prediction method provided by the present invention; Figure 4 A schematic diagram of another embodiment of the network structure in the retired power battery health status prediction method provided by the present invention; Figure 5 It is a schematic diagram of an embodiment of a retired power battery health status prediction device provided by the present invention; Figure 6 It is a schematic diagram of the operating environment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0019] A specific embodiment of the present invention discloses a method for predicting the health status of retired power batteries. Figure 1 ,include: S101, classifying the acquired battery parameters into different levels to obtain battery parameters of multiple different levels; S102, constructing a graph network structure with a timestamp using the battery numbers and battery parameter level values ​​as nodes and the battery parameter names as edges; S103, constructing a time-based battery state feature matrix according to the graph network structure; S104: Input the time-based battery status feature matrix into a well-trained battery health status estimation model to obtain the health status of the retired power battery for reutilization.

[0020] In this embodiment, the obtained battery parameters are divided into levels to obtain multiple battery parameters of different levels, and a graph network structure is constructed according to the levels of the battery parameters, battery numbers and names of the battery parameters, so as to capture the hidden relationship between different battery parameters and their levels on the health status of retired power batteries based on the graph network structure, and construct a battery state feature matrix based on the influence of battery parameters on consistency represented by the graph network structure. Finally, a well-trained battery health status assessment model is used to assess the health status of retired power batteries based on the battery state feature matrix, and the estimation accuracy is improved by combining inconsistency and timing information.

[0021] Specifically, the amount of battery data is huge, and the decay of battery health status is a long time series problem. In order to strengthen the connection between time series information and the decay of battery health status, the data in the battery data set needs to be divided according to certain time intervals, thereby strengthening the expression of time series information.

[0022] In step S101, the collected historical data is preprocessed, including data cleaning and data processing. Data cleaning mainly includes processing abnormal values ​​and missing values ​​of battery data in the battery management system to obtain a complete battery data set; data processing is to reprocess the battery data set so that its inconsistency impact can be captured by the graph network structure and improve the accuracy of neural network estimation. The steps include: interval processing of time series information to ensure the diversity of time series information; taking the attribute value of the battery parameter as the tail entity node of the graph network structure, and classifying the value of the tail entity node.

[0023] The essence of a graph is a network structure. Batteries with different numbers are used as the head entity nodes of the graph network structure, and the attribute values ​​of the battery parameters are used as the tail entity nodes. This will lead to too many tail entity nodes in the graph network structure, and the inconsistency between the different parameter attributes of different batteries and the battery health state decay law is sparse. Therefore, a reasonable level division of the parameter attribute values ​​as the tail entity nodes helps to reduce the number of tail entity nodes, strengthen the capture of the potential laws between inconsistency and battery health state decay, and build a more reasonable graph network structure. It should be noted that the battery parameters include at least voltage, capacity, internal resistance, temperature and self-discharge rate. The level division of the obtained battery parameters includes dividing into different levels according to the parameter values ​​of the battery parameters, and constructing a graph network structure according to the divided levels, so as to characterize the degree of influence of consistency on the battery health state according to the graph network structure, and determine the battery health state to improve the efficiency of recycling.

[0024] In some embodiments, the battery number and battery parameter level value are used as nodes, and the battery parameter name is used as an edge to construct a graph network structure with a timestamp. Figure 2 ,include: S201, constructing an initial graph network structure and defining a head entity node and a tail entity node; S202, constructing a static graph network structure at a certain moment with the battery number as the head entity node, the battery parameter level value as the tail entity node, and the battery parameter name as the edge; S203: Determine the association between the battery state, the battery parameter level value, and the battery parameter based on the static graph network structure.

[0025] In step S201, the head entity node is usually defined as a battery number (such as a battery ID), representing a specific battery instance. The tail entity node is defined as a level value of a battery parameter, such as a voltage level value (A, B, C), a capacity level value, etc. The edge is defined as the name of a battery parameter, such as "voltage", "capacity", "internal resistance", etc., connecting the head entity node and the tail entity node.

