A method and system for evaluating the health status and predicting the life of a battery in a train unit

Abstract

The present application relates to the technical field of battery pack assessment for electric trains, and provides a method and system for battery health status assessment and life prediction for electric trains. The method includes establishing a database using voltage fluctuation and temperature data in a floating charge state, analyzing the correlation mechanism between target parameters such as battery capacity retention and active lithium inventory and internal chemical substance loss; extracting driving behavior characteristics, temperature parameters, and electrochemical impedance spectra at three time granularities, namely, seconds, minutes, and hours, and aligning multi-frequency data using a dynamic time warping algorithm; constructing a graph structure model with electrochemical parameters as nodes and cross-level associations as edges, and extracting the topological relationship of battery aging characteristics through a graph neural network; and finally establishing a prediction model that integrates multi-dimensional features to achieve accurate prediction of the remaining battery life and capacity decay inflection point. Combining multi-scale features with electrochemical mechanisms improves the accuracy of battery health status assessment under complex working conditions.

Description

Technical Field

[0001] The present application relates to the technical field of EMU battery evaluation, and in particular to a method and system for evaluating the health status and predicting the life of EMU batteries. Background Art

[0002] The contents of this section merely provide background information related to this application and may not constitute prior art.

[0003] In the field of EMU operation, EMU batteries are the key power source, and their health status is directly related to the operational safety and performance of the EMU. In the management and supervision of public facilities such as EMUs, accurately assessing the health status of batteries and predicting their lifespan are of great significance to ensuring the reliable operation of EMUs.

[0004] With the development of technology, some existing methods attempt to predict battery life. For example, patent publication number CN113393064A proposes a life prediction method and terminal equipment for cadmium nickel batteries in EMUs. This method uses an algorithm that combines particle filtering and extended Kalman filtering to establish a degradation model based on the battery's historical capacity data and discharge depth data, thereby achieving life prediction.

[0005] However, such methods have numerous limitations. For one thing, their feature data extraction is incomplete, focusing only on a limited number of features while ignoring a variety of key characteristic information about the battery at different time scales, such as the frequency of sudden acceleration or deceleration events during EMU operation, subtle changes in battery temperature data, and the frequency domain characteristics of voltage and current signals. This makes the assessment of battery health status lacking sufficient comprehensiveness and accuracy. Furthermore, faced with the complex and ever-changing driving modes and operating conditions in actual EMU operation, this method lacks adaptability to battery life prediction under different driving modes, making it difficult to accurately reflect the differential impact of different operating conditions on battery life.

[0006] Therefore, there is an urgent need for a method and system for evaluating the health status and life prediction of EMU batteries to evaluate the battery pack in a more comprehensive, accurate and adaptable manner. Summary of the Invention

[0007] In order to solve the above technical problems, the purpose of this application is to provide a method and system for evaluating the health status and predicting the life of batteries in EMUs, which obtains a variety of key characteristic information of batteries at different time scales by extracting feature data at different time levels, thereby improving the comprehensiveness and accuracy of battery health status evaluation; at the same time, by adopting a dynamic time warping algorithm to align data with different sampling frequencies and identify the driving mode of the EMU, the adaptability of battery life prediction under different driving modes is improved.

[0008] The purpose of this application is achieved through the following technical solutions:

[0009] In a first aspect, the present invention provides a method for evaluating the health status and predicting the life of a battery in a train unit, comprising:

[0010] Based on the EMU running status at a preset speed within a preset time in the floating charge state, the floating charge voltage fluctuation data and EMU battery temperature data are collected and a database is established;

[0011] Using the data corresponding to the charge and discharge cycles in the database, the target parameters of the battery are monitored, including capacity retention, active lithium inventory, and phase variation of the cathode material. A correlation mapping relationship is established between the target parameters and the loss of chemically active substances within the battery, electrode structure degradation, and interface side reactions, to obtain information on aging mechanisms.

[0012] Extract feature data at different time levels, including extracting the frequency of sudden acceleration or deceleration events during EMU operation at a time granularity of seconds, using a dynamic time warping algorithm to align data with different sampling frequencies and identify the EMU's driving mode; extracting EMU battery temperature data at a time granularity of minutes; extracting EMU voltage and current signals at a time granularity of hours; and performing frequency domain analysis on the voltage and current signals to obtain an electrochemical impedance spectrum.

[0013] The aging mechanism information is constructed as a graph structure, where nodes represent electrochemical impedance spectroscopy or feature data corresponding to different time levels, and edges represent the association between feature data at different time levels. Based on the graph structure, graph topology features reflecting the battery aging state are extracted, and a corresponding graph neural network model is established.

[0014] The graph neural network model is trained with feature data as input and the remaining life and capacity decay inflection point of battery samples as output to obtain a prediction model; new voltage data, current data, temperature data and electrochemical impedance spectrum are input into the prediction model to obtain the prediction results.

[0015] Furthermore, the step of extracting the voltage signal and current signal of the motor vehicle at the hourly time granularity specifically includes:

[0016] Select the preset number of decomposition layers and perform multi-scale decomposition and extraction of voltage and current signals using the wavelet packet decomposition method.

