Battery Pack Health State Prediction Method, Device, Equipment, Medium and Program Product

Abstract

The present disclosure provides a method, device, equipment, medium and program product for predicting the health state of a battery pack, relating to the technical field of data processing. The method includes: obtaining the battery pack temperature data, battery pack current data, and battery pack voltage data of at least two battery packs within a preset time period; calculating the temperature difference of the battery pack at each time point based on the battery pack temperature data, and calculating the temperature similarity between battery packs accordingly; calculating the current similarity between battery packs based on the frequency domain representation of the battery pack current data; calculating the covariance matrix of voltage changes based on the battery pack voltage data, and calculating the voltage similarity between battery packs accordingly; constructing a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph based on the temperature similarity, current similarity, and voltage similarity; predicting the health state of the battery pack through a preset health state prediction model based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph. Using the present disclosure can improve the accuracy of the SOH prediction result.

Description

Technical Field

[0001] The present disclosure relates to the technical field of data processing, and particularly to a method, device, equipment, medium and program product for predicting the state of health of a battery pack. Background Art

[0002] In the related art, the prediction of the state of health (SOH) of a battery pack is crucial for the efficient operation of an energy storage power station. At present, the prediction of the SOH of a battery pack mostly relies on traditional physical models or statistical models. Among them, physical models are usually constructed based on the electrochemical principles and physical properties of batteries. Since the chemical reactions and physical properties inside the battery are difficult to be described and simulated by accurate physical equations, the constructed physical models are relatively simplified, thus affecting the prediction accuracy of the SOH of the battery pack. Statistical models usually analyze historical data to find the correlation rules between the battery health state and various factors. However, due to the characteristics of high dimensionality, non-linearity, strong noise, etc. of the data generated during the operation of the battery pack, statistical models have obvious deficiencies in processing such complex data, resulting in problems of low accuracy when dealing with complex data. Summary of the Invention

[0003] The present disclosure provides a method, device, equipment, medium and program product for predicting the state of health of a battery pack. The technical solution of the present disclosure is as follows:

[0004] In a first aspect, the present disclosure provides a method for predicting the state of health of a battery pack, including:

[0005] Obtaining the battery pack operation data of at least two battery packs within a preset time period; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series;

[0006] Based on the battery pack temperature data, calculating the temperature difference of the battery pack at each time point, and based on the temperature difference at each time point, calculating the temperature similarity between the battery packs;

[0007] Based on the frequency domain representation of the battery pack current data, calculating the current similarity between the battery packs;

[0008] Based on the battery pack voltage data, calculating the covariance matrix of the voltage change, and based on the covariance matrix, calculating the voltage similarity between the battery packs;

[0009] Based on the temperature similarity, current similarity, and voltage similarity between the battery packs, constructing a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph;

[0010] Based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph, predict the health state of each battery pack through a preset health state prediction model; wherein, the preset health state prediction model is constructed based on a graph convolutional network.

[0011] In a possible implementation manner, calculating the current similarity between the battery packs based on the frequency domain representation of the battery pack current data includes:

[0012] Perform Fourier transform on the battery pack current data of each battery pack to obtain the frequency domain representation of the battery pack current data of each battery pack;

[0013] Based on the frequency domain representation of the battery pack current data of each battery pack, calculate the current similarity between the battery packs;

[0014] Calculating the covariance matrix of the voltage change based on the battery pack voltage data includes:

[0015] Based on the battery pack voltage data of each battery pack at each time point, calculate the covariance matrix of the voltage change.

[0016] In a possible implementation manner, when the battery pack includes battery pack i and battery pack j in the case of ， The calculation formula for the temperature similarity between the battery packs is:

[0017]

[0018] Wherein, α The weighting factor of the temperature difference, n is an exponential parameter that adjusts the influence of temperature change on the adjacency relationship, △T ij (t ) is the temperature difference of the battery pack at each time point t , T i (t), T j (t) respectively represent the battery packs i and battery pack j at the time point t temperature, T is the length of the time series, S temp (i, j) is the temperature similarity between the battery packs, i represents the battery pack i , j represents the battery pack j .

[0019] In a possible implementation, the calculation formula for the current similarity between the battery packs is:

[0020]

[0021] Wherein, I i (f) represents the frequency-domain current signal of battery pack i and I j (f) represents the frequency-domain current signal of battery pack j and I i (t) represents the frequency-domain current signal of battery pack i at time point t and i represents battery pack i , j represents battery pack j , S current (i, j) represents the current similarity between the battery packs.

[0022] In a possible implementation, the calculation formula for the voltage similarity between the battery packs is:

[0023]

[0024] Wherein, and are respectively the average values of the voltages of battery pack i and battery pack j , V i (t) is the voltage of battery pack i at time point t , V j (t) is the voltage of battery pack j at time point t , T is the length of the time series, represents battery pack i and battery pack j represents the voltage similarity between them.

[0025] In a possible implementation, predicting the battery pack health state of each battery pack based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph through a preset health state prediction model includes:

[0026] Input the temperature adjacency graph, current adjacency graph, and voltage adjacency graph into the preset health state prediction model, and process the temperature adjacency graph, current adjacency graph, and voltage adjacency graph through the preset health state prediction model to obtain the graph convolution processing results of the temperature adjacency graph, current adjacency graph, and voltage adjacency graph;

[0027] Based on a time series model, process the graph convolution processing results to obtain the long-term and short-term dependencies and long-term time series patterns of the battery capacity change of the battery pack; wherein, the time series model includes a long short-term memory network and a Transformer;

[0028] Predict the health state of each battery pack based on the long-term dependencies and long-term time series patterns of the battery capacity change of the battery pack.