[0026] For example: Battery No.: 001 Voltage level: A Side: Voltage In this example, see Figure 3 and Figure 4, a graph network structure is constructed with the level values ​​of battery parameters. With the help of nodes and edges in the graph network structure, the relationship between the battery status and the battery parameter level values ​​can be intuitively reflected. Specifically, the overall health status of the battery can be judged by the level value of the tail entity node. For example, if most parameter levels are A, the battery status is good; if multiple parameter levels are C, the battery status is poor.

[0027] Association between battery status and parameter names: Through the edge type (such as "voltage", "capacity", "internal resistance"), you can identify which parameters have a greater impact on the battery status. For example, if a parameter (such as capacity) has a low level value, it may reflect that the battery's endurance has decreased.

[0028] It should be noted that the finer the parameter value level division, the more value levels there are, and the more nodes there are in the corresponding graph network structure. Therefore, graph network structures with different numbers of nodes are constructed according to different parameter value level divisions.

[0029] In some embodiments, determining the association between the battery state, the battery parameter level value, and the battery parameter based on the static graph network structure includes: Based on the static graph network structure, a four-tuple model of battery status, battery status value, battery number and time is constructed; Based on the four-tuple model, the time sequence relationship between the battery state value and the battery parameter level value is determined according to the battery number.

[0030] In this embodiment, the retired power battery reorganization utilizes the health state graph network structure can be expressed as ,in , represents the static graph network structure at time i, E represents the entity set in the graph network structure, R represents the relationship set in the graph network structure, It represents the set of knowledge facts that occurred at time i, that is, the knowledge facts that occurred in the real world are described by the relationship R between different entities E at time i. Combined with the historical data in the power battery management system within a certain period of time, the retired power battery reorganization is constructed based on this. The battery number in the battery pack at a certain time The four-tuple of the state of health (SOH) is represented as , through this example, we can see that at different times , different battery numbers and different battery parameter information can be represented by four-tuples.

[0031] In some embodiments, constructing a time-based battery state feature matrix according to the graph network structure includes: Using a distance-based translation model TransE, determining a feature coding value of each entity node in the graph network structure, and constructing an initial vector according to the feature coding value; According to the initial vector of each entity node, a battery state feature matrix is ​​constructed.

[0032] In this embodiment, based on the above-mentioned completed graph network structure, the distance-based translation model TransE is used to embed the entity nodes and attribute edges in the graph network structure, and determine the low-dimensional vector representation of each entity node, including: using the distance-based translation model TransE to implement the low-dimensional vector representation of the entity nodes and attribute edges in the graph network structure; splicing the low-dimensional vector representation of each entity node to obtain a feature matrix. The translation model TransE is inspired by word turn vectors and regards the relationship (attribute) in the triple as a translation operation. The judgment basis is that the sum of the head entity vector and the attribute vector should be as equal to the tail entity as possible, that is, h+r≈t; therefore, a scoring function based on Euclidean distance is used. To measure the distance between the head entity and the tail entity, and optimize the low-dimensional vector representation of the triplet according to the minimum objective function, and then determine the low-dimensional vector representation of each entity.

[0033] The TransE algorithm defines a scoring function based on Euclidean distance:

[0034] In order to better learn the representation of entities and attributes in low-dimensional vector space, the minimized objective function is:

[0035]

[0036] in >0 is a marginal constant, S is the correct set of triples, is a set of wrong triples, E is a set of entities, and the wrong triples are obtained by randomly replacing one of the head entity or the tail entity of the correct triple.

[0037] Specifically, the graph network structure is divided into entity node sets, attribute edge sets and Figure 3 Tuple set, firstly initialize the entity nodes and attribute edges in the entity node set and attribute edge set with low-dimensional vectors and Figure 3The tuple set is divided into training set, validation set and test set. The model training process first extracts triplets from the training set as positive samples, and randomly replaces the head entity node or the tail entity node of the triplets as negative samples. The scoring function is used to calculate the scores of positive and negative samples respectively. According to the algorithm expectation, the score of positive samples should be close to 0, and the score of negative samples should be far away from 0. Then, the low-dimensional vector representation of entity nodes and attribute edges is updated in the back propagation by minimizing the objective function and the optimizer to meet the requirements of the algorithm. Finally, the two indicators MRR and HITS@n are used to measure the embedding level of entity nodes and attribute edges in the graph network structure. After obtaining the low-dimensional vector representation of each entity node, it is combined to obtain the feature matrix as the input of the LSTM estimation model, which retains the data information of the graph network structure to the greatest extent.