[0017] Furthermore, a dynamic time warping algorithm is used to align data with different sampling frequencies and identify the driving mode of the train, specifically including:

[0018] Obtain test sequences and standard sequences for EMU driving modes;

[0019] Construct the distance matrix between the test sequence and the standard sequence;

[0020] Based on the distance matrix, find an alignment path from the starting point to the end point of the matrix. The alignment path satisfies the preset boundary conditions, preset continuity conditions, and preset monotonic conditions. The alignment distance is obtained by accumulating the distances of each point on the path. The preset boundary condition is that the alignment path starts from the starting point of the distance matrix and ends at the other end of the distance matrix diagonal. The preset continuity condition is that only matrices adjacent to the current path point can be selected for alignment. The preset monotonic condition is that the alignment path is monotonic on the time axis.

[0021] If the alignment distance is less than or equal to the threshold, it indicates that the driving modes in the test sequence and the standard sequence are similar. The driving mode corresponding to the standard sequence is output when the alignment distance is greater than the threshold. The driving mode corresponding to the test sequence is obtained by comparing the known driving modes in the standard sequence.

[0022] Furthermore, graph topology features reflecting the battery aging status are extracted based on the graph structure, including:

[0023] The battery voltage, current and electrochemical impedance spectrum are used as input parameters. The Butler-Volmer equation is used to express the relationship between the effect of voltage on the electrode polarization state and the electrochemical impedance spectrum, and the graph topological characteristics are obtained.

[0024] Furthermore, the formula corresponding to the influence relationship is:

[0025]

[0026] The boundary conditions are:

[0027]

[0028] Where, 、 They are the electron current of the battery external circuit and the ion current of the battery internal circuit; 、 They are the electron potential of the battery's external circuit and the ion potential of the battery's internal circuit, 、 are the electrical conductivities of the electrode solid skeleton and the electrolyte respectively; is the working current output by the electrode; is the electrode thickness; It is the product of the specific surface area of ​​the porous electrode and the exchange current density; is the transfer coefficient; The number of electrons converted by the electrode reaction; is a constant; is the corresponding position of the electrode.

[0029] Furthermore, after establishing the corresponding graph neural network model, it also includes:

[0030] Based on the node features and adjacency relationships of the graph structure, a spatiotemporal attention module is embedded between each graph neural network layer. The spatiotemporal attention module includes:

[0031] For any node, the temporal attention weight matrix and the spatial attention weight matrix are generated respectively according to the feature change trend of its adjacent nodes in the time series and the spatial topological connection strength. The temporal attention weight matrix is ​​coupled with the spatial attention weight matrix to obtain the spatiotemporal fusion attention coefficient. Based on the spatiotemporal fusion attention coefficient, the features of the adjacent nodes are weighted averaged to update the feature representation of the current node.

[0032] Furthermore, the capacity decay inflection point is obtained by a knee algorithm; the knee algorithm includes:

[0033] Extract the time series data of the capacity retention rate of the battery throughout its life cycle from the database, smooth the time series data, and obtain the aging trajectory curve after denoising;

[0034] Connect the starting point and the ending point of the aging trajectory curve to generate a reference straight line, and calculate the vertical distance from each sampling point on the curve to the reference straight line to form a distance sequence;

[0035] Traverse the distance sequence and select the point with the largest vertical distance as the capacity decay inflection point.

[0036] Furthermore, after obtaining the prediction results, it also includes:

[0037] Taking temperature as the accelerating stress and the thickening of the negative electrode SEI film as the main attenuation mechanism, the capacity attenuation law is described by the Arrhenius model to obtain the target model;

[0038] According to the relative capacity attenuation of the battery at different cycle times, the target model is used to calculate the capacity value of the battery after different cycle times.

[0039] Furthermore, the formula corresponding to the target model is:

[0040]

[0041]

[0042] in, is the relative capacity attenuation after n cycles; is a constant greater than 0; is the activation energy; is the gas constant; is the absolute temperature; is the number of cycles; is the index; The capacity value estimated by the model after the battery has been cycled n times; is the initial capacity of the battery.

[0043] In a second aspect, the present invention provides a system for evaluating the health status and predicting the life of a battery in a train unit, comprising:

[0044] The data establishment module collects the floating charge voltage fluctuation data and the battery temperature data of the EMU based on the EMU running status at a preset speed within a preset time in the floating charge state, and establishes a database;

[0045] The aging mechanism information acquisition module is used to monitor the target parameters of the battery through the data corresponding to the charge and discharge cycles in the database. The target parameters include capacity retention rate, active lithium inventory and positive electrode material phase variation; establish a correlation mapping relationship between the target parameters and the loss of chemically active substances inside the battery, electrode structure degradation and interface side reactions, and obtain aging mechanism information;

[0046] The feature data extraction module is used to extract feature data at different time levels. This includes extracting the frequency of sudden acceleration or deceleration events during EMU operation at a time granularity of seconds, using a dynamic time warping algorithm to align data with different sampling frequencies and identify the EMU's driving mode; extracting EMU battery temperature data at a time granularity of minutes; extracting EMU voltage and current signals at a time granularity of hours; and performing frequency domain analysis on the voltage and current signals to obtain electrochemical impedance spectroscopy.