[0029] In a second aspect, the present disclosure provides a device for predicting the health state of a battery pack, including:

[0030] A data acquisition module, configured to acquire the battery pack operation data of at least two battery packs within a preset time period; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series;

[0031] A temperature similarity calculation module, configured to calculate the temperature difference of the battery packs at each time point based on the battery pack temperature data, and calculate the temperature similarity between the battery packs based on the temperature difference at each time point;

[0032] A current similarity calculation module, configured to calculate the current similarity between the battery packs based on the frequency domain representation of the battery pack current data;

[0033] A voltage similarity calculation module, configured to calculate the covariance matrix of the voltage change based on the battery pack voltage data, and calculate the voltage similarity between the battery packs based on the covariance matrix;

[0034] A graph construction module, configured to construct a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph based on the temperature similarity, current similarity, and voltage similarity between the battery packs;

[0035] A prediction module, configured to predict the health state of each battery pack through a preset health state prediction model based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph; wherein, the preset health state prediction model is constructed based on a graph convolutional network.

[0036] In a third aspect, the present disclosure provides an electronic device, including:

[0037] A processor;

[0038] A memory for storing the processor-executable instructions;

[0039] Wherein, the processor is configured to execute the instructions to implement the battery pack state of health prediction method described in the first aspect.

[0040] In a fourth aspect, the present disclosure provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the battery pack state of health prediction method described in the first aspect is implemented.

[0041] In a fifth aspect, the present disclosure provides a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the battery pack state of health prediction method described in the first aspect is implemented.

[0042] The technical solution of the present disclosure at least brings the following beneficial effects:

[0043] In the embodiments of the present disclosure, by obtaining the battery pack operation data of at least two battery packs within a preset time period; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series; based on the battery pack temperature data, calculating the temperature difference of the battery pack at each time point, and based on the temperature difference at each time point, calculating the temperature similarity between the battery packs; based on the frequency domain representation of the battery pack current data, calculating the current similarity between the battery packs; based on the battery pack voltage data, calculating the covariance matrix of the voltage change, and based on the covariance matrix, calculating the voltage similarity between the battery packs; based on the temperature similarity, current similarity, and voltage similarity between the battery packs, constructing a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph; through a preset state of health prediction model, based on the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph, predicting the state of health of each battery pack; wherein, the preset state of health prediction model is constructed based on a graph convolutional network. In this way, by capturing the multi-dimensional correlation features (temperature, current, voltage) between the battery packs, constructing multiple graphs (temperature adjacency graph, current adjacency graph, voltage adjacency graph, etc.), and combining the graph convolutional network to predict the SOH of the battery pack. In this way, the multi-dimensional spatio-temporal characteristics of the battery pack can be fused, the mutual influence between the battery packs and the multi-dimensional characteristics of the battery pack can be considered more comprehensively, and the accurate prediction of the SOH of the battery pack can be realized; at the same time, by combining the graph convolutional network to predict the SOH of the battery pack, the spatial dependence relationship and time dependence relationship of the battery pack can be better processed, so that the SOH prediction result is more accurate.

[0044] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure, and do not constitute an undue limitation of the present disclosure.

[0046] Figure 1 A flowchart of a method for predicting the health state of a battery pack provided by an embodiment of the present disclosure;

[0047] Figure 2 A flowchart of another method for predicting the health state of a battery pack provided by an embodiment of the present disclosure;

[0048] Figure 3 A structural diagram of a device for predicting the health state of a battery pack provided by an embodiment of the present disclosure;

[0049] Figure 4 A structural diagram of an electronic device provided by an embodiment of the present disclosure. Detailed implementation manners

[0050] In order to enable those of ordinary skill in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0051] It should be noted that in the embodiments of the present disclosure, there may be some existing solutions in the industry such as certain software, components, models, etc. They should be considered exemplary. Their purpose is only to illustrate the feasibility in the implementation of the technical solutions of the present disclosure, but it does not mean that the applicant has already or necessarily used this solution.

[0052] In the related art, the prediction of the health state of a battery pack is crucial for the efficient operation of an energy storage power station. The capacity of the battery pack is the core indicator of battery health. As the battery is used, the capacity gradually decays, and the state of health (SOH) of the battery pack is usually measured by the change in capacity. With the changes in factors such as the charge-discharge cycle times, temperature, current, and voltage of the battery, the health state of the battery will show complex fluctuations, which makes the prediction of SOH challenging. Currently, most of the prediction methods for the SOH of battery packs rely on traditional physical models or statistical models, and these methods may have problems of low accuracy and poor generalization ability when dealing with complex data. In the related art, deep learning and machine learning methods have also been widely applied to the field of battery health prediction. In particular, methods such as the long short-term memory network (LSTM) and convolutional neural network (CNN) based on time series data have achieved certain results. However, these methods usually focus on single features or information in the time dimension and ignore the spatial correlation between battery packs and the comprehensive influence of multi-dimensional features. Especially in an energy storage power station, since the battery packs are physically related, their health states are affected not only by their own characteristics but also by the interaction with other battery packs. Therefore, traditional models are difficult to capture the spatial relationship between battery packs, thus affecting the accuracy and reliability of SOH prediction.

[0053] To this end, the embodiments of the present disclosure provide a method, device, equipment, medium and program product for predicting the state of health of a battery pack. The method for predicting the state of health of a battery pack provided by the embodiments of the present disclosure is a method for predicting the state of health of a battery pack that integrates multi-dimensional spatio-temporal characteristics and multi-model integration. It can comprehensively consider the spatio-temporal characteristics of the battery pack, and by capturing multi-dimensional correlation features (such as capacity, temperature, current, voltage, etc.) between battery packs, achieve accurate prediction of the state of health (SOH) of the battery pack. Exemplarily, the method for predicting the state of health of a battery pack provided by the embodiments of the present disclosure can predict the SOH of the battery pack by constructing multiple graphs (such as a temperature adjacency graph, a current adjacency graph, a voltage adjacency graph, etc.) and combining a graph convolutional network (Graph Convolutional Network, GCN). In this way, it can more comprehensively consider the mutual influence between battery packs and the multi-dimensional characteristics of the battery pack; moreover, it is also an SOH prediction method that combines a multi-graph convolutional network with a time series model LSTM and a Transformer model architecture. By combining GCN, LSTM and Transformer to process the spatial and temporal dependencies of the battery pack, the SOH prediction result can be made more accurate. At the same time, the method for predicting the SOH of the entire battery pack that comprehensively considers the mutual influence between battery packs can optimize the SOH prediction of multiple battery packs as a whole, consider the collaborative working state and mutual influence of each battery pack in the energy storage power station, and provide a more accurate overall state of health prediction.

