A method and system for evaluating EMU battery aging based on mechanism and data fusion

Abstract

The present invention provides a method and system for evaluating the aging of EMU batteries by integrating mechanism and data, which relates to the field of battery health management technology. The method comprises extracting and processing multi-physical characteristic parameters of batteries from experimental data to obtain a first structured data parameter and a first unstructured data parameter after processing; constructing an aging mechanism model of high-speed EMU batteries with electro-thermal-mechanical multi-physical characteristics by combining the Arrhenius equation, a quasi-two-dimensional model, and an equivalent circuit model to predict the first aging state of the battery; establishing a mapping relationship between battery performance and multi-physical characteristics to obtain a second aging state as a predicted output; treating the first aging state and the second aging state as two independent bodies of evidence, and dynamically adjusting the weight of the Dempster combination rule based on the conflict factor, thereby determining the final aging state by the maximum confidence criterion. The present invention improves the accuracy and reliability of aging state prediction, and provides technical support for life assessment and maintenance strategy optimization of high-speed EMU batteries.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health management, and in particular to a method and system for evaluating battery aging in electric train sets by integrating mechanism and data. Background Art

[0002] With the development of intelligent and green high-speed trains, the reliability of their onboard power battery systems has become a core factor affecting operational safety. Accurate characterization of battery aging status is a key technical challenge for optimizing battery life management and preventing sudden failures. However, existing methods still have limitations in aging assessment under complex operating conditions. Among them, traditional electrochemical models (such as single-particle models) have difficulty quantifying the impact of multi-physics field coupling on aging, and rely on simplified laboratory conditions, which cannot adapt to actual complex operating conditions such as long-term floating charging and impact loads of EMUs. Purely data-driven methods are limited by the high-dimensional noise and feature redundancy of battery operating data, resulting in poor model generalization ability and significantly increased prediction errors under extreme temperatures or mechanical vibrations.

[0003] Existing technologies primarily focus on monitoring battery electrical performance parameters, ignoring the correlation between mechanical and thermodynamic characteristics such as temperature field distribution and electrode material stress and strain and aging, resulting in insufficient early warning capabilities for aging. Existing fusion methods fail to address the dynamic conflict between the confidence of the outputs of mechanism models and data models, and lack a collaborative modeling framework for experimental and operational data, limiting the interpretability and accuracy of fusion results. Traditional monitoring systems rely on offline laboratory test data and lack the integration of edge computing and real-time feedback mechanisms, making it difficult to meet the high-frequency, low-latency online health management needs of high-speed trains. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for evaluating the aging of EMU batteries by integrating mechanism and data to improve the above-mentioned problems. To achieve the above-mentioned purpose, the technical solutions adopted by the present invention are as follows:

[0005] In the first aspect, the present application provides a method for evaluating the aging of EMU batteries by integrating mechanism and data, including:

[0006] Through the high-speed train battery aging experiment, the operating conditions of different temperatures, different loads, and different charge and discharge modes in actual operation are simulated. The battery is tested for electrochemical impedance spectroscopy and open circuit voltage. The experimental data of the battery aging process is recorded. The multi-physical characteristic parameters of the battery in the experimental data are extracted and processed to obtain the processed first structured data parameters and first unstructured data parameters.

[0007] Physically decomposing the first unstructured data parameters into second structured data parameters, fusing the first and second structured data parameters, and combining them with the Arrhenius equation, quasi-two-dimensional model, and equivalent circuit model to construct a high-speed train battery aging mechanism model with electrical, thermal, and mechanical multi-physical characteristics. The high-speed train battery aging mechanism model is solved using numerical simulation methods to predict the first aging state of the battery.

[0008] A local linear embedding method is used to decompose and extract key information of multi-physical feature parameters. This key information is input as a sequence into a long short-term memory network and a self-attention mechanism. A mapping relationship between battery performance and multi-physical features is established. A deep learning model for battery aging big data is generated and trained to obtain the predicted output of the second aging state. The key information includes key features, local structural information, and key features after dimensionality reduction.

[0009] The first aging state and the second aging state are regarded as two independent evidence bodies, and the weight of the Dempster combination rule is dynamically adjusted based on the conflict factor to complete the joint confidence allocation of the two independent evidence bodies, and then the final aging state is determined by the maximum confidence criterion.

[0010] Preferably, the multi-physical characteristic parameters of the battery in the extracted experimental data are processed to obtain the processed first structured data parameters and the first unstructured data parameters, wherein the processing process of the processed first structured data parameters includes:

[0011] The battery performance parameters and multi-physical characteristic parameters obtained from the high-speed train battery aging test were rank-transformed, including: sorting the values ​​of each variable from small to large and assigning a rank to each value; if there are identical values, the average rank of the identical values ​​in the variable value sorting process is assigned;

[0012] Dividing the rank-converted parameters into a first rank and a second rank, wherein the first rank is the rank of the battery performance parameters and the second rank is the rank of the multi-physical characteristic parameters;

[0013] Based on the first rank and the second rank, the Spearman correlation coefficient between each pair of performance parameters and multi-physical feature parameters is calculated, and multi-physical feature parameters with a value greater than 0.8 are screened and recorded as the processed first structured data parameters.

[0014] Preferably, the multi-physical characteristic parameters of the battery in the extracted experimental data are processed to obtain a processed first structured data parameter and a first unstructured data parameter, wherein the processing process of the processed first unstructured data parameter includes:

[0015] Obtain the Nyquist plot from the electrochemical impedance spectroscopy test, perform differential geometry processing on the Nyquist plot, and extract the characteristic frequency set of the curvature mutation point;

[0016] A dynamic weighted tensor is constructed based on a set of characteristic frequencies, an open circuit voltage curve is decomposed into a time-domain differential intrinsic mode function, and the time-domain differential intrinsic mode function is recorded as a first unstructured data parameter after processing.

[0017] Preferably, the physically decomposing the first unstructured data parameter into the second structured data parameter and fusing the first structured data parameter with the second structured data parameter comprises:

[0018] Calculating the mean, variance, energy, and time-frequency characteristics of each mode in the first unstructured data parameter, and correlating the battery health state through the aging sensitivity weighting coefficient to generate a second structured data parameter reflecting the microscopic aging mechanism;

[0019] The first structured data parameter and the second structured data parameter are fused using a multimodal fusion algorithm.