[0047] A neural network building module is used to construct a graph structure based on aging mechanism information. Nodes in the graph structure represent electrochemical impedance spectroscopy or feature data corresponding to different time levels, and edges in the graph structure represent the association between feature data at different time levels. Based on the graph structure, the module extracts graph topological features that reflect the battery aging state and establishes a corresponding graph neural network model.

[0048] The result module is used to train the graph neural network model with feature data as input and the remaining life and capacity decay inflection point of battery samples as output to obtain a prediction model; new voltage data, current data, temperature data and electrochemical impedance spectrum are input into the prediction model to obtain prediction results.

[0049] In summary, the technical solutions of the embodiments of the present application have at least the following advantages and beneficial effects:

[0050] The present invention establishes a battery operation database through float charge voltage fluctuation monitoring and temperature data collection, and then extracts the correlation between parameters such as capacity retention rate and active lithium inventory and electrode degradation and side reactions to form an aging mechanism model. In order to solve the problem of incomplete feature extraction, the frequency of rapid acceleration, temperature fluctuation curve and frequency domain impedance spectrum are extracted for the second-level dynamic events, minute-level temperature change trends and hour-level electrical signal cycles in the operation of the EMU, and a dynamic time warping algorithm is used to eliminate data deviations under different driving modes. By constructing a graph neural network model with aging mechanism information as the graph neural network model and multi-time scale features as nodes, and utilizing the adaptive topological correlation characteristics of the graph neural network, the internal chemical decay process of the battery and the dynamic characteristics of the external working conditions are cross-level coupled and analyzed, realizing a deep fusion of microscopic mechanisms and macroscopic data. It improves the adaptability and accuracy of battery life prediction under complex working conditions, and can accurately identify the capacity decay inflection point and the remaining life decay trajectory. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flow chart of a method for evaluating the health status and predicting the life of a battery in an EMU provided by the present invention;

[0052] Figure 2 The voltage, temperature, current and EIS of the battery system in the present invention are used to describe the input relationship diagram of the graph neural network feature data;

[0053] Figure 3 is the adjacency matrix of the input relationship graph of the graph neural network feature data in the present invention;

[0054] Figure 4 This is the spatiotemporal fusion attention module of the graph neural network in the present invention;

[0055] Figure 5 This is a structural schematic diagram of a battery health status assessment and life prediction system for an EMU provided by the present invention. DETAILED DESCRIPTION

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0057] like Figure 1 As shown, a method for evaluating the health status and predicting the life of a battery in a train set proposed in an embodiment of the present application includes:

[0058] S101, based on the EMU running status at a preset speed within a preset time in a floating charge state, collect floating charge voltage fluctuation data and EMU battery temperature data, and establish a database.

[0059] Specifically, a basic data set reflecting the dynamic characteristics of batteries is constructed by real-time monitoring of EMU operating parameters during float charge. Float charge refers to a charging mode in which, after reaching full charge, a battery maintains a stable voltage through a continuous low current input. During this state, the internal electrochemical reactions in the battery are in a state of dynamic equilibrium, but voltage fluctuations and temperature changes still reveal key information about battery aging and performance degradation. To this end, the present invention first uses onboard sensors to simultaneously collect float charge voltage fluctuation data and battery body temperature data during long-term (e.g., one month) operation at high speeds (e.g., 300 km / h). The float charge voltage fluctuation data reflects battery polarization characteristics and internal resistance changes, while the temperature data characterizes the thermodynamic characteristics of internal side reactions. Specifically, small fluctuations in float charge voltage (e.g., between 52.5 V and 53.5 V) are directly related to the loss of active materials and the intensification of interfacial side reactions, while temperature changes (e.g., between 25°C and 45°C) affect electrochemical reaction rates and lithium ion diffusion kinetics through the Arrhenius effect. During this process, all collected data are dynamically associated with the EMU operating status (such as traction or braking) through time stamps and stored in the database in a standardized format.

[0060] S102, monitor the target parameters of the battery through the data corresponding to the charge and discharge cycles in the database, the target parameters include capacity retention rate, active lithium inventory and positive electrode material phase variable; establish a correlation mapping relationship between the target parameters and the loss of chemically active substances inside the battery, electrode structure degradation and interface side reactions, and obtain aging mechanism information.

[0061] Specifically, with the battery's capacity retention, active lithium inventory, and cathode material phase variables as the core monitoring targets, combined with voltage-capacity curves, differential capacity analysis, and electrochemical impedance spectroscopy (EIS), the coupling relationship between battery performance degradation and internal chemical active material loss, electrode structure degradation, and interface side reactions is revealed. The decline in capacity retention not only reflects the irreversible loss of active materials during lithium ion insertion / extraction, but is also directly related to the lithium inventory consumption caused by lattice distortion of the cathode material and thickening of the negative electrode solid electrolyte interface (SEI) film. By comparing the changes in the phase variables of the cathode material before and after cycling, material structure collapse or phase separation phenomena can be identified, such as the local enrichment of nickel in the ternary material or the dissolution and loss of cobalt. These structural degradations will aggravate the obstruction of lithium ion diffusion paths, thereby accelerating capacity fading.