[0054] The technical solutions provided by the embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0055] Figure 1 The following is a flowchart of a method for predicting the state of health of a battery pack provided by an embodiment of the present disclosure. This method can be applied to a server, such as a single server or a server cluster. As Figure 1 shown, the method for predicting the state of health of a battery pack may include the following steps:

[0056] S101, obtain the battery pack operation data of at least two battery packs within a preset time period.

[0057] Among them, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series.

[0058] In an embodiment of the present disclosure, when predicting the state of health of a battery pack, the operating data of the battery pack within a preset time period (e.g., a past period of time) can be obtained first. Exemplarily, the operating data of at least two battery packs within the preset time period can be obtained, and the operating data of each battery pack can include data such as battery pack temperature data, battery pack current data, and battery pack voltage data. Moreover, for each type of battery pack operating data, it can be represented in the form of a dynamic time series. As a specific example, considering that in the actual operation of an energy storage power station, the operating parameters of the battery pack have significant dynamic characteristics and multi-dimensional influences, the battery pack operating data can be processed into the following form:

[0059] Capacity time series: The historical capacity change of each battery pack C i (t) , where t is time, i is the number of the battery pack.

[0060] Environmental and operating conditions characteristics: including the battery pack temperature data, battery pack current data, and battery pack voltage data of the battery pack, which are respectively represented as the temperature T i (t) 、current I i (t) 、voltage V i (t) and other time series.

[0061] Each type of data is represented as a dynamic time series, as follows:

[0062]

[0063] where X i is the input feature matrix of the battery pack i , which contains multi-dimensional data of time steps t 1 ，t 2 ，……，t T .

[0064] S102. Calculate the temperature difference of the battery pack at each time point based on the battery pack temperature data, and calculate the temperature similarity between the battery packs based on the temperature difference at each time point.

[0065] In an embodiment of the present disclosure, after obtaining the battery pack operation data of the battery pack within a preset time period, the similarity between battery packs can be calculated, including temperature similarity. Exemplarily, based on the obtained battery pack temperature data, the temperature difference of the battery pack at each time point can be calculated. As an example, considering that temperature is a key factor affecting battery performance, different battery packs may experience similar degradation patterns under similar temperature conditions. To accurately capture the temperature correlation between battery packs, a high-order similarity metric based on temperature change can be adopted. Based on this, when determining the temperature similarity between battery packs, the high-order temperature difference can be defined as follows first, and the non-linear relationship of the temperature difference is processed through a weighted exponential function to calculate the temperature difference (high-order temperature difference) of the battery pack at each time point △T ij (t) .

[0066]

[0067] Wherein, is the weighting factor of the temperature difference, n is the exponential parameter that adjusts the influence of temperature change on the adjacency relationship, controls the attenuation speed of the temperature difference on the similarity, and the battery pack i and j at the time point t The temperatures are and .

[0068] After obtaining the temperature difference at each time point, based on the temperature difference at each time point, the weighted similarity between the battery packs i and j can be calculated to obtain the overall temperature similarity between the battery packs i and j as follows:

[0069]

[0070] Wherein, α is the weighting factor of the temperature difference, n is the exponential parameter that adjusts the influence of temperature change on the adjacency relationship, △T ij (t ) is the temperature difference of the battery pack at each time point t , T i (t), T j (t) respectively represent the battery packs i and the battery pack j at the time point t The temperature of, Tis the length of the time series, S temp (i, j) is the temperature similarity between battery packs, i represents battery pack i , j represents battery pack j .

[0071] It can be understood that the temperature similarity between battery packs can be used to evaluate the similarity between the temperature data of different battery packs.

[0072] S103. Calculate the current similarity between battery packs based on the frequency-domain representation of the battery pack current data.

[0073] In an embodiment of the present disclosure, the current similarity between battery packs can also be calculated. Considering that during the charging and discharging process of the battery, the current fluctuation affects the battery health state. In order to accurately capture the similarity of the current fluctuations between battery packs, a dynamic frequency-domain similarity metric can be adopted to effectively reflect the periodic fluctuations of the current between battery packs. Exemplarily, the frequency-domain feature extraction of the battery pack current data can be performed first, and then the current similarity between battery packs can be calculated based on the frequency-domain representation of each battery pack current signal. It can be understood that the current similarity between battery packs can be used to evaluate the similarity between the current data of different battery packs.

[0074] S104. Calculate the covariance matrix of the voltage change based on the battery pack voltage data, and calculate the voltage similarity between battery packs based on the covariance matrix.

[0075] In an embodiment of the present disclosure, the voltage similarity between battery packs can also be calculated. Considering that the voltage fluctuation of the battery directly affects the working efficiency and capacity of the battery. In order to fully consider the correlation of the voltage changes between battery packs, a voltage change modeling based on a high-order covariance matrix can be adopted to calculate the voltage similarity. Exemplarily, the covariance matrix characterizing the voltage change of the battery pack can be calculated first based on the battery pack voltage data of the battery pack. Then, the voltage similarity between battery packs can be calculated based on the covariance matrix characterizing the voltage change of the battery pack. It can be understood that the voltage similarity between battery packs can be used to evaluate the similarity between the voltage data of different battery packs. As an example, taking the example of including battery pack 1 and battery pack 2, the temperature similarity between battery pack 1 and battery pack 2 can be calculated, the current similarity between battery pack 1 and battery pack 2 can be calculated, and the voltage similarity between battery pack 1 and battery pack 2 can be calculated.

[0076] S105. Construct a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph based on the temperature similarity, current similarity, and voltage similarity between battery packs.