[0020] Preferably, the locally linear embedding method is used to decompose and extract key information of multiple physical characteristic parameters, including:

[0021] Based on the data of multiple physical feature parameters, the neighborhood size of each data point is determined by Euclidean distance to obtain domain information;

[0022] Combined with the neighborhood information, find the r nearest neighbor points for each data point and calculate the corresponding reconstruction weight;

[0023] Using the calculated reconstruction weights, a local neighborhood matrix is ​​constructed for each data point, where each row represents the difference between a neighboring point and the data point.

[0024] Based on the local neighborhood matrix, the low-dimensional embedding results are extracted by solving the least squares problem and matrix eigenvalue and eigenvector analysis, and the low-dimensional embedding results are recorded as key information.

[0025] Preferably, the first aging state and the second aging state are used as two independent bodies of evidence, and the weight of the Dempster combination rule is dynamically adjusted based on the conflict factor to complete the joint confidence allocation of the two independent bodies of evidence, and then the final aging state is determined by the maximum confidence criterion, which includes:

[0026] The first aging state and the second aging state are regarded as independent evidence bodies, and the confidence of the first aging state is , the confidence of the second aging state is ;

[0027] Using Dempster's combination rule, calculate the conflict factor , and dynamically adjust the weight of Dempster combination rules based on the size of the conflict factor;

[0028] Calculate the joint confidence of each aging state according to the adjusted Dempster combination rule;

[0029] The joint confidence of each aging state is arranged from high to low, and the state with the highest confidence is selected as the final aging state for output.

[0030] Secondly, the present application also provides a mechanism and data fusion EMU battery aging assessment system, including:

[0031] Extraction and processing module: used to simulate the operating conditions of different temperatures, different loads, and different charge and discharge modes in actual operation through high-speed train battery aging experiments, perform electrochemical impedance spectroscopy testing and open circuit voltage measurement on the battery, record experimental data during the battery aging process, extract and process the multi-physical characteristic parameters of the battery in the experimental data, and obtain the processed first structured data parameters and first unstructured data parameters;

[0032] The first prediction module is used to physically decompose the first unstructured data parameters into second structured data parameters, fuse the first structured data parameters with the second structured data parameters, and combine the Arrhenius equation, quasi-two-dimensional model, and equivalent circuit model to construct a high-speed train battery aging mechanism model with electrical, thermal, and mechanical multi-physical characteristics. The high-speed train battery aging mechanism model is solved using a numerical simulation method to predict the first aging state of the battery.

[0033] The second prediction module is used to decompose and extract key information of multi-physical feature parameters using a local linear embedding method. The key information is input as a sequence into the long short-term memory network and the self-attention mechanism to establish a mapping relationship between battery performance and multi-physical features. A deep learning model of battery aging big data is generated and trained to obtain the predicted output of the second aging state. The key information includes key features, local structural information, and key features after dimensionality reduction.

[0034] Determination evaluation module: used to treat the first aging state and the second aging state as two independent evidence bodies, and dynamically adjust the Dempster combination rule weights based on the conflict factor to complete the joint confidence allocation of the two independent evidence bodies, and then determine the final aging state through the maximum confidence criterion.

[0035] In a third aspect, the present application also provides a mechanism and data fusion EMU battery aging assessment device, including:

[0036] memory for storing computer programs;

[0037] A processor is used to implement the steps of the EMU battery aging assessment method that integrates the mechanism and data when executing the computer program.

[0038] In a fourth aspect, the present application also provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for evaluating the aging of EMU batteries based on mechanism and data fusion.

[0039] The beneficial effects of the present invention are:

[0040] The present invention extracts the electric-thermal-mechanical multi-physical characteristic information from the source of experimental data, studies the mapping relationship between the high-speed rail EMU battery performance and the electric-thermal-mechanical multi-physical characteristics under actual complex working conditions such as long-term floating charge and impact load, completes the performance calibration of the EMU battery by mechanism and data model, and realizes dynamic monitoring of the high-speed rail EMU battery performance and characterization of aging performance.

[0041] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the embodiments of the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 Schematic diagram of the process flow of the EMU battery aging assessment method with mechanism and data fusion described in an embodiment of the present invention;

[0044] Figure 2 Schematic diagram of the structure of the EMU battery aging assessment system with mechanism and data fusion described in an embodiment of the present invention;

[0045] Figure 3 Schematic diagram of the structure of the EMU battery aging assessment device with the mechanism and data fusion described in an embodiment of the present invention.

[0046] In the figure: 701, extraction and processing module; 702, first prediction module; 703, second prediction module; 704, determination and evaluation module; 800, EMU battery aging evaluation equipment with mechanism and data fusion; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are used only to distinguish the description and should not be understood as indicating or implying relative importance.

[0049] Example 1:

[0050] This embodiment provides a method for evaluating battery aging in electric trains by integrating mechanism and data.

[0051] See also Figure 1 , the figure shows that the method includes step S100, step S200, step S300 and step S400.

[0052] S100. Conduct an aging experiment on high-speed train battery conditions to simulate the operating conditions of different temperatures, different loads, and different charge and discharge modes in actual operation, perform electrochemical impedance spectroscopy testing and open-circuit voltage measurement on the battery, record experimental data during the battery aging process, extract multiple physical characteristic parameters of the battery from the experimental data, and process them to obtain the processed first structured data parameters and first unstructured data parameters.

[0053] It should be noted that the complex operating condition aging experiment of high-speed train batteries also includes performance testing of the batteries under different cycle numbers to obtain data on the aging trend of the batteries during long-term use, and the physical characteristics include at least three of the battery voltage and current impedance, electrochemical impedance spectrum characteristic frequency, charge and discharge curve differential parameters, battery surface temperature field distribution and temperature rise time, and electrode material stress and strain parameters.

[0054] It can be understood that, in this step S100, including S101, S102 and S103, the processing of the processed first structured data parameter includes:

[0055] S101. Performing rank conversion on the battery performance parameters and multi-physical characteristic parameters obtained from the high-speed train battery aging test, including: sorting the value of each variable from small to large and assigning a rank to each value; if there are identical values, assigning the average rank of the identical values ​​that appear in the variable value sorting process;

[0056] S102, dividing the parameters after rank conversion into a first rank and a second rank, wherein the first rank is the rank of the battery performance parameters, and the second rank is the rank of the multi-physical characteristic parameters;

[0057] S103. Based on the first rank and the second rank, calculate the Spearman correlation coefficient between each pair of performance parameters and the multi-physical feature parameters, and select the multi-physical feature parameters greater than 0.8, and record them as the processed first structured data parameters. The calculation formula of the Spearman correlation coefficient is:

[0058]

[0059] Where, is the correlation coefficient, is the battery performance parameter without rank conversion, are the multi-physical characteristic parameters that have not been rank-transformed, is the first rank, is the second rank, is the sample size, is the number of battery performance parameters, is the number of physical characteristic parameters.