[0062] At the same time, dynamic monitoring of active lithium inventory, combined with quantitative analysis of temperature sensitivity using the Arrhenius equation, can pinpoint the thermodynamically dominant mechanisms of interfacial side reactions (such as electrolyte decomposition and metallic lithium precipitation). When battery temperature fluctuations exceed the stable range of 25°C to 45°C, the desolvation barrier of lithium ions at the electrode / electrolyte interface decreases, intensifying competition between SEI film repair reactions and lithium dendrite growth, further consuming active lithium and causing an increase in internal resistance. By simultaneously analyzing the voltage hysteresis phenomenon during charge and discharge and the evolution of charge transfer resistance (Rct) in the EIS spectrum, a multi-parameter coordinated aging fingerprint can be established. For example, when the Rct increase exceeds 20% and the differential capacity peak shifts by 0.1V, it indicates that a significant kinetic barrier has formed in the passivation layer on the electrode surface. This mapping relationship based on the multi-dimensional correlation of electrochemistry, materials and thermodynamics not only achieves accurate classification of battery aging modes (such as positive electrode-dominated attenuation, lithium inventory loss or electrolyte depletion), but also can provide physically interpretable feature inputs for subsequent graph neural network models by reversely deducing the quantitative relationship between electrode polarization voltage and side reaction rate, thereby significantly improving the generalization ability of the life prediction model for complex operating conditions.

[0063] S103, extracting feature data at different time levels, including extracting the frequency of sudden acceleration or deceleration events during EMU operation at a time granularity of seconds, using a dynamic time warping algorithm to align data with different sampling frequencies and identify the EMU's driving mode; extracting EMU battery temperature data at a time granularity of minutes; extracting EMU voltage and current signals at a time granularity of hours; and performing frequency domain analysis on the voltage and current signals to obtain an electrochemical impedance spectrum.

[0064] Specifically, this step constructs a holographic representation of the battery's dynamic behavior through feature extraction and fusion technology at multiple levels of time granularity. First, for driving events with a time granularity of seconds, the system monitors in real time the instantaneous frequency of sudden acceleration or deceleration during vehicle operation. These highly dynamic events can cause dramatic fluctuations in battery load, leading to a sudden increase in the lithium ion concentration gradient at the electrode interface, which in turn accelerates the accumulation of structural stress in the active material. To accurately correlate driving behavior with battery response, a dynamic time warping (DTW) algorithm is used to align sensor data with different sampling frequencies.

[0065] In detail, the dynamic time warping algorithm mainly constructs a distance matrix between the test sequence (such as the current driving operation data) and the standard sequence (preset driving mode template). Based on the distance matrix, it finds an alignment path from the starting point to the end point of the matrix. The alignment path satisfies the preset boundary conditions, preset continuity conditions, and preset monotonicity conditions. The alignment distance is obtained by accumulating the distances of each point on the path. The corresponding calculation process is as follows:

[0066] Suppose there are two time series and , is a test sequence with a length of ; is a standard sequence with a length of :

[0067] (1)

[0068] (2)

[0069] in, Representing time series Operation data at the nth time point in , Representing time series The operation data at the mth time point in .

[0070] To align the time series and , for the sequence and Construct an n×m distance matrix M, where the element M in the i-th row and j-th column of the matrix is i,j for point and The distance between Euclidean distance is usually used. After T and S are aligned, the alignment path W formed by the matrix box can be expressed as:

[0071] (3)

[0072] (4)

[0073] in represents the middle node of the alignment path W, Represents the last node of the alignment path W.

[0074] The constraints for the alignment path W are:

[0075] Preset boundary conditions: The alignment path starts from the starting point of the distance matrix and ends at the other end of the distance matrix diagonal, that is, , .

[0076] Preset continuity condition: Only matrices adjacent to the current path point can be selected for alignment, that is, if , Must satisfy and .

[0077] Preset monotonic condition: the alignment path is monotonic on the time axis. , , then it satisfies and .

[0078] Starting from the initial point (1,1) of the matrix, each subsequent point is the accumulation of the distance calculated by the previous points on the path. When the cumulative distance The final alignment distance , the expression is as follows:

[0079] (5)

[0080] (6)

[0081] If the alignment distance is less than or equal to the threshold, it indicates that the driving modes in the test sequence and the standard sequence are similar. The driving mode corresponding to the standard sequence is output when the alignment distance is greater than the threshold. The driving mode corresponding to the test sequence is obtained by comparing the known driving modes in the standard sequence.

[0082] In addition, temperature data collection with minute-level granularity focuses on the thermal dynamic characteristics of the battery itself, especially the impact of temperature fluctuations on the kinetics of electrochemical reactions. For example, when the temperature rises from 25°C to 45°C, the lithium ion diffusion coefficient increases exponentially with the Arrhenius relationship, but overheating (such as exceeding 50°C) will trigger a chain reaction of electrolyte decomposition. This level of data is processed through sliding window mean filtering to effectively suppress the interference of environmental noise on temperature trend analysis.