[0038] The calculation formulas for MRR and HITS@n indicators are as follows:

[0039]

[0040] Where S represents a set of triples, represents the number of triples, represents the ranking of the i-th triple in link prediction, Represents the indicator function. The larger the MRR and HITS@n index values, the better.

[0041] The well-trained battery health state estimation model includes at least one of a long short-term memory recurrent network model, a deep learning model, and a support vector machine. Specifically, the battery health state estimation model is an LSTM estimation model, which includes an input layer, the input object is time window data, two LSTM layers, a fully connected layer, and an output layer; the problem of power battery health state decay is a long time series problem. In addition to being affected by battery parameter changes, external environment, inconsistency, etc., its decay process also shows an obvious trend. The long short-term memory network (LSTM) selects to store and delete information through different gating units, which can handle long-term dependency problems in data very well.

[0042] The LSTM network formula is as follows:

[0043] In the formula is the hidden state at time step t, is the cell state at time step t, is the input at time step t, W and b are the weight matrix and bias vector, , , , They are input gate, forget gate, cell state and output gate. is the sigmoid activation function, is the hyperbolic tangent activation function, is the Hadamard product.

[0044] Specifically, according to the feature matrix that has been embedded above, the battery data is re-divided into training set, validation set and test set, and a time window sample for multivariate rolling estimation is constructed as the input of the model. The window size (how many previous sample data are used to estimate the target data) needs to be set according to the actual situation; the training set is used to train the model to obtain the graph-embedded retired power battery health status prediction model; the validation set is used for model parameter adjustment, the trained estimation model is estimated on the validation set and the parameters are adjusted in time to obtain the optimal estimation model parameters; the test set is used for model evaluation, the estimation model that meets the accuracy is estimated on the test set, and the obtained estimated value is compared with the true value to verify the effectiveness of the model; Furthermore, three evaluation indicators are used to verify the effectiveness of the model: root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination ( ), the calculation formulas of the three evaluation indicators are:

[0045]

[0046]

[0047] In the formula, m is the number of samples, is the true value, is the true average value, is the estimated value. The smaller the RMSE and MAE values ​​are, the higher the estimation accuracy is. The larger the value, the better the model fit.

[0048] Based on this, the key points of the embodiments of the present invention mainly lie in the following aspects: 1. Capture the potential factor of battery inconsistency that affects the health state degradation of power batteries by modeling a graph network structure with time series characteristics.

[0049] 2. Set different parameter value levels to adjust the number of nodes in the graph network structure so that it can capture battery inconsistency to the greatest extent while providing sufficient graph structure information and time series information for subsequent graph embedding tasks and time series estimation tasks.

[0050] 3. Use graph embedding technology to complete the connection between the upstream graph network structure and the downstream time series estimation.

[0051] Based on the above retired power battery health status prediction method, the embodiment of the present invention also provides a retired power battery health status prediction device, see Figure 5 ,include: A classification module 510 is used to classify the acquired battery parameters to obtain battery parameters of multiple different levels; A graph network structure building module 520, for building a graph network structure with a timestamp using battery numbers and battery parameter level values ​​as nodes and battery parameter names as edges; A feature matrix construction module 530 is used to construct a time-based battery state feature matrix according to the graph network structure; The health state determination module 540 is used to input the time-based battery state feature matrix into a well-trained battery health state estimation model to obtain the health state of the retired power battery for re-utilization.

[0052] like Figure 6 As shown, based on the above retired power battery health status prediction method, the present invention also provides an electronic device, which can be a computing electronic device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The electronic device includes a processor 610, a memory 620, and a display 630. Figure 6 Only some components of the electronic device are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0053] In some embodiments, the memory 620 may be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device. In other embodiments, the memory 620 may also be an external storage electronic device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 620 may also include both an internal storage unit of the electronic device and an external storage electronic device. The memory 620 is used to store application software and various types of data installed in the electronic device, such as program codes installed in the electronic device. The memory 620 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a retired power battery health state prediction program 640 is stored on the memory 620, and the retired power battery health state prediction program 640 can be executed by the processor 610, thereby realizing the retired power battery health state prediction method of each embodiment of the present application.