[0060] It should be noted that the complex operating conditions aging test for high-speed train batteries simulates the different temperatures, loads, and charge-discharge modes encountered in actual operation, allowing for long-term battery aging testing. During the experiment, battery performance data at different cycle times was recorded, including but not limited to the following multi-physical characteristic parameters: battery voltage, current, and impedance; battery surface temperature field distribution and temperature rise time; and electrode material stress and strain parameters.

[0061] Data recording and preprocessing The above experimental data are recorded and organized to form a database, and the data are preprocessed to ensure the accuracy and consistency of the data, including removing outliers and normalization.

[0062] Use empirical rules to remove outliers in the data, where The calculation formula is:

[0063]

[0064] in, is the standard deviation of the same group of data, is the voltage in the data record, is the voltage in the data record, is the sample size.

[0065] It should be noted that It can also be any one of the original sequence data of current, impedance, temperature rise time and electrode material stress and strain parameters. It can also be any one of the average values ​​of the original sequence data of current, impedance, temperature rise time and electrode material stress and strain parameters.

[0066] The maximum and minimum method is used to normalize the data, and the calculation formula is:

[0067]

[0068] Where, Refers to any of the time series data of voltage, current, impedance, temperature rise time, and electrode material stress and strain parameters before normalization after outliers are removed. Refers to any of the time series data of voltage, current, impedance, temperature rise time, and electrode material stress and strain parameters after normalization after outliers are removed. Refers to any of the maximum values ​​of the original sequence data of voltage, current, impedance, temperature rise time and electrode material stress and strain parameters in the data record after the outliers are removed. Refers to any of the minimum values ​​of the original sequence data of voltage, current, impedance, temperature rise time and electrode material stress and strain parameters in the data record after outliers are removed.

[0069] It should be noted that step S100 also includes steps S104 and S105, wherein the processing of the processed first unstructured data parameter includes:

[0070] S104. Obtain the Nyquist diagram from the electrochemical impedance spectroscopy test, perform differential geometry processing on the Nyquist diagram, and extract the characteristic frequency set of the curvature mutation point. The calculation formula is as follows:

[0071]

[0072] Where, is the angle between the tangent direction at the corresponding frequency point and the real axis, is the set of characteristic frequencies extracted from the Nyquist plot of electrochemical impedance spectroscopy, For the angle The corresponding characteristic frequency;

[0073] S105. Construct a dynamic weighted tensor based on the characteristic frequency set, decompose the open circuit voltage curve into a time-domain differential intrinsic mode function, and record the time-domain differential intrinsic mode function as the first unstructured data parameter after processing. The calculation formula is as follows:

[0074]

[0075] Where, is a dynamic weighted tensor, For the angle The corresponding characteristic frequency, is the charge and discharge cycle, Angle The charge transfer resistance at Angle The double layer capacitance at .

[0076] The time-domain differential intrinsic mode function (TDIEMF) transforms the unstructured voltage curve into structured characteristic parameters with clear aging correlations through physical decomposition. Together with the operating condition parameters selected by Spearman, these parameters form the "processed multi-physical characteristic parameters." This dual-driven fusion of mechanism and data significantly improves the accuracy and interpretability of EMU battery aging assessment.

[0077] It should be noted that the experimental data used to extract and process the multi-physical characteristic parameters of the battery from the experimental data includes the following two categories: structured data: directly quantifiable parameters such as voltage, current, and temperature (used for rank transformation and Spearman analysis); and unstructured data: waveform data such as Nyquist plots of electrochemical impedance spectroscopy (EIS) and open-circuit voltage (OCV) curves (used for differential geometry processing and dynamic weighted tensor construction). It is understandable that structured data processing (Spearman analysis) solves the problem of initial feature screening and reduces dimensionality, while unstructured data processing (differential geometry and modal decomposition) solves the problem of physical interpretability and enhances the significance of features. Differential geometry analysis of Nyquist plots captures impedance spectrum distortion caused by SEI film growth through curvature mutation points, but traditional methods rely solely on frequency domain analysis.

[0078] S200. Physically decompose the first unstructured data parameters and convert them into second structured data parameters. Fuse the first structured data parameters with the second structured data parameters. Combine the Arrhenius equation, the quasi-two-dimensional model, and the equivalent circuit model to construct a high-speed train battery aging mechanism model with electric-thermal-mechanical multi-physical characteristics. Use a numerical simulation method to solve the high-speed train battery aging mechanism model, and then predict the first aging state of the battery.

[0079] It is understandable that step S200 includes steps S201 and S202, wherein:

[0080] S201, calculating the mean, variance, energy, and time-frequency characteristics of each mode in the first unstructured data parameter, and correlating the parameters with the battery health status through an aging sensitivity weighting coefficient to generate a second structured data parameter reflecting the microscopic aging mechanism;

[0081] S202: A multimodal fusion algorithm is used to fuse the first structured data parameter and the second structured data parameter. The calculation formula is as follows:

[0082]

[0083] Where, X fused As the fusion result, and are the first dynamic weight matrix and the second dynamic weight matrix, X struct is the first structured data parameter, X IMF is the second structuring parameter, It is Hadamard.

[0084] It should be noted that the variable The value of , calculate its rank The battery performance parameters and multi-physical characteristics are rank transformed respectively. The battery performance parameters are , the multi-physics features are , calculate their ranks respectively and , is the battery performance parameter without rank conversion, are the multi-physical characteristic parameters that have not been rank-transformed, is the number of battery performance parameters, j is the number of physical parameters, where , .