[0077] In an embodiment of the present disclosure, after calculating the temperature similarity, current similarity, and voltage similarity between battery packs, multiple adjacency graphs can be constructed based on these similarity evaluation results between battery packs (including the temperature similarity, current similarity, and voltage similarity between battery packs), with each adjacency graph corresponding to a feature (such as temperature similarity, current similarity, voltage similarity) to capture the multi-dimensional relationships between battery packs. In this way, the spatial dependence, temporal dependence, and steady-state dependence between battery packs can be comprehensively modeled at different levels. Exemplarily, a temperature adjacency graph associated with the thermal environment space can be constructed based on the temperature similarity between battery packs. For example, the quantization result of the battery pack temperature similarity can be mapped to the adjacency graph structure to capture the thermal environment dependence relationship between battery packs, and the temperature adjacency matrix (temperature adjacency graph) is defined as follows:

[0078]

[0079] Wherein, S temp (i, j) is the temperature similarity between battery packs i, j and

[0080] Exemplarily, the dynamic similarity of current fluctuations (current similarity) can be utilized to reflect the dynamic dependence characteristics of battery packs during the charging and discharging process. The elements of the current adjacency matrix (current adjacency graph) are defined as follows:

[0081]

[0082] Wherein, is the current similarity calculated through the frequency domain similarity metric, is the time alignment error of the current fluctuations between battery packs measured through dynamic time warping (DTW).

[0083] Exemplarily, through the covariance quantization of the steady-state voltage characteristics, that is, a long-term steady-state association graph between battery packs can be constructed based on the voltage similarity. The elements of the voltage adjacency matrix (voltage adjacency graph) are defined as:

[0084]

[0085] Wherein, represents the voltage similarity between battery pack i and battery pack j and

[0086] In this way, by designing multi-dimensional similarity metrics based on temperature similarity, current similarity, and voltage similarity, the spatial dependence, temporal dependence, and steady-state dependence between battery packs can be comprehensively modeled at different levels.

[0087] S106. By presetting a health state prediction model, based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph, predict the battery pack health state of each battery pack.

[0088] Among them, the preset health state prediction model is constructed based on a graph convolutional network.

[0089] In an embodiment of the present disclosure, after obtaining the temperature adjacency graph, current adjacency graph, and voltage adjacency graph, the battery pack health state of the battery pack can be predicted through a preset health state prediction model. Exemplarily, the temperature adjacency graph, current adjacency graph, and voltage adjacency graph can be input into the preset health state prediction model, and the preset health state prediction model processes the temperature adjacency graph, current adjacency graph, and voltage adjacency graph to predict the battery pack health state of each battery pack. As an example, the preset health state prediction model can be pre-constructed based on a graph convolutional network. The graph convolutional network can process graph-structured data, extract features by learning the connection patterns between nodes, and thus perform node-level predictions. In the prediction of the battery pack health state, the battery pack can be regarded as a node in the graph, and parameters such as temperature, current, and voltage can be used as the features of the node to predict the battery pack health state. In this way, based on the graph convolutional network, the complex relationships inside the battery pack can be fully utilized, thereby improving the accuracy and reliability of the health state prediction.

[0090] In an embodiment of the present disclosure, battery pack operation data of at least two battery packs within a preset time period is obtained; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series; based on the battery pack temperature data, battery pack current data, and battery pack voltage data of each battery pack, the temperature similarity, current similarity, and voltage similarity between the battery packs are respectively determined; based on the temperature similarity, current similarity, and voltage similarity between the battery packs, a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph are constructed; through a preset health state prediction model, based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph, the battery pack health state of each battery pack is predicted; wherein, the preset health state prediction model is constructed based on a graph convolutional network. In this way, by capturing multi-dimensional correlation features (temperature, current, voltage) between battery packs, multiple graphs (temperature adjacency graph, current adjacency graph, voltage adjacency graph, etc.) are constructed, and the state of health (SOH) of the battery packs is predicted in combination with the graph convolutional network. In this way, the multi-dimensional spatio-temporal characteristics of the battery packs can be fused, the mutual influence between the battery packs and the multi-dimensional characteristics of the battery packs can be considered more comprehensively, and accurate prediction of the SOH of the battery packs can be realized; at the same time, the SOH of the battery packs is predicted in combination with the graph convolutional network, and the spatial dependence relationship and time dependence relationship of the battery packs are better processed, so that the SOH prediction result is more accurate.

[0091] In some possible implementation manners, calculating the current similarity between battery packs based on the frequency-domain representation of the battery pack current data includes:

[0092] Performing a Fourier transform on the battery pack current data of each battery pack to obtain the frequency-domain representation of the battery pack current data of each battery pack;

[0093] Based on the frequency-domain representation of the battery pack current data of each battery pack, calculating the current similarity between the battery packs;

[0094] Calculating the covariance matrix of voltage changes based on the battery pack voltage data includes:

[0095] Based on the battery pack voltage data of each battery pack at each time point, calculating the covariance matrix of voltage changes.

[0096] In an embodiment of the present disclosure, when calculating the current similarity between battery packs based on the frequency-domain representation of the battery pack current data, frequency-domain feature extraction of current fluctuations can be performed first. First, perform a Fourier transform on the current time series of the battery pack current data of the battery pack to obtain the frequency-domain representation of each battery pack current signal, as follows:

[0097]

[0098] Battery pack iand j The current similarity (current fluctuation similarity) between can be represented by the cross - correlation function in the frequency domain, defined as follows:

[0099]

[0100] where, I i (f) represents the frequency - domain current signal of the battery pack i ; I j (f) represents the frequency - domain current signal of the battery pack j ; I i (t) represents the frequency - domain current signal of the battery pack i at the time point t ; i represents the battery pack i , j represents the battery pack j , S current (i, j) represents the current similarity between battery packs.

[0101] When calculating the covariance matrix of voltage changes based on battery - pack voltage data, it can be assumed that the voltages of battery packs i and j at the time point t are and respectively. The covariance matrix of voltage changes can be calculated, and the correlation of voltage changes between battery packs is reflected by the high - order covariance matrix. The covariance matrix of voltage changes is defined as follows:

[0102]

[0103] where, and are the mean values of the voltages of battery packs i and battery pack j respectively, V i (t) is the voltage of battery pack i at the time point t , V j (t) is the voltage of battery pack j at the time point t , T is the length of the time series.