[0054] In some embodiments, the processor 610 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 620, such as executing a retired power battery health status prediction method.

[0055] In some embodiments, the display 630 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 630 is used to display information of the retired power battery health status prediction electronic device and to display a visual user interface. The components 610-630 of the electronic device communicate with each other through a system bus.

[0056] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0057] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting the health status of retired power batteries, characterized in that: include: The obtained battery parameters are divided into different levels to obtain battery parameters of multiple different levels; Using battery numbers and battery parameter level values ​​as nodes and battery parameter names as edges, a graph network structure with timestamps is constructed; And according to the graph network structure, a time-based battery status feature matrix is ​​constructed; The time-based battery state feature matrix is ​​input into the well-trained battery health state estimation model to obtain the health state of the retired power battery for reutilization.

2. The method for predicting the health status of retired power batteries according to claim 1, characterized in that: The method uses the battery number and the battery parameter level value as nodes and the battery parameter name as edges to construct a graph network structure with a timestamp, including: Construct the initial graph network structure and define the head entity node and the tail entity node; A static graph network structure at a certain moment is constructed with the battery number as the head entity node, the battery parameter level value as the tail entity node, and the battery parameter name as the edge; Based on the static graph network structure, the correlation between the battery status and the battery parameter level value and the battery parameters is determined.

3. The method for predicting the health status of retired power batteries according to claim 2, characterized in that: The determining of the association between the battery state, the battery parameter level value and the battery parameter based on the static graph network structure includes: Based on the static graph network structure, a four-tuple model of battery status, battery status value, battery number and time is constructed; Based on the four-tuple model, the time sequence relationship between the battery state value and the battery parameter level value is determined according to the battery number.

4. The method for predicting the health status of retired power batteries according to claim 2, characterized in that: The method of constructing a time-based battery state feature matrix according to the graph network structure includes: Using a distance-based translation model TransE, determining a feature coding value of each entity node in the graph network structure, and constructing an initial vector according to the feature coding value; According to the initial vector of each entity node, a battery state feature matrix is ​​constructed.

5. The method for predicting the health status of retired power batteries according to claim 4, characterized in that: The distance-based translation model TransE is adopted to determine the feature coding value of each entity node in the graph network structure, and the initial vector is constructed according to the feature coding value, including: Dividing the graph network structure into a head entity node set, a tail entity node set and an attribute edge set; Based on the equivalence relationship between the sum of the head entity initial vector and the attribute edge initial vector and the tail entity initial vector, the initial distance between the head entity initial vector and the tail entity initial vector is determined according to a preset Euclidean distance scoring function; The initial distance is optimized according to the minimum objective function to determine the initial distance of each entity.

6. The method for predicting the health status of retired power batteries according to claim 5, characterized in that: The preset Euclidean distance scoring function is expressed by the following formula: in, represents the initial distance, Indicates the head entity node, represents the attribute edge, t represents the tail entity node, and Represent the first distance and the second distance respectively.

7. The method for predicting the health status of retired power batteries according to claim 1, characterized in that: The fully trained battery health state estimation model includes at least one of a long short-term memory recurrent network model, a deep learning model and a support vector machine.

8. A device for predicting the health status of retired power batteries, characterized in that: include: A classification module is used to classify the acquired battery parameters to obtain battery parameters of multiple different levels; A graph network structure construction module, which is used to construct a graph network structure with timestamps by using battery numbers and battery parameter level values ​​as nodes and battery parameter names as edges; A feature matrix construction module is used to construct a time-based battery state feature matrix according to a graph network structure; The health state determination module is used to input the time-based battery state feature matrix into the well-trained battery health state estimation model to obtain the health state of the retired power battery for re-utilization.

9. An electronic device, characterized in that: include: Processor and memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the method for predicting the health status of retired power batteries as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the retired power battery health status prediction method as described in any one of claims 1-7.

Images