[0085] This step S200 also includes combining the Arrhenius equation, the quasi-two-dimensional model, and the equivalent circuit model to construct a high-speed train battery aging mechanism model with electrical-thermal-mechanical multi-physical characteristics, wherein the calculation formula of the Arrhenius equation is as follows:

[0086]

[0087] Where, is a computational symbol representing a function, The temperature is The reaction rate constant, is the pre-exponential factor, is the activation energy, is the universal gas constant, is the absolute temperature;

[0088] The quasi-two-dimensional model includes the charge conservation equation, the material conservation equation, and the heat conservation equation. The calculation formulas of the three equations are as follows:

[0089] The charge conservation equation , where is the potential in the electrode, is the electrode conductivity, is the current density in the electrode, For time;

[0090] The material conservation equation ,

[0091] Where, is the solid phase lithium ion concentration, is the solid phase diffusion coefficient, is the current density, is the Faraday constant, is the volume fraction of electrode particles, is the solid phase lithium ion concentration, is the solid phase diffusion coefficient For time, is the absolute temperature;

[0092] The heat conservation equation , where is the density, is the specific heat capacity, is the thermal conductivity, The heat generated by the electrochemical reaction, For heat loss, For time, is the absolute temperature;

[0093] The calculation formula of the equivalent circuit model is as follows:

[0094]

[0095] Where, 、 and are ohmic internal resistance, electrochemical polarization resistance and diffusion polarization resistance, respectively. and are the electrochemical polarization capacitance and the diffusion polarization capacitance, is the battery open circuit voltage, is the terminal voltage of the battery, is the electrochemical polarization resistance The voltage, is the diffusion polarization resistance The voltage, For flow Current;

[0096] Considering the electrochemical reaction, heat transfer and mechanical behavior inside the battery, combined with the Arrhenius equation, quasi-two-dimensional model and equivalent circuit model, an electric-thermal-mechanical coupling model is established. The calculation formula is as follows:

[0097]

[0098] Where, is the current density, is the exchange current density, 、 、 and are the liquid surface concentration, solid surface concentration reference concentration and maximum concentration, respectively. 、 、 and are total strain, elastic strain, diffusion-induced strain and thermal strain, respectively. is stress, is the elastic modulus.

[0099] It should be noted that the above coupling equations are solved by numerical simulation methods to obtain the mechanism model prediction results of the battery aging state.

[0100] S300. Use the local linear embedding method to decompose and extract the key information of multi-physical feature parameters, input the key information as a sequence into the long short-term memory network and the self-attention mechanism, establish a mapping relationship between battery performance and multi-physical features, generate a deep learning model of battery aging big data and train it, and obtain the predicted output of the second aging state, where the key information includes key features, local structure information and key features after dimensionality reduction.

[0101] It can be understood that step S300 includes S301, S302, S303 and S304, wherein:

[0102] S301, based on the data of multiple physical feature parameters, determine the neighborhood size of each data point through Euclidean distance to obtain domain information;

[0103] S302: Combined with the neighborhood information, find the r nearest neighbor points for each data point and calculate the corresponding reconstruction weight. The calculation formula is as follows:

[0104]

[0105] In the formula, in the formula, To reconstruct the weights, is the identity matrix, is the weight matrix, is the transpose of the matrix;

[0106] S303. Using the calculated reconstruction weights, construct a local neighborhood matrix for each data point, where each row represents the difference between a neighboring point and the data point. The calculation formula for the local neighborhood matrix is ​​as follows:

[0107] For each data point , calculated by the formula :

[0108]

[0109] Where, is the local neighborhood matrix, For all neighboring points, for point The reconstruction weight of

[0110] S304: Based on the local neighborhood matrix, extract the low-dimensional embedding result by solving the least squares problem and matrix eigenvalue eigenvector analysis, and record the low-dimensional embedding result as key information

[0111] It should be noted that the neighborhood is determined by using a distance metric (such as Euclidean distance) to balance the capture of local structure and noise robustness to determine the neighborhood size r.

[0112] For each data point , find its r nearest neighbor points and calculate their corresponding reconstruction weights For each data point , construct a local neighborhood matrix , where each row represents a neighboring point and , solve the following least squares problem:

[0113]

[0114] Where, is the weight vector, N is the weight matrix, solve the above least squares problem, and get the reconstruction weight of each data point , and then calculate the reconstruction weight of each data point Then, construct the matrix.

[0115] It can be understood that the long short-term memory network in this step introduces a complex internal structure, including a memory unit, a forget gate, an input gate, and an output gate, so that the model can maintain effective information transmission over a longer time span.

[0116] The long short-term memory network is composed of a chain structure of multiple long short-term memory network units, and the information output of the previous time step will be used as the operation input of the current time step. As the key information obtained by the local linear embedding method is used as the time series input of the current time step, after the operation processing of the long short-term memory network unit, the aging status of the high-speed rail battery pack is finally obtained. The state of the memory unit will be gradually updated with each time step of the sequence. New information can be written and old information can be cleared by "forgetting". The forget gate determines whether information is removed from the memory unit. In each time step, LSTM reads the current input and the hidden state of the previous time step, and outputs a value of 0 to 1 through the sigmoid function, corresponding to the proportion of forgetting. Information with an output value close to 1 is retained; information close to 0 is forgotten. The calculation formula of the forget gate is:

[0117]

[0118] Where, the output of the forget gate at the current moment is , is the weight matrix of the forget gate, is the hidden state of the previous time step, is the input of the current time step, Bias vector for the forget gate.

[0119] The output gate determines how the memory cell state at the current time step affects the output hidden state, controlling the output of the memory cell's contents at the current time step. After the forget gate executes, the input gate determines which of the input information at the current time step is added to the memory cell. LSTM uses the sigmoid and tanh functions to process the current input, generating a vector of new information and adding it to the memory cell state to ensure that only information associated with the current context is updated. The calculation process is:

[0120]

[0121] Where, is the value of the input gate, is the weight matrix of the input gate, is the hidden state of the previous time step, is the input of the current time step, is the bias vector of the input gate, is the candidate memory cell value, is the weight matrix of the memory unit, is the bias vector of the memory unit, To update the memory cell state, is the storage state of the memory unit at the previous time step.

[0122] The output gate determines the hidden state of the current time step, that is, the output value of the LSTM. After processing by the output gate, the information of the memory cell state is selectively output to the hidden state and used as the input of the next time step, thus completing the information transfer of the current time step. The calculation formula of the output gate is:

[0123] The output gate determines the hidden state of the current time step, that is, the output value of the LSTM. After processing by the output gate, the information of the memory cell state is selectively output to the hidden state and used as the input of the next time step, thus completing the information transfer of the current time step. The calculation formula of the output gate is:

[0124]

[0125] Where, is the value of the output gate, is the weight matrix of the output gate, is the hidden state of the previous time step, is the input of the current time step, is the bias vector of the output gate.