[0104] Then, the voltage similarity between battery packs can be defined as follows:

[0105]

[0106] Wherein, and are the average voltages of battery pack i and battery pack j respectively, V i (t) is the voltage of battery pack i at time point t ; V j (t) is the voltage of battery pack j at time point t ; T is the length of the time series, represents the voltage similarity between battery pack i and battery pack j .

[0107] In a possible implementation, through a preset health state prediction model, based on the temperature adjacency graph, current adjacency graph, and voltage adjacency graph, the battery pack health state of each battery pack is predicted, including:

[0108] Input the temperature adjacency graph, current adjacency graph, and voltage adjacency graph into the preset health state prediction model, and process the temperature adjacency graph, current adjacency graph, and voltage adjacency graph through the preset health state prediction model to obtain the graph convolution processing results of the temperature adjacency graph, current adjacency graph, and voltage adjacency graph;

[0109] Process the graph convolution processing results based on the time series model to obtain the long - short - term dependence and long - term time series pattern of the battery capacity change of the battery pack; wherein, the time series model includes a long - short - term memory network and a Transformer;

[0110] Predict the battery pack health state of each battery pack based on the long - term dependence and long - term time series pattern of the battery capacity change of the battery pack.

[0111] In an embodiment of the present disclosure, when predicting the battery pack health state of each battery pack through a preset health state prediction model, a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph can be input into the preset health state prediction model, and these graphs are processed through a graph convolutional network to obtain a graph convolutional processing result. The graph convolutional network can process graph-structured data, extract features by learning the connection patterns between nodes, and thus perform node-level prediction. Then, the graph convolutional processing result can be input into a time series model, and the graph convolutional processing result is processed based on the time series model to obtain the long-term and short-term dependencies and long-term time series patterns of the battery capacity change of the battery pack. Exemplarily, the time series model includes a long short-term memory network (LSTM) and a Transformer. These models can effectively capture short-term patterns and long-term dependence problems in time series data. For example, the long short-term memory network can control the flow of information through a gating mechanism (input gate, forget gate, output gate), model local time series patterns, and capture the short-term dynamic changes of the sequence. After that, based on the long-term dependencies and long-term time series patterns of the battery capacity change of the battery pack, the health state of each battery pack can be predicted. For example, the output of the time series model can be used for a classification or regression task, and machine learning or deep learning algorithms can be used to predict the battery pack health state.

[0112] It can be understood that, first, the embodiment of the present disclosure can construct a battery health state (SOH) prediction model based on a multi-graph convolutional network (GCN). Through layer-by-layer information transfer, graph convolutional operations, and multi-graph feature fusion, the complex spatial relationship between battery packs can be effectively learned, and key features helpful for SOH prediction can be extracted from multi-dimensional features. The steps include graph convolutional operations, fusion of multiple graphs, and non-linear activation and other processes. Specifically, the specific implementation of this process can include:

[0113] 1. Graph convolutional operation and dynamic update of node features: Graph convolutional operation is the core of GCN. By propagating node information in the graph structure, the features of each node are updated. For each node of the battery pack , its feature will be updated according to the features of neighboring nodes. Specifically, it includes:

[0114] 1.1 Processing of the adjacency matrix:

[0115] Suppose there is an adjacency graph of a battery pack G = (V, E) , where V is the set of battery pack nodes and E is the set of edges between nodes. The dependency relationship between nodes is represented by the adjacency matrix A . In the graph convolutional operation, the role of the adjacency matrix is to define the relationship between nodes and use its weights for feature update.

[0116] 1.2 Multidimensional graph convolutional update formula for fusing heterogeneous features:

[0117] The graph convolution operation updates by aggregating and weighting the features of neighboring nodes. Suppose the feature of node i is , and the feature of neighboring node j is . Then, the weighted aggregation method can be used to update the node feature as follows:

[0118]

[0119] where is the updated feature of node i at the l -th layer, is an element of the adjacency matrix, representing the connection weight between node i and node j , and are the degrees of nodes i and j respectively (i.e., the number of adjacent nodes), is the weight matrix of the l -th layer, is the bias term, is the non-linear activation function, represents the set of all nodes directly connected to node i .

[0120] 2. Multi-graph convolution feature fusion and global dependence enhancement mechanism: In the prediction of the state of health of a battery pack, different adjacency graphs represent different dependence relationships between battery packs (such as temperature adjacency graph, current adjacency graph, voltage adjacency graph, etc.). To comprehensively consider these multi-dimensional features, the graph convolution layers can be fused by fusing the results of multiple graph convolution operations to improve the prediction ability of the model as follows:

[0121]

[0122] where is the element of the adjacency matrix at the l-th layer, l = 1: (temperature adjacency matrix); l = 2: (current adjacency matrix); l = 3: (voltage adjacency matrix).

[0123] Then, considering that the SOH of the battery pack has strong time dependence, especially the periodic fluctuations during the capacity attenuation process. Therefore, the long short-term memory network (LSTM) and the Transformer time series model can be combined to construct a time series modeling and adaptive optimization mechanism to further improve the accuracy of SOH prediction. Specifically, the specific implementation of this process can include:

[0124] 1. Long-Term Temporal Dependency Modeling and Dynamic Learning Based on LSTM: LSTM can effectively capture the long-term dependencies of battery capacity changes and learn the dynamic characteristics of the health state of the battery pack over time. The key update steps of LSTM are as follows:

[0125]

[0126]

[0127]

[0128]

[0129] Among them, is the input gate, is the forget gate, is the output gate, is the cell state, is the hidden state. is the input feature at time t (i.e., the node feature output by GCN), is the hidden state of the previous time step.

[0130] 2. Long-Term Temporal Pattern Recognition and Adaptive Attention Mechanism Based on Transformer: By using a biased cosine similarity and a weighted fusion strategy, the self-attention mechanism becomes more flexible and takes into account the dynamic weights of different features.