[0083] Furthermore, at the hourly time granularity, wavelet packet decomposition is used to extract multi-scale features from the voltage and current signals. Using a preset number of decomposition levels (e.g., 5), the original signal is decomposed into different frequency bands. The low-frequency components reflect the macroscopic evolution of the battery polarization process, while the high-frequency components capture the characteristic harmonic components of microscopic side reactions on the electrode surface. Combined with simultaneously collected charge-discharge cycle data, the reconstructed sub-signals are subjected to a fast Fourier transform to generate an electrochemical impedance spectrum (EIS). The Cole-Cole plot (i.e., Cole-Cole plot) of the real and imaginary components quantifies the synergistic growth of the charge transfer resistance (Rct) and the ohmic internal resistance. For example, when the Rct increases by more than 20% and the phase angle in the low-frequency region decreases significantly, it indicates that the formation of a passivation layer at the electrode interface has severely hindered lithium ion migration. This multi-temporal feature extraction framework, through the cross-scale fusion of second-level event-driven, minute-level thermodynamic tracking, and hour-level impedance spectrum analysis, provides full-dimensional data support for subsequent graph neural network models, from transient shocks to long-term decay. The wavelet decomposition expression is as follows:

[0084] (7)

[0085] In the formula, is the wavelet packet coefficient on the jth sub-band of the i-th layer after signal decomposition; i represents the number of layers, j represents the number, 、 are the low-pass and high-pass filter coefficients respectively; is the number of sub-bands. The above formula is iterated again to obtain the wavelet packet coefficients of the current signal and voltage signal in different frequency bands. Then the wavelet packet coefficients are processed by multi-threshold to extract useful harmonic information. The expression is:

[0086] (8)

[0087] In the formula, is the wavelet packet coefficient after multi-threshold processing; is the nth threshold.

[0088] S104, using the aging mechanism information as a graph structure, where nodes in the graph structure represent electrochemical impedance spectra or feature data corresponding to different time levels, and edges in the graph structure represent the association between feature data at different time levels; extracting graph topology features reflecting the battery aging state based on the graph structure, and establishing a corresponding graph neural network model; Figure 2 As shown in the figure, EIS is the electrochemical impedance spectroscopy, I is the current, U1, U2...Un are the voltage values ​​extracted at different time granularities of hours, and T1, T2...Tn are the temperature values ​​extracted at different time granularities of minutes.

[0089] Specifically, a graph structure is constructed based on information about battery aging mechanisms. In this graph, nodes represent electrochemical impedance spectroscopy (EIS) or characteristic data corresponding to different time levels (seconds, minutes, and hours). This characteristic data includes battery voltage, current, temperature, and other key parameters derived from this data (such as the frequency of rapid acceleration / deceleration events, battery temperature and humidity, etc.). The edges in the graph represent the relationships between these characteristic data at different time levels, reflecting the interactions and influences between various battery parameters under different operating conditions.

[0090] Next, based on the above graph structure, the graph topological features that reflect the battery aging state are extracted. This process is achieved by introducing the Butler-Volmer equation (i.e., the Butler-Volmer equation), which describes how voltage changes affect the electrochemical impedance spectroscopy (EIS) by changing the polarization state of the electrode. Specifically, the Butler-Volmer equation reveals the intrinsic connection between voltage and electrochemical impedance by expressing the relationship between parameters such as the external circuit electron current, the battery internal circuit ion current, the battery external circuit electron, the battery internal circuit ion potential, and conductivity. By solving this equation, the graph topological features of each node in the graph structure can be obtained. These features contain key information about the battery aging state. The corresponding formula is:

[0091] (9)

[0092] The boundary conditions are:

[0093] (10)

[0094] Where, 、 They are the electron current of the battery external circuit and the ion current of the battery internal circuit; 、 They are the electron potential of the battery's external circuit and the ion potential of the battery's internal circuit, 、 are the electrical conductivities of the electrode solid skeleton and the electrolyte respectively; is the working current output by the electrode; is the electrode thickness; It is the product of the specific surface area of ​​the porous electrode and the exchange current density; is the transfer coefficient; The number of electrons converted by the electrode reaction; is a constant; is the corresponding position of the electrode.

[0095] In addition, after constructing the graph neural network model, considering that the battery aging process is a complex spatiotemporal dynamic process, a spatiotemporal attention module is embedded between the graph neural network layers. The principle is as follows Figure 4 As shown in Figure 2. This module generates a temporal attention weight matrix and a spatial attention weight matrix for any node based on the feature change trend of its adjacent nodes in the time series (temporal attention) and the spatial topological connection strength (spatial attention). Subsequently, these two weight matrices are coupled to obtain the spatiotemporal fusion attention coefficient. Based on this coefficient, the features of the adjacent nodes (i.e., the adjacent node matrix, as shown in Figure 2) are added. Figure 3) to update the feature representation of the current node. This process not only considers the temporal dynamics of battery aging but also the spatial correlations between parameters, significantly improving the model's ability to generalize and reason about battery aging paths.

[0096] S105, using the feature data as input and the remaining life and capacity decay inflection point of the battery sample as output, trains the graph neural network model to obtain a prediction model; inputs the new voltage data, current data, temperature data and electrochemical impedance spectrum into the prediction model to obtain the prediction results.