[0126] The memory cell state is then processed through the tanh activation function and then compared with the value of the output gate. Multiply to get the final hidden state output :

[0127]

[0128] Where, is the hidden state at the current time step, is the value of the output gate, To update the memory unit status.

[0129] Among them, the attention mechanism has three ternary variables , , , the above variables are obtained by inputting By multiplying with the weight matrix. , , To complete the self-attention layer mapping, the self-attention score of each element is used and The value is calculated by dot product; the self-attention score is processed by softmax function to obtain the self-attention weight of each element; the attention weight is obtained by The output of the attention mechanism ensemble is obtained through a complete connection layer and is calculated as follows:

[0130]

[0131] In the formula, the matrix , for The input sequence length is is the dimension of each position in the sequence; , , Indicates the indivual , , The corresponding weight matrix; for The corresponding weight vector; For output, For the The dimension of a key.

[0132] It can be understood that the number of nodes in the input layer of the long short-term memory network-attention mechanism is consistent with the number of low-dimensional embeddings after local linear embedding dimensionality reduction, and the output layer is the battery health status (SOH) prediction value.

[0133] S400: The first aging state and the second aging state are treated as two independent evidence bodies, and the weight of the Dempster combination rule is dynamically adjusted based on the conflict factor to complete the joint confidence allocation of the two independent evidence bodies, and then the final aging state is determined by the maximum confidence criterion.

[0134] It can be understood that step S400 includes S401, S402, S403 and S404, wherein:

[0135] S401, take the first aging state and the second aging state as independent evidence bodies, and set the confidence of the first aging state to be , the confidence of the second aging state is ;

[0136] S402. Calculate the conflict factor using Dempster's combination rule , and dynamically adjust the Dempster combination rule weight based on the size of the conflict factor, where the conflict factor The calculation formula is as follows:

[0137]

[0138] Where, is the conflict factor, is an empty set, 、 Represent the first aging state and the second aging state respectively, is the confidence of the first aging state, is the confidence level of the second aging state;

[0139] S403. Calculate the joint confidence of each aging state according to the adjusted Dempster combination rule. The calculation formula is as follows:

[0140]

[0141] Where, is the joint confidence, is the conflict factor, 、 、 Respectively represent the final output aging state, the first aging state and the second aging state, is the confidence of the first aging state, is the confidence level of the second aging state;

[0142] S404: Arrange the joint confidence of each aging state from high to low, and select the state with the highest confidence as the final aging state for output.

[0143] It should be noted that the weight of the combination rules is dynamically adjusted according to the size of the conflict factor. When the conflict factor is large, the combination weight is reduced to avoid unreasonable results caused by the conflict.

[0144] The final aging state is determined by the maximum confidence criterion to determine the final aging state output, calculate the joint confidence of each aging state, and select the state with the highest confidence as the final result output.

[0145] It should be noted that in this embodiment, a complex operating-condition aging test for high-speed train batteries was conducted, simulating actual operating conditions such as different temperatures (e.g., -20°C, 0°C, 25°C, 40°C), different loads (e.g., traction loads of varying power, auxiliary loads, and impact loads), and different charge and discharge modes (e.g., constant current charging, constant voltage charging, long-term float charging, and pulse discharge). During the experiment, battery performance data was recorded at different cycle times, including but not limited to the following multi-physical characteristic parameters: battery voltage, current, and impedance; battery surface temperature field distribution and temperature rise time; and electrode material stress and strain parameters.

[0146] Next, the experimental data was recorded and organized into a database using data logging and preprocessing. Preprocessing was also performed to ensure accuracy and consistency, including outlier removal and normalization. Through complex aging tests on high-speed train batteries, data was collected on the electrical characteristics (voltage, current, impedance), internal electrochemical heat generation and heat transfer, battery temperature rise, and battery material reaction, decomposition, and mechanical deformation of high-speed train lithium-ion batteries.

[0147] Subsequently, voltage-current-time curves, capacity decay curves, electrochemical impedance spectroscopy characteristic frequency curves, capacity increment curves, battery surface temperature field distribution, and electrode material stress-strain curves were established. This foundation allowed for the extraction of battery characteristics and the exploration of battery aging patterns. The characteristic parameter correlation analysis employed the Spearman analysis method to quantify the correlation between battery performance parameters (such as capacity decay rate and internal resistance change) and multiple physical characteristic parameters. Characteristic parameters with a significance level above 0.8 were selected for subsequent model construction.

[0148] Then, a DP model based on the equivalent circuit is constructed, and the aging mechanism model is constructed using the Arrhenius equation and the quasi-two-dimensional model theory. The equation is constructed, and then the experimental data is used to complete the solution of the temperature parameters. The above coupling equations are solved by numerical simulation methods to obtain the mechanism model prediction results of the battery aging state.

[0149] Finally, the local linear embedding method is used to extract the value information of multiple physical features and use it as the input sequence of the long short-term memory network.

[0150] The input of the LSTM model is a sequence , the output is a sequence In the attention mechanism, the original gate calculation in LSTM can be Set as , is the output of the self-attention mechanism. , , the matrix with sequence Splicing generates a new forget gate:

[0151]

[0152] The output gate is:

[0153]

[0154] The input gate is:

[0155]

[0156] The candidate memory cell values ​​are:

[0157]

[0158] Output of new time series through attention mechanism Perform weighted processing, where the weighted sequence , the weight is calculated as:

[0159]

[0160] Then have:

[0161]

[0162] The number of nodes in the input layer of the long short-term memory network-attention mechanism is consistent with the number of low-dimensional embeddings after local linear embedding dimensionality reduction, and the output layer It is the battery state of health (SOH) prediction value.

[0163] The aging status confidence output by the mechanism model and the aging probability distribution output by the data model are regarded as independent evidence bodies. The weight of the Dempster combination rule is dynamically adjusted based on the conflict factor to complete the joint confidence allocation of the two types of evidence, and the final aging status is determined by the maximum confidence criterion.

[0164] The aging state confidence output by the mechanism model and the aging probability distribution output by the data model are taken as independent evidence bodies. The aging state confidence output by the mechanism model is assumed to be , the confidence of the aging state output by the data model is By using the Dempster combination rule and dynamically adjusting the weight of the Dempster combination rule based on the conflict factor, the joint confidence allocation of the two types of evidence is completed.