[0131] Introduce the attention weights of the weighted kernel function and the bias term:

[0132]

[0133]

[0134] Among them, is the learned bias term, representing the preference for certain elements. At this time, the attention matrix will take into account this additional learned information to adjust the similarity scores of different features. The symbol ⊙ represents element-wise multiplication, which can make the attention mechanism not only a weighted sum but also emphasize which features are more important and which parts of the information may be "ignored" or have higher weights.

[0135] For the multi-head attention mechanism, considering that each head may capture different health patterns of the battery under different operating conditions (such as voltage, current, temperature, etc.), a "weighted fusion" strategy can be introduced to dynamically adjust the contribution of each head in order to better capture the temporal dependencies of the battery in different states. The weighted fusion method of the multi-head self-attention mechanism is as follows:

[0136]

[0137] Among them, is the learnable weight of each attention head, indicating the contribution of that head in the final calculation, h is the number of heads.

[0138] Secondly, spatio-temporal data fusion and multi-model integration can be performed to predict the state of health of the battery pack: Based on the spatial features extracted by the graph convolutional network, the temporal features of the battery capacity are further processed by combining the LSTM and Transformer modules to improve the adaptability of SOH prediction to time changes. Specifically, the specific implementation of this process can include:

[0139] 1. GCN feature space extraction: The input graph data is processed through the GCN network to obtain the graph structure feature representation of the nodes. The GCN layer captures the spatial relationships between the nodes.

[0140]

[0141] Among them, X is the node feature, A is the adjacency matrix, is the node feature representation output by the GCN.

[0142] 2. Parallel input of LSTM and Transformer and multi-level temporal feature fusion: The LSTM captures the long-term and short-term dependencies between time steps, and the Transformer captures the longer-range dependencies through the self-attention mechanism, as follows:

[0143]

[0144]

[0145] Among them, H LSTM represents the long-term and short-term dependencies, H Transformer represents the long-term temporal pattern.

[0146] 3. Adaptive spatio-temporal feature fusion and weighted decision-making mechanism: After the parallel operation of the LSTM and the Transformer, the outputs of the two can be fused. Through a weighted fusion mechanism, the outputs of the two are weighted and combined through a non-linear transformation, as follows:

[0147]

[0148] Among them, H LSTM and H Transformeris the output from LSTM and Transformer, is the learnable weight matrix, is the bias term.

[0149] This approach can provide a more flexible non - linear transformation for the final output, allowing the model to dynamically adjust the contributions of LSTM and Transformer at each time step.

[0150] Finally, SOH prediction and calculation are performed, including:

[0151] Based on the fused features from the trained model H final , the current capacity of each battery pack is predicted. The SOH of a battery pack is usually calculated by the ratio of its current capacity to the initial capacity, as follows:

[0152]

[0153] where, is the predicted capacity of the battery pack, is the initial capacity of the battery pack.

[0154] Based on this, the battery pack state of health prediction method provided by the embodiments of the present disclosure uses a graph convolutional network (GCN) to simultaneously process graph-structured data from multiple dimensions (such as temperature, current, voltage, etc.), which can effectively capture the spatial dependence relationships between battery packs. By constructing multiple adjacency graphs to represent different mutual relationships between battery packs, the prediction ability for complex battery systems is improved. The battery pack state of health prediction method provided by the embodiments of the present disclosure comprehensively considers multi-dimensional features of the battery pack (such as capacity, current, voltage, temperature, etc.). Based on the deep modeling of the battery pack state of health with spatio-temporal fusion, the spatial relationships between battery packs are learned through a graph convolutional network, and at the same time, the temporal features of the battery capacity change are captured through temporal modeling. This way of fusing multi-dimensional features can improve the comprehensiveness, accuracy, and robustness of SOH prediction. The battery pack state of health prediction method provided by the embodiments of the present disclosure also proposes an adaptive learning framework based on multi-layer feature representation, which can dynamically adjust feature weights and convolutional operation parameters at different levels to enhance the model's understanding ability of the battery pack state of health (SOH). Thus, not only can the local features of the battery pack (such as the relationships between adjacent nodes) be captured, but also a unified description of the battery pack state of health can be constructed through global feature integration. Moreover, the battery pack state of health prediction method provided by the embodiments of the present disclosure also considers the periodic volatility of the battery capacity. By analyzing the change rules of the battery capacity and further optimizing the model in combination with temporal data, the model can more accurately predict the periodic change of the battery capacity, thereby improving the SOH prediction accuracy. Further, by accurately predicting the SOH of the battery pack, it can also help the energy storage power station to implement precise battery maintenance and replacement strategies, thereby significantly reducing the operation and maintenance costs, timely discovering battery health problems, adopting optimized charge and discharge strategies, and improving the benefits of the energy storage system.

[0155] To make the battery pack state of health prediction method provided by the embodiments of the present disclosure clearer, the following is described in conjunction with the attached Figure 2 figures. As Figure 2 shown, the battery pack state of health prediction method includes:

[0156] 1. Construct a data set, including obtaining the battery pack operation data of at least two battery packs within a preset time period and constructing a data set; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data.

[0157] 2. Construct multi-dimensional features, including processing the data set into a form represented by a dynamic time series.

[0158] 3. Construction of quantifying the spatial dependence relationship between battery packs based on feature similarity, including constructing temperature similarity, current similarity, and voltage similarity between battery packs.

[0159] 4. Construction of spatial, dynamic, and steady-state adjacency graph structures, including the construction of temperature adjacency graphs, current adjacency graphs, and voltage adjacency graphs.

[0160] 5. Multi-graph convolutional network (GCN) modeling and feature fusion, including constructing a state of health (SOH) prediction model for batteries based on a multi-graph convolutional network, learning the complex spatial relationships between battery packs through layer-by-layer information transfer, graph convolutional operations, and multi-graph feature fusion, and extracting key features helpful for SOH prediction from multi-dimensional features. It includes graph convolutional operations, fusion of multiple graphs, and non-linear activation and other processes.