[0097] Specifically, by introducing a graph neural network (GNN) that incorporates multi-dimensional feature fusion and an improved spatiotemporal attention mechanism, dynamic modeling of EMU battery aging characteristics and lifespan degradation paths is achieved. The core principle is to abstract battery operational features at different time levels (such as second-level sudden acceleration events, minute-level temperature fluctuations, and hour-level electrochemical impedance spectroscopy) into nodes in a graph structure. After aligning the time series data using dynamic time warping (DTW), a graph convolutional network is used to extract the spatiotemporal correlations between nodes. For example, a sudden temperature rise and a sudden change in charge transfer resistance (Rct) are mapped as strong edges between nodes. The spatiotemporal attention mechanism further dynamically adjusts feature weights. In the temporal dimension, it quantifies the nonlinear acceleration of lithium dendrite growth under high-temperature conditions by analyzing the temporal correlation between voltage hysteresis and the increase in Rct (e.g., a 1.8-fold increase in dendrite growth rate for every 10°C increase in temperature). In the spatial dimension, it identifies the dominant role of cathode material phase shift anomalies (e.g., a 30% decrease in lithium diffusion coefficient due to nickel enrichment) in capacity degradation based on the coupled relationship between diffusion impedance and differential capacity peak shift in electrochemical impedance spectroscopy (EIS). Through end-to-end training, the model learns the complex co-evolution of multiple parameters during battery aging, ultimately outputting a predicted remaining life and capacity inflection point (e.g., a sudden capacity drop to 80% ± 2% at 500 cycles).

[0098] Among them, the capacity decay inflection point is obtained through the knee algorithm; the knee algorithm includes: extracting the capacity retention rate time series data of the battery throughout its life cycle from the database, smoothing the time series data, and obtaining the denoised aging trajectory curve; connecting the starting point and the end point of the aging trajectory curve to generate a reference straight line, calculating the vertical distance from each sampling point on the curve to the reference straight line to form a distance sequence; traversing the distance sequence, and selecting the point with the largest vertical distance as the capacity decay inflection point.

[0099] Specifically, first, the capacity retention rate time series data of the battery in multiple charge and discharge cycles is extracted from the database, and the capacity retention rate is defined as the percentage of the current cycle discharge capacity to the rated capacity. Due to the influence of sensor noise and operating condition fluctuations in actual operation, there are local outliers in the original time series data. Therefore, the Savitzky-Golay filter is used to smooth the time series data to obtain the denoised aging trajectory curve. For example, for a certain model of lithium-ion battery, its original capacity retention rate shows a fluctuating downward trend in 0 to 1000 cycles. After Savitzky-Golay filtering with a window width of 15 and a polynomial order of 3, the high-frequency noise is effectively suppressed, and a smooth attenuation curve is obtained.

[0100] Connect the starting point of the aging trajectory curve (corresponding to the initial capacity when the number of cycles is 0) and the ending point (corresponding to the number of cycles when the capacity retention rate decays to a preset threshold, such as 80%) to generate a reference straight line. This straight line represents the theoretical aging path of the battery under the ideal linear decay mode. By calculating the vertical distance from each sampling point on the curve to the reference straight line, a distance sequence is formed. Specifically, for the number of cycles corresponding to the i-th sampling point, and its capacity retention rate , whose coordinates in the rectangular coordinate system are , and the mathematical expression of the reference straight line is C, where the slope k and intercept b are determined by the starting and ending points. According to the perpendicular distance formula from a point to a line, the distance of the i-th sampling point is It can be expressed as:

[0101] (11)

[0102] After traversing all sampling points, select The sampling point corresponding to the maximum value is taken as the capacity decay inflection point. For example, the vertical distance between the aging trajectory curve and the reference line reaches its maximum value at the 500th cycle. At this point, the capacity retention rate drops from the initial 100% to 92.8%. The decay rate after the inflection point increases by about 3.5 times compared to before the inflection point, indicating that irreversible electrode structure collapse or intensified interface side reactions have occurred within the battery.

[0103] Furthermore, after obtaining the prediction results, it also includes: using temperature as the accelerating stress and the thickening of the negative electrode SEI film as the main attenuation mechanism, describing the capacity attenuation law through the Arrhenius model, and obtaining a target model; according to the relative capacity attenuation of the battery under different cycle numbers, using the target model to calculate the capacity value of the battery after different cycle numbers.

[0104] Specifically, a quantitative analysis model for capacity decay based on temperature-accelerated stress is introduced. This model takes the thickening of the negative electrode solid electrolyte interface (SEI) film as the dominant decay mechanism and combines the Arrhenius thermodynamic equation to establish a temperature-aging rate correlation. That is, when the EMU battery operates under high temperature conditions, the interfacial side reaction activity between the electrolyte and the negative electrode graphite is significantly enhanced, resulting in the disruption of the dynamic balance between SEI film repair and growth. At this time, lithium ions need to consume additional active lithium during the embedding process of the negative electrode to compensate for the lithium inventory loss caused by the thickening of the SEI film. By using temperature as an accelerated aging factor and establishing the capacity decay activation energy parameters corresponding to different temperature ranges, the nonlinear effect of temperature fluctuations on the long-term battery decay process can be quantified.

[0105] In specific implementation, a certain type of ternary lithium-ion power battery was used as an example to conduct an accelerated aging test at 55°C under laboratory conditions. By regularly disassembling the battery to measure the thickness of the negative electrode SEI film, it was found that when the number of cycles reached 300, the SEI film thickness increased from the initial 2.1 nm to 5.8 nm, and the corresponding capacity retention rate decreased from 100% to 82.3%. Based on the Arrhenius model, the activation energy of capacity decay at this temperature was calculated to be 45 kJ / mol, compared with 35 kJ / mol under the 25°C baseline operating condition, indicating that high temperature significantly exacerbated the kinetic process of interfacial side reactions. After the corrected activation energy parameters were input into the target model, the capacity prediction error of the battery at the 150th cycle in actual vehicle operation was reduced from 8.7% before compensation to 2.3%, verifying the effectiveness of this method.