[0165] Finally, the weights of the combined rules are dynamically adjusted based on the size of the conflict factor. When the conflict factor is large, the combined weight is reduced to avoid unreasonable results caused by the conflict. The final aging state is determined by the maximum confidence criterion, and the joint confidence of each aging state is calculated. The state with the highest confidence is selected as the final result.

[0166] Example 2:

[0167] like Figure 2 As shown, this embodiment provides a mechanism and data fusion EMU battery aging assessment system, see Figure 2 The system comprises:

[0168] Extraction and processing module 701: used to simulate the operating conditions of different temperatures, different loads, and different charge and discharge modes in actual operation through high-speed train battery aging experiments, perform electrochemical impedance spectroscopy testing and open circuit voltage measurement on the battery, record experimental data during the battery aging process, extract and process multiple physical characteristic parameters of the battery from the experimental data, and obtain processed first structured data parameters and first unstructured data parameters;

[0169] First prediction module 702: used to physically decompose the first unstructured data parameters into second structured data parameters, fuse the first structured data parameters with the second structured data parameters, and combine the Arrhenius equation, quasi-two-dimensional model, and equivalent circuit model to construct a high-speed train battery aging mechanism model with electrical, thermal, and mechanical multi-physical characteristics. The high-speed train battery aging mechanism model is solved using a numerical simulation method to predict the first aging state of the battery.

[0170] Second prediction module 703: used to decompose and extract key information of multi-physical feature parameters using a local linear embedding method, input the key information as a sequence into the long short-term memory network and self-attention mechanism, establish a mapping relationship between battery performance and multi-physical features, generate a deep learning model for battery aging big data and train it, and obtain a predicted output of the second aging state, where the key information includes key features, local structure information, and key features after dimensionality reduction;

[0171] Determination evaluation module 704: used to treat the first aging state and the second aging state as two independent evidence bodies, and dynamically adjust the Dempster combination rule weight based on the conflict factor to complete the joint confidence allocation of the two independent evidence bodies, and then determine the final aging state through the maximum confidence criterion.

[0172] Specifically, in the extraction processing module 701, the processing of the processed first structured data parameters includes:

[0173] The conversion unit is used to perform rank conversion on the battery performance parameters and multi-physical characteristic parameters obtained from the high-speed train battery aging test, including: sorting the values ​​of each variable from small to large and assigning a rank to each value; if there are identical values, the average rank of the identical values ​​appearing in the variable value sorting process is assigned;

[0174] A division unit: used to divide the parameters after rank conversion into a first rank and a second rank, wherein the first rank is the rank of the battery performance parameter and the second rank is the rank of the multi-physical characteristic parameter;

[0175] A calculation and screening unit is used to calculate the Spearman correlation coefficient between each pair of performance parameters and multi-physical feature parameters based on the first rank and the second rank, and screen the multi-physical feature parameters greater than 0.8, and record them as the processed first structured data parameters. The calculation formula of the Spearman correlation coefficient is:

[0176]

[0177] Where, is the correlation coefficient, is the battery performance parameter without rank conversion, are the multi-physical characteristic parameters that have not been rank-transformed, is the first rank, is the second rank, is the sample size, is the number of battery performance parameters, is the number of physical characteristic parameters.

[0178] Specifically, in the extraction processing module 701, the processing of the processed first unstructured data parameter includes:

[0179] Processing and extraction unit: used to obtain the Nyquist diagram in the electrochemical impedance spectroscopy test, perform differential geometry processing on the Nyquist diagram, and extract the characteristic frequency set of the curvature mutation point. The calculation formula is as follows:

[0180]

[0181] Where, θ is the angle between the tangent direction at the corresponding frequency point and the real axis, is the set of characteristic frequencies extracted from the Nyquist plot of electrochemical impedance spectroscopy, For the angle θ The corresponding characteristic frequency;

[0182] Decomposition unit: used to construct a dynamic weighted tensor based on the characteristic frequency set, decompose the open circuit voltage curve into the time domain differential intrinsic mode function, and record the time domain differential intrinsic mode function as the first unstructured data parameter after processing. The calculation formula is as follows:

[0183]

[0184] Where, is a dynamic weighted tensor, is the characteristic frequency corresponding to the angle θ, τ is the charge and discharge cycle, is the charge transfer resistance at angle θ, is the double layer capacitance at angle θ.

[0185] Specifically, the first prediction module 702 includes:

[0186] A generation unit is used to calculate the mean, variance, energy, and time-frequency characteristics of each mode in the first unstructured data parameter, and associate the battery health state with the aging sensitivity weighted coefficient to generate a second structured data parameter reflecting the microscopic aging mechanism;

[0187] Fusion unit: used to fuse the first structured data parameter and the second structured data parameter using a multimodal fusion algorithm. The calculation formula is as follows:

[0188]

[0189] Where, X fused As the fusion result, and are the first dynamic weight matrix and the second dynamic weight matrix, X struct is the first structured data parameter, X IMF is the second structuring parameter, It is Hadamard.

[0190] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0191] Example 3:

[0192] Corresponding to the above method embodiment, this embodiment also provides a EMU battery aging assessment device with mechanism and data fusion. The EMU battery aging assessment device with mechanism and data fusion described below and the EMU battery aging assessment method with mechanism and data fusion described above can be referenced to each other.

[0193] Figure 3 FIG. 8 is a block diagram of a device 800 for evaluating battery aging of a train set according to an exemplary embodiment, showing a mechanism and data fusion. Figure 3 As shown, the mechanism and data fusion EMU battery aging assessment device 800 includes: a processor 801 and a memory 802. The mechanism and data fusion EMU battery aging assessment device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0194] The processor 801 is used to control the overall operation of the mechanism and data fusion EMU battery aging assessment device 800 to complete all or part of the steps in the mechanism and data fusion EMU battery aging assessment method described above. The memory 802 is used to store various types of data to support the operation of the mechanism and data fusion EMU battery aging assessment device 800. Such data may include, for example, instructions for any application or method operating on the mechanism and data fusion EMU battery aging assessment device 800, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be a keyboard, a mouse or buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the EMU battery aging assessment device 800 and other devices that integrate the mechanism and data. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: a Wi-Fi module, a Bluetooth module or an NFC module.

[0195] In an exemplary embodiment, the mechanism and data fusion EMU battery aging assessment device 800 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned mechanism and data fusion EMU battery aging assessment method.