[0161] 6. Long-term time series dependence modeling and dynamic learning based on LSTM, including capturing the long-term and short-term dependencies of battery capacity changes and learning the dynamic features of the health state of battery packs over time.

[0162] 7. Long-term time series pattern recognition and adaptive attention mechanism based on Transformer, including Transformer capturing longer-distance dependencies through self-attention mechanism.

[0163] 8. Adaptive spatio-temporal feature fusion and weighted decision-making mechanism, including after parallelizing LSTM and Transformer, fusing their outputs, and through a weighted fusion mechanism, making the outputs of both pass through a non-linear transformation for weighted combination.

[0164] 9. Test error <θ, including checking whether the test error of the model is less than the preset threshold θ.

[0165] If so, proceed to the next step.

[0166] If not, update the hyperparameters (adjust the non-learning parameters of the model), and return to the multi-graph convolutional network modeling and feature fusion step.

[0167] 10. SOH prediction and calculation, including calculating the state of health (SOH) of the battery pack based on the prediction results of the model.

[0168] The specific implementation and technical effects of each step in this embodiment are similar to those of the above method embodiment, and will not be elaborated here.

[0169] Based on the same inventive concept, the embodiments of the present disclosure also provide a device for predicting the state of health of a battery pack. As Figure 3 shown, the device 300 for predicting the state of health of the battery pack includes:

[0170] A data acquisition module 310, configured to acquire battery pack operation data of at least two battery packs within a preset time period; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series;

[0171] A temperature similarity calculation module 320, configured to calculate the temperature difference of the battery packs at each time point based on the battery pack temperature data, and calculate the temperature similarity between the battery packs based on the temperature difference at each time point;

[0172] A current similarity calculation module 330, configured to calculate the current similarity between the battery packs based on the frequency domain representation of the battery pack current data;

[0173] A voltage similarity calculation module 340, configured to calculate the covariance matrix of voltage changes based on the battery pack voltage data, and calculate the voltage similarity between the battery packs based on the covariance matrix;

[0174] A graph construction module 350, configured to construct a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph based on the temperature similarity, current similarity, and voltage similarity between the battery packs;

[0175] A prediction module 360, configured to predict the battery pack health state of each battery pack through a preset health state prediction model based on the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph; wherein, the preset health state prediction model is constructed based on a graph convolutional network.

[0176] In a possible implementation manner, the current similarity calculation module 330 is configured to:

[0177] Perform Fourier transform on the battery pack current data of each battery pack to obtain the frequency domain representation of the battery pack current data of each battery pack;

[0178] Calculate the current similarity between the battery packs based on the frequency domain representation of the battery pack current data of each battery pack;

[0179] The voltage similarity calculation module 340 is configured to:

[0180] Calculate the covariance matrix of voltage changes based on the battery pack voltage data of each battery pack at each time point.

[0181] In a possible implementation manner, when the battery pack includes battery pack i and battery pack j in the case of ， The calculation formula for the temperature similarity between the battery packs is:

[0182]

[0183] Among them, α The weighting factor of the temperature difference n is an exponential parameter that adjusts the influence of temperature changes on the adjacency relationship △T ij (t ) is the temperature difference of the battery pack at each time point t of T i (t), T j (t) respectively represent the battery pack i and the battery pack j at the time point t of T is the length of the time series S temp (i, j) is the temperature similarity between the battery packs i represents the battery pack i , j represents the battery pack j .

[0184] In a possible implementation manner, the calculation formula for the current similarity between the battery packs is:

[0185]

[0186] Among them, I i (f) represents the frequency-domain current signal of the battery pack i of I j (f) represents the frequency-domain current signal of the battery pack j of I i (t) represents the frequency-domain current signal of the battery pack i at the time point t of i represents the battery pack i , j represents the battery pack j , S current (i, j) represents the current similarity between the battery packs.

[0187] In a possible implementation manner, the calculation formula for the voltage similarity between the battery packs is:

[0188]

[0189] Wherein, and are the mean values of the voltages of battery pack i and battery pack j respectively, V i (t) is the voltage of battery pack i at time point t respectively, V j (t) is the voltage of battery pack j at time point t respectively, T is the length of the time series, represents the voltage similarity between battery pack i and battery pack j respectively.

[0190] In a possible implementation manner, the prediction module 360 is configured to:

[0191] Input the temperature adjacency graph, current adjacency graph, and voltage adjacency graph into the preset health state prediction model, and process the temperature adjacency graph, current adjacency graph, and voltage adjacency graph through the preset health state prediction model to obtain the graph convolution processing results of the temperature adjacency graph, current adjacency graph, and voltage adjacency graph;

[0192] Process the graph convolution processing results based on a time series model to obtain the long-term and short-term dependencies and long-term time series patterns of the battery capacity change of the battery pack; wherein, the time series model includes a long short-term memory network and a Transformer;

[0193] Predict the battery pack health state of each battery pack based on the long-term dependencies and long-term time series patterns of the battery capacity change of the battery pack.

[0194] The specific implementation manners and technical effects of the device provided by the embodiments of the present disclosure are similar to those of the above method embodiments, and will not be elaborated herein.

[0195] According to the embodiments of the present disclosure, the present disclosure also discloses an electronic device, a computer-readable storage medium, and a computer program product.

[0196] Figure 4FIG. 0 shows a schematic block diagram of an electronic device 400 that can be used to implement embodiments of the present disclosure. The electronic device 400 is intended to represent various forms of digital computers, such as, for example, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as, for example, personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0197] As Figure 4 shown, the electronic device 400 includes a computing unit 401 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 400 can also be stored. The computing unit 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0198] A plurality of components in the electronic device 400 are connected to the I / O interface 405, including: an input unit 406, such as a keyboard, a mouse, etc.; an output unit 407, such as various types of displays, speakers, etc.; a storage unit 408, such as a magnetic disk, an optical disk, etc.; and a communication unit 409, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 409 allows the electronic device 400 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0199] The computing unit 401 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 executes the various methods and processes described above, such as the battery pack health state prediction method. For example, in some embodiments, the battery pack health state prediction method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded into the RAM 403 and executed by the computing unit 401, one or more steps of the battery pack health state prediction method described above can be executed. Alternatively, in other embodiments, the computing unit 401 can be configured to execute the battery pack health state prediction method in any other suitable manner (e.g., by means of firmware).