[0106] Among them, the formula corresponding to the target model is:

[0107] (12)

[0108] (13)

[0109] in, is the relative capacity attenuation after n cycles; is a constant greater than 0; is the activation energy; is the gas constant; is the absolute temperature; is the number of cycles; is the index; The capacity value estimated by the model after the battery has been cycled n times; is the initial capacity of the battery.

[0110] Based on the same inventive concept, Figure 5 As shown, the present invention provides a system for evaluating the health status and predicting the life of a battery in a train set, comprising:

[0111] The data establishment module 201 collects the floating charge voltage fluctuation data and the battery temperature data of the EMU based on the EMU running status at a preset speed within a preset time in the floating charge state, and establishes a database;

[0112] The aging mechanism information acquisition module 202 is used to monitor the target parameters of the battery using the data corresponding to the charge and discharge cycles in the database. The target parameters include capacity retention rate, active lithium inventory, and positive electrode material phase variation; establish a correlation mapping relationship between the target parameters and the loss of chemically active substances inside the battery, electrode structure degradation, and interface side reactions to obtain aging mechanism information;

[0113] Feature data extraction module 203 is used to extract feature data at different time levels, including extracting the frequency of sudden acceleration or deceleration events during EMU operation at a time granularity of seconds, using a dynamic time warping algorithm to align data with different sampling frequencies and identify the EMU's driving mode; extracting EMU battery temperature data at a time granularity of minutes; extracting EMU voltage and current signals at a time granularity of hours; and performing frequency domain analysis on the voltage and current signals to obtain an electrochemical impedance spectrum.

[0114] A neural network building module 204 is configured to extract graph topological features reflecting the battery aging state based on the aging mechanism information as a graph structure, wherein nodes in the graph structure represent electrochemical impedance spectroscopy or feature data corresponding to different time levels, and edges in the graph structure represent associations between feature data at different time levels, and to establish a corresponding graph neural network model.

[0115] The result module 205 is used to train the graph neural network model with the feature data as input and the remaining life and capacity decay inflection point of the battery sample as output to obtain a prediction model; the new voltage data, current data, temperature data and electrochemical impedance spectrum are input into the prediction model to obtain the prediction results.

[0116] The above are merely preferred embodiments of the present application and are not intended to limit the present application. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for evaluating the health status and predicting the life of a battery in a train unit, characterized in that: include: Based on the EMU running status at a preset speed within a preset time in the floating charge state, the floating charge voltage fluctuation data and EMU battery temperature data are collected and a database is established; Monitoring target parameters of the battery using data corresponding to charge and discharge cycles in the database, including capacity retention, active lithium inventory, and positive electrode material phase variation; establishing a correlation mapping relationship between the target parameters and the loss of chemically active substances within the battery, electrode structure degradation, and interface side reactions, to obtain aging mechanism information; Extract feature data at different time levels, including extracting the frequency of sudden acceleration or deceleration events during EMU operation at a time granularity of seconds, using a dynamic time warping algorithm to align data with different sampling frequencies and identify EMU driving patterns; extracting EMU battery temperature data at a time granularity of minutes; and extracting EMU voltage and current signals at a time granularity of hours. Performing frequency domain analysis on the voltage signal and the current signal to obtain an electrochemical impedance spectroscopy; The aging mechanism information is formed into a graph structure, wherein nodes in the graph structure represent electrochemical impedance spectra or feature data corresponding to different time levels, and edges in the graph structure represent associations between feature data at different time levels; graph topology features reflecting the battery aging state are extracted based on the graph structure, and a corresponding graph neural network model is established; The graph neural network model is trained with characteristic data as input and the remaining life and capacity decay inflection point of the battery sample as output to obtain a prediction model; new voltage data, current data, temperature data and electrochemical impedance spectrum are input into the prediction model to obtain a prediction result.

2. The method for evaluating the health status and predicting the life of a train battery according to claim 1, characterized in that: The step of extracting the voltage signal and current signal of the motor vehicle at the hourly time granularity specifically includes: A preset number of decomposition layers is selected, and the voltage signal and the current signal are subjected to multi-scale decomposition and extraction using a wavelet packet decomposition method.

3. The method for evaluating the health status and predicting the life of a train battery according to claim 1, characterized in that: The steps for using the dynamic time warping algorithm to align data with different sampling frequencies and identify the driving mode of the train include: Obtain test sequences and standard sequences for EMU driving modes; Constructing a distance matrix between the test sequence and the standard sequence; Based on the distance matrix, an alignment path is found from the starting point to the end point of the matrix, and the alignment path satisfies preset boundary conditions, preset continuity conditions, and preset monotonic conditions, and the alignment distance is obtained by accumulating the distances of each point on the path; the preset boundary condition is that the alignment path starts from the starting point of the distance matrix and ends at the other end of the distance matrix diagonal; the preset continuity condition is that only matrices adjacent to the current path point can be selected for alignment; the preset monotonic condition is that the alignment path is monotonic on the time axis; If the alignment distance is less than or equal to a threshold, it indicates that the driving modes in the test sequence and the standard sequence are similar, and the driving mode corresponding to the standard sequence is output. If the alignment distance is greater than the threshold, the driving mode corresponding to the test sequence is obtained by comparing with the known driving modes in the standard sequence.