[0196] In another exemplary embodiment, a computer-readable storage medium including program instructions is further provided. When executed by a processor, the program instructions implement the steps of the aforementioned method for evaluating EMU battery aging using mechanism and data fusion. For example, the computer-readable storage medium may be the aforementioned memory 802 including the program instructions. The program instructions may be executed by the processor 801 of the EMU battery aging evaluation device 800 using mechanism and data fusion to implement the aforementioned method for evaluating EMU battery aging using mechanism and data fusion.

[0197] Example 4:

[0198] Corresponding to the above method embodiment, this embodiment further provides a readable storage medium. The readable storage medium described below and the EMU battery aging assessment method with mechanism and data fusion described above can refer to each other.

[0199] A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, the steps of the EMU battery aging assessment method with the mechanism and data fusion of the above method embodiment are implemented.

[0200] The readable storage medium may specifically be any readable storage medium that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0201] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0202] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for evaluating battery aging in electric trains by integrating mechanism and data, characterized in that: include: Through the high-speed train battery aging experiment, the operating conditions of different temperatures, different loads, and different charge and discharge modes in actual operation are simulated. The battery is tested for electrochemical impedance spectroscopy and open circuit voltage. The experimental data of the battery aging process is recorded. The multi-physical characteristic parameters of the battery in the experimental data are extracted and processed to obtain the processed first structured data parameters and first unstructured data parameters. Physically decomposing the first unstructured data parameters into second structured data parameters, fusing the first and second structured data parameters, and combining them with the Arrhenius equation, quasi-two-dimensional model, and equivalent circuit model to construct a high-speed train battery aging mechanism model with electrical, thermal, and mechanical multi-physical characteristics. The high-speed train battery aging mechanism model is solved using numerical simulation methods to predict the first aging state of the battery. A local linear embedding method is used to decompose and extract key information of multi-physical feature parameters. This key information is input as a sequence into a long short-term memory network and a self-attention mechanism. A mapping relationship between battery performance and multi-physical features is established. A deep learning model for battery aging big data is generated and trained to obtain the predicted output of the second aging state. The key information includes key features, local structural information, and key features after dimensionality reduction. The first aging state and the second aging state are treated as two independent evidence bodies, and the weight of the Dempster combination rule is dynamically adjusted based on the conflict factor to complete the joint confidence allocation of the two independent evidence bodies, and then the final aging state is determined by the maximum confidence criterion; Among them, include: The first aging state and the second aging state are regarded as independent evidence bodies, and the confidence of the first aging state is , the confidence of the second aging state is ; Using Dempster's combination rule, calculate the conflict factor , and dynamically adjust the Dempster combination rule weight based on the size of the conflict factor, where the conflict factor The calculation formula is as follows: Where, is the conflict factor, is an empty set, 、 Represent the first aging state and the second aging state respectively, is the confidence of the first aging state, is the confidence level of the second aging state; According to the adjusted Dempster combination rule, the joint confidence of each aging state is calculated, and the calculation formula is as follows: Where, is the joint confidence, is the conflict factor, 、 、 Respectively represent the final output aging state, the first aging state and the second aging state, is the confidence of the first aging state, is the confidence level of the second aging state; The joint confidence of each aging state is arranged from high to low, and the state with the highest confidence is selected as the final aging state for output.

2. The EMU battery aging assessment method based on mechanism and data fusion according to claim 1 is characterized in that: The multi-physical characteristic parameters of the battery in the experimental data are extracted and processed to obtain a processed first structured data parameter and a first unstructured data parameter, wherein the processing process of the processed first structured data parameter includes: The battery performance parameters and multi-physical characteristic parameters obtained from the high-speed train battery aging test were rank-transformed, including: sorting the values ​​of each variable from small to large and assigning a rank to each value; if there are identical values, the average rank of the identical values ​​in the variable value sorting process is assigned; Dividing the rank-converted parameters into a first rank and a second rank, wherein the first rank is the rank of the battery performance parameters and the second rank is the rank of the multi-physical characteristic parameters; Based on the first rank and the second rank, the Spearman correlation coefficient between each pair of performance parameters and multi-physical feature parameters is calculated, and multi-physical feature parameters greater than 0.8 are selected and recorded as the first structured data parameters after processing. The calculation formula of the Spearman correlation coefficient is: Where, is the correlation coefficient, is the battery performance parameter without rank conversion, are the multi-physical characteristic parameters that have not been rank-transformed, is the first rank, is the second rank, is the sample size, is the number of battery performance parameters, and j is the number of physical parameters.

3. The EMU battery aging assessment method based on mechanism and data fusion according to claim 1 is characterized in that: The multi-physical characteristic parameters of the battery in the experimental data are extracted and processed to obtain a processed first structured data parameter and a first unstructured data parameter, wherein the processing process of the processed first unstructured data parameter includes: Obtain the Nyquist plot from the electrochemical impedance spectroscopy test, perform differential geometry processing on the Nyquist plot, and extract the characteristic frequency set of the curvature mutation point. The calculation formula is as follows: Where, is the angle between the tangent direction at the corresponding frequency point and the real axis, is the set of characteristic frequencies extracted from the Nyquist plot of electrochemical impedance spectroscopy, For the angle The corresponding characteristic frequency; A dynamic weighted tensor is constructed based on the characteristic frequency set, and the open circuit voltage curve is decomposed into the time-domain differential intrinsic mode function. The time-domain differential intrinsic mode function is recorded as the first unstructured data parameter after processing. Its calculation formula is as follows: Where, is a dynamic weighted tensor, For the angle The corresponding characteristic frequency, is the charge and discharge cycle, Angle The charge transfer resistance at Angle The double layer capacitance at .

4. The EMU battery aging assessment method based on mechanism and data fusion according to claim 1 is characterized in that: The physically decomposing the first unstructured data parameter into the second structured data parameter and fusing the first structured data parameter with the second structured data parameter includes: Calculating the mean, variance, energy, and time-frequency characteristics of each mode in the first unstructured data parameter, and correlating the battery health state through the aging sensitivity weighting coefficient to generate a second structured data parameter reflecting the microscopic aging mechanism; The first structured data parameter and the second structured data parameter are fused using a multimodal fusion algorithm, and the calculation formula is as follows: Where, X fused For the fusion result, and are the first dynamic weight matrix and the second dynamic weight matrix, X struct is the first structured data parameter, X IMF is the second structuring parameter, It is Hadamard.