[0200] The various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), systems-on-a-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special or general-purpose programmable processor that receives data and instructions from a storage system, at least one input device, and at least one output device, and transmits the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0201] The program code of the computer program product for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program codes can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0202] In the context of this disclosure, a computer-readable storage medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A computer-readable storage medium can be a machine-readable signal medium or a machine-readable storage medium. A computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a computer-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0203] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0204] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0205] A computer system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is created by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services ("Virtual Private Server", or simply "VPS"). The server may also be a server of a distributed system or a server combined with a blockchain.

[0206] It should be understood that various forms of processes shown above can be used, steps can be reordered, added or deleted. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of this disclosure can be achieved, and no limitation is imposed herein.

[0207] The above specific embodiments do not constitute a limitation on the protection scope of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of this disclosure shall be included within the protection scope of this disclosure.

Claims

1. A method for predicting the state of health of a battery pack, characterized in that, Including: Obtain the battery pack operation data of at least two battery packs within a preset time period; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series; Based on the battery pack temperature data, calculate the temperature difference of the battery pack at each time point, and based on the temperature difference at each time point, calculate the temperature similarity between the battery packs; Based on the frequency domain representation of the battery pack current data, calculate the current similarity between the battery packs; Based on the battery pack voltage data, calculate the covariance matrix of the voltage change, and based on the covariance matrix, calculate the voltage similarity between the battery packs; Based on the temperature similarity, current similarity, and voltage similarity between the battery packs, construct a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph; Through a preset health state prediction model, based on the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph, predict the battery pack health state of each battery pack; wherein, the preset health state prediction model is constructed based on a graph convolutional network.

2. The method for predicting the state of health of a battery pack according to claim 1, wherein, The calculating the current similarity between the battery packs based on the frequency domain representation of the battery pack current data includes: Perform Fourier transform on the battery pack current data of each battery pack to obtain the frequency domain representation of the battery pack current data of each battery pack; Based on the frequency domain representation of the battery pack current data of each battery pack, calculate the current similarity between the battery packs; The calculating the covariance matrix of the voltage change based on the battery pack voltage data includes: Based on the battery pack voltage data of each battery pack at each time point, calculate the covariance matrix of the voltage change.

3. The method for predicting the state of health of a battery pack according to claim 2, wherein In the case where the battery pack includes a battery pack i and a battery pack j the formula for calculating the temperature similarity between the battery packs is as follows: ， In the case Among them, α The weighting factor of the temperature difference, n is an exponential parameter that adjusts the influence of temperature changes on the adjacency relationship, △T ij (t ) is the temperature difference of the battery pack at each time point t of, T i (t), T j (t) respectively represent the battery pack i and the battery pack j at the time point t of temperature, T is the length of the time series, S temp (i, j) is the temperature similarity between battery packs, i represents the battery pack i , j represents the battery pack j .

4. The method for predicting the state of health of a battery pack according to claim 3, wherein The calculation formula for the current similarity between the battery packs is: Among them, I i (f) represents the frequency-domain current signal of the battery pack i ; I j (f) represents the frequency-domain current signal of the battery pack j ; I i (t) represents the frequency-domain current signal of the battery pack i at the time point t ; i represents the battery pack i , j represents the battery pack j , S current (i,j) represents the current similarity between battery packs.

5. The method for predicting the state of health of a battery pack according to claim 3, wherein The calculation formula for the voltage similarity between the battery packs is: Among them, and are the mean values of the voltages of battery packs i and battery packs j respectively, V i (t) is the voltage of battery pack i at time point t ; V j (t) is the voltage of battery pack j at time point t ; T is the length of the time series, represents the voltage similarity between battery pack i and battery pack j .

6. The method for predicting the state of health of a battery pack according to claim 1, wherein The predicting the battery pack health state of each battery pack through the preset health state prediction model based on the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph includes: Input the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph into the preset health state prediction model, and process the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph through the preset health state prediction model to obtain the graph convolutional processing results of the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph; Based on a time series model, process the graph convolutional processing results to obtain the long-term dependence and long-term time series pattern of the battery capacity change of the battery pack; wherein, the time series model includes a long short-term memory network and a Transformer; Based on the long-term dependence and long-term time series pattern of the battery capacity change of the battery pack, predict the battery pack health state of each battery pack.

7. A battery pack state of health prediction device, characterized in that, Including: A data acquisition module for obtaining the battery pack operation data of at least two battery packs within a preset time period; wherein, the battery pack operation data includes battery pack temperature data, battery pack current data, and battery pack voltage data, and the battery pack operation data is represented in the form of a time series; A temperature similarity calculation module, configured to calculate the temperature difference of the battery pack at each time point based on the battery pack temperature data, and calculate the temperature similarity between the battery packs based on the temperature differences at each time point; A current similarity calculation module, configured to calculate the current similarity between the battery packs based on the frequency domain representation of the battery pack current data; A voltage similarity calculation module, configured to calculate the covariance matrix of the voltage change based on the battery pack voltage data, and calculate the voltage similarity between the battery packs based on the covariance matrix; A graph construction module, configured to construct a temperature adjacency graph, a current adjacency graph, and a voltage adjacency graph based on the temperature similarity, current similarity, and voltage similarity between the battery packs; A prediction module, configured to predict the battery pack health state of each battery pack through a preset health state prediction model based on the temperature adjacency graph, the current adjacency graph, and the voltage adjacency graph; wherein, the preset health state prediction model is constructed based on a graph convolutional network.

8. An electronic device, characterized in that, Comprising: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to execute the instructions to implement the battery pack health state prediction method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the battery pack health state prediction method according to any one of claims 1-6.

10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, it implements the battery pack health state prediction method according to any one of claims 1-6.

Images