4. The method for evaluating the health status and predicting the life of a train battery according to claim 1, characterized in that: Extracting graph topology features reflecting the battery aging status based on the graph structure specifically includes: The battery voltage, current and electrochemical impedance spectrum are used as input parameters. The Butler-Volmer equation is used to express the relationship between the effect of voltage on the electrode polarization state and the electrochemical impedance spectrum, and the graph topological characteristics are obtained.

5. The method for evaluating the health status and predicting the life of a train battery according to claim 4, characterized in that: The formula corresponding to the influence relationship is: The boundary conditions are: Where, 、 They are the electron current of the battery external circuit and the ion current of the battery internal circuit; 、 They are the electron potential of the battery's external circuit and the ion potential of the battery's internal circuit, 、 are the electrical conductivities of the electrode solid skeleton and the electrolyte respectively; is the working current output by the electrode; is the electrode thickness; It is the product of the specific surface area of ​​the porous electrode and the exchange current density; is the transfer coefficient; The number of electrons converted by the electrode reaction; is a constant; is the corresponding position of the electrode.

6. The method for evaluating the health status and predicting the life of a train battery according to claim 1, characterized in that: After establishing the corresponding graph neural network model, the following steps are also included: Based on the node features and adjacency relationships of the graph structure, a spatiotemporal attention module is embedded between each graph neural network layer. The spatiotemporal attention module includes: For any node, a temporal attention weight matrix and a spatial attention weight matrix are generated respectively according to the feature change trend of its adjacent nodes in the time series and the spatial topological connection strength; the temporal attention weight matrix and the spatial attention weight matrix are coupled to obtain the spatiotemporal fusion attention coefficient; based on the spatiotemporal fusion attention coefficient, the features of the adjacent nodes are weighted averaged to update the feature representation of the current node.

7. The method for evaluating the health status and predicting the life of a train battery according to claim 1, characterized in that: The capacity decay inflection point is obtained by a knee algorithm; the knee algorithm includes: Extracting time series data of capacity retention rate over the entire life cycle of the battery from a database, smoothing the time series data to obtain a denoised aging trajectory curve; Connecting the starting point and the ending point of the aging trajectory curve to generate a reference straight line, and calculating the vertical distance from each sampling point on the curve to the reference straight line to form a distance sequence; The distance sequence is traversed, and the point with the largest vertical distance is selected as the capacity decay inflection point.

8. The method for evaluating the health status and predicting the life of a train battery according to claim 1, characterized in that: After obtaining the prediction result, the method further includes: Taking temperature as the accelerating stress and the thickening of the negative electrode SEI film as the main attenuation mechanism, the capacity attenuation law is described by the Arrhenius model to obtain the target model; According to the relative capacity attenuation of the battery at different cycle times, the target model is used to calculate the capacity value of the battery after different cycle times.

9. The method for evaluating the health status and predicting the life of a train battery according to claim 8, characterized in that: The formula corresponding to the target model is: in, is the relative capacity attenuation after n cycles; is a constant greater than 0; is the activation energy; is the gas constant; is the absolute temperature; is the number of cycles; is the index; The capacity value estimated by the model after the battery has been cycled n times; is the initial capacity of the battery.

10. A system for evaluating the health status and predicting the life of a battery in a train unit, characterized in that: include: A data establishment module is used to collect float charge voltage fluctuation data and EMU battery temperature data based on the EMU running status at a preset speed within a preset time in a float charge state, and establish a database; an aging mechanism information acquisition module, configured to monitor target parameters of the battery using data corresponding to charge and discharge cycles in the database, the target parameters including capacity retention, active lithium inventory, and positive electrode material phase variation; establish a correlation mapping relationship between the target parameters and the loss of chemically active substances within the battery, electrode structure degradation, and interface side reactions, to obtain aging mechanism information; The feature data extraction module is used to extract feature data at different time levels. This includes extracting the frequency of sudden acceleration or deceleration events during EMU operation at a time granularity of seconds, using a dynamic time warping algorithm to align data with different sampling frequencies and identify EMU driving patterns; extracting EMU battery temperature data at a time granularity of minutes; and extracting EMU voltage and current signals at a time granularity of hours. Performing frequency domain analysis on the voltage signal and the current signal to obtain an electrochemical impedance spectroscopy; A neural network establishment module is configured to use the aging mechanism information as a graph structure, wherein nodes in the graph structure represent electrochemical impedance spectra or feature data corresponding to different time levels, and edges in the graph structure represent associations between feature data at different time levels; extract graph topology features reflecting the battery aging state based on the graph structure, and establish a corresponding graph neural network model; The result module is used to train the graph neural network model with feature data as input and the remaining life and capacity decay inflection point of the battery sample as output to obtain a prediction model; and input new voltage data, current data, temperature data and electrochemical impedance spectrum into the prediction model to obtain prediction results.

Images