5. The EMU battery aging assessment method based on mechanism and data fusion according to claim 1 is characterized in that: The local linear embedding method is used to decompose and extract key information of multiple physical feature parameters, including: Based on the data of multiple physical feature parameters, the neighborhood size of each data point is determined by Euclidean distance to obtain domain information; Combined with the neighborhood information, find the r nearest neighbor points for each data point and calculate the corresponding reconstruction weight. The calculation formula is as follows: Where, To reconstruct the weights, is the identity matrix, N is the weight matrix, is the transpose of the matrix; The calculated reconstruction weights are used to construct a local neighborhood matrix for each data point, where each row represents the difference between a neighboring point and the data point. The calculation formula for the local neighborhood matrix is ​​as follows: For each data point , calculated by the formula : Where, is the local neighborhood matrix, For all neighboring points, for point The reconstruction weight of Based on the local neighborhood matrix, the low-dimensional embedding results are extracted by solving the least squares problem and matrix eigenvalue and eigenvector analysis, and the low-dimensional embedding results are recorded as key information.

6. A mechanism and data fusion EMU battery aging assessment system, based on the mechanism and data fusion EMU battery aging assessment method according to claim 1, characterized in that: include: Extraction and processing module: used to simulate the operating conditions of different temperatures, different loads, and different charge and discharge modes in actual operation through high-speed train battery aging experiments, perform electrochemical impedance spectroscopy testing and open circuit voltage measurement on the battery, record experimental data during the battery aging process, extract and process the multi-physical characteristic parameters of the battery in the experimental data, and obtain the processed first structured data parameters and first unstructured data parameters; The first prediction module is used to physically decompose the first unstructured data parameters into second structured data parameters, fuse the first structured data parameters with the second structured data parameters, and combine the Arrhenius equation, quasi-two-dimensional model, and equivalent circuit model to construct a high-speed train battery aging mechanism model with electrical, thermal, and mechanical multi-physical characteristics. The high-speed train battery aging mechanism model is solved using a numerical simulation method to predict the first aging state of the battery. The second prediction module is used to decompose and extract key information of multi-physical feature parameters using a local linear embedding method. The key information is input as a sequence into the long short-term memory network and the self-attention mechanism to establish a mapping relationship between battery performance and multi-physical features. A deep learning model of battery aging big data is generated and trained to obtain the predicted output of the second aging state. The key information includes key features, local structural information, and key features after dimensionality reduction. Determine the evaluation module: It is used to treat the first aging state and the second aging state as two independent evidence bodies, and dynamically adjust the Dempster combination rule weight based on the conflict factor to complete the joint confidence allocation of the two independent evidence bodies, and then determine the final aging state through the maximum confidence criterion; which specifically includes: treating the first aging state and the second aging state as independent evidence bodies, assuming the confidence of the first aging state is , the confidence of the second aging state is ; Using Dempster's combination rule, calculate the conflict factor , and dynamically adjust the Dempster combination rule weight based on the size of the conflict factor, where the conflict factor The calculation formula is as follows: Where, is the conflict factor, is an empty set, 、 Represent the first aging state and the second aging state respectively, is the confidence of the first aging state, is the confidence level of the second aging state; According to the adjusted Dempster combination rule, the joint confidence of each aging state is calculated, and the calculation formula is as follows: Where, is the joint confidence, is the conflict factor, 、 、 Respectively represent the final output aging state, the first aging state and the second aging state, is the confidence of the first aging state, is the confidence level of the second aging state; The joint confidence of each aging state is arranged from high to low, and the state with the highest confidence is selected as the final aging state for output.

7. The EMU battery aging assessment system based on mechanism and data fusion according to claim 6 is characterized in that: The extraction processing module, wherein the processing process of the processed first structured data parameter includes: The conversion unit is used to perform rank conversion on the battery performance parameters and multi-physical characteristic parameters obtained from the high-speed train battery aging test, including: sorting the values ​​of each variable from small to large and assigning a rank to each value; if there are identical values, the average rank of the identical values ​​appearing in the variable value sorting process is assigned; A division unit: used to divide the parameters after rank conversion into a first rank and a second rank, wherein the first rank is the rank of the battery performance parameter and the second rank is the rank of the multi-physical characteristic parameter; A calculation and screening unit is used to calculate the Spearman correlation coefficient between each pair of performance parameters and multi-physical feature parameters based on the first rank and the second rank, and screen the multi-physical feature parameters greater than 0.8, and record them as the processed first structured data parameters. The calculation formula of the Spearman correlation coefficient is: Where, is the correlation coefficient, is the battery performance parameter without rank conversion, are the multi-physical characteristic parameters that have not been rank-transformed, is the first rank, is the second rank, is the sample size, is the number of battery performance parameters, and j is the number of physical characteristic parameters.

8. The EMU battery aging assessment system based on mechanism and data fusion according to claim 6 is characterized in that: The extraction processing module, wherein the processing process of the processed first unstructured data parameter includes: Processing and extraction unit: used to obtain the Nyquist diagram in the electrochemical impedance spectroscopy test, perform differential geometry processing on the Nyquist diagram, and extract the characteristic frequency set of the curvature mutation point. The calculation formula is as follows: Where, is the angle between the tangent direction at the corresponding frequency point and the real axis, is the set of characteristic frequencies extracted from the Nyquist plot of electrochemical impedance spectroscopy, For the angle The corresponding characteristic frequency; Decomposition unit: used to construct a dynamic weighted tensor based on the characteristic frequency set, decompose the open circuit voltage curve into the time domain differential intrinsic mode function, and record the time domain differential intrinsic mode function as the first unstructured data parameter after processing. The calculation formula is as follows: Where, is a dynamic weighted tensor, For the angle The corresponding characteristic frequency, is the charge and discharge cycle, Angle The charge transfer resistance at Angle The double layer capacitance at .

9. The EMU battery aging assessment system based on mechanism and data fusion according to claim 6 is characterized in that: The first prediction module includes: A generation unit is used to calculate the mean, variance, energy, and time-frequency characteristics of each mode in the first unstructured data parameter, and associate the battery health state with the aging sensitivity weighted coefficient to generate a second structured data parameter reflecting the microscopic aging mechanism; Fusion unit: used to fuse the first structured data parameter and the second structured data parameter using a multimodal fusion algorithm. The calculation formula is as follows: Where, X fused For the fusion result, and are the first dynamic weight matrix and the second dynamic weight matrix, X struct is the first structured data parameter, X IMF is the second structuring parameter, It is Hadamard.

Images