Multi-dimensional Testing Method and Device for Golf Cart Batteries

Abstract

The present invention relates to a multi-dimensional testing method and device for a golf cart battery. The method includes: collecting standard operating condition data of the golf cart battery, constructing an electrochemical-thermal-mechanical coupling theoretical model, and obtaining a battery multi-dimensional coupling theoretical matrix; setting temperature sensors and strain gauges on the surface of the golf cart battery and performing real-time data synchronous acquisition to obtain battery multi-dimensional measured data, and conducting an electrochemical impedance spectroscopy test on the battery to obtain a battery impedance characteristic vector; discretizing the battery volume into control volume units and performing thermal-mechanical coupling iterative solution to obtain a battery thermal-mechanical coupling characteristic matrix; and performing battery performance evaluation based on the battery impedance characteristic vector, the battery thermal-mechanical coupling characteristic matrix, and the battery multi-dimensional measured data to obtain a battery performance evaluation result. The present invention can comprehensively capture the interaction between various physical fields inside the battery, and realizes real-time monitoring and rapid response of battery multi-dimensional physical quantities.

Description

Technical Field

[0001] The present invention relates to the technical field of battery testing, and particularly relates to a multi-dimensional testing method and device for golf cart batteries. Background Art

[0002] Golf carts are increasingly widely used as short-distance transportation tools, and the performance testing and state evaluation of their core power source batteries have become key links to ensure the safety of use. At present, the testing methods for golf cart batteries mainly focus on the measurement and analysis of single physical quantities, such as the independent monitoring of parameters such as voltage, current, and temperature. This testing method ignores the coupling effect between various physical fields inside the battery, resulting in the test results being unable to comprehensively reflect the true working state of the battery.

[0003] Traditional battery testing methods have three main problems: there are multi-field coupling effects of electrochemical reactions, heat conduction, and mechanical stress during the actual operation of the battery. These physical fields influence and restrict each other, and the measurement of a single physical quantity cannot accurately describe the performance characteristics of the battery; secondly, existing test data often contains a large amount of interference information related to working conditions, and there is a lack of effective feature decoupling methods, making it difficult to extract characteristic indicators reflecting the intrinsic performance of the battery; there is a lack of a systematic cooperation mechanism in the process of collecting and analyzing test data, resulting in information redundancy and inconsistency between various test results. Summary of the Invention

[0004] The main object of the present invention is to provide a multi-dimensional testing method and device for golf cart batteries. The present invention can comprehensively capture the interaction between various physical fields inside the battery, and realizes the real-time monitoring and rapid response of multi-dimensional physical quantities of the battery.

[0005] To achieve the above object, the present invention provides a multi-dimensional testing method for golf cart batteries, including the following steps:

[0006] Collect the standard working condition operation data of the golf cart battery, construct an electrochemical-thermal-mechanical coupling theory model, and obtain the battery multi-dimensional coupling theory matrix;

[0007] Set temperature sensors and strain gauges on the surface of the golf cart battery and perform real-time data synchronous acquisition to obtain the battery multi-dimensional measured data, and perform battery electrochemical impedance spectroscopy testing to obtain the battery impedance characteristic vector;

[0008] According to the battery multi-dimensional measured data and the battery multi-dimensional coupling theory matrix, discretize the battery volume into control volume units and perform thermal-mechanical coupling iterative solution to obtain the battery thermal-mechanical coupling characteristic matrix;

[0009] Based on the battery impedance characteristic vector, the battery thermo-mechanical coupling characteristic matrix, and the multi-dimensional measured data of the battery, battery performance evaluation is performed to obtain a battery performance evaluation result.

[0010] The present invention also provides a multi-dimensional testing device for a golf cart battery, including:

[0011] A data acquisition unit for collecting standard operating condition data of the golf cart battery, constructing an electrochemistry-thermodynamics-mechanics coupling theory model, and obtaining a battery multi-dimensional coupling theory matrix;

[0012] A testing unit for setting temperature sensors and strain gauges on the surface of the golf cart battery and performing real-time data synchronization acquisition to obtain multi-dimensional measured data of the battery, and performing an electrochemical impedance spectroscopy test on the battery to obtain a battery impedance characteristic vector;

[0013] An iterative solution unit for discretizing the battery volume into control volume units and performing thermo-mechanical coupling iterative solution according to the multi-dimensional measured data of the battery and the battery multi-dimensional coupling theory matrix to obtain a battery thermo-mechanical coupling characteristic matrix;

[0014] A performance evaluation unit for performing battery performance evaluation based on the battery impedance characteristic vector, the battery thermo-mechanical coupling characteristic matrix, and the multi-dimensional measured data of the battery to obtain a battery performance evaluation result.

[0015] In summary, the technical solution provided by the present invention realizes the accurate measurement of the multi-physical field coupling characteristics of the battery: by establishing an electrochemistry-thermodynamics-mechanics three-field coupling model and combining with the real-time data acquisition of the distributed sensing network, the interaction between various physical fields inside the battery can be comprehensively captured, improving the accuracy and integrity of the test data. An effective decoupling solution for working condition interference is provided: by adopting a feature extraction method combining contrast learning and adversarial training, the working condition-related features and the battery intrinsic performance features are successfully separated, enhancing the reliability and generality of the test results. A multi-dimensional collaborative testing system is established: through thermo-mechanical coupling iterative solution and multi-scale feature fusion, the unified evaluation of various performance indicators of the battery is realized, avoiding the information redundancy and inconsistency problems in traditional testing methods. The accuracy of performance evaluation is optimized: based on the comprehensive analysis of the battery impedance characteristics, thermo-mechanical coupling characteristics, and measured data, a complete performance evaluation index system is constructed, improving the accuracy and predictability of battery performance evaluation. The real-time performance of the testing process is enhanced: through the distributed data acquisition and synchronous control strategy, the real-time monitoring and rapid response of the multi-dimensional physical quantities of the battery are realized. Description of the Drawings

[0016] Figure 1 It is a schematic diagram of the steps of a multi-dimensional testing method for a golf cart battery in an embodiment of the present invention;

[0017] Figure 2 It is a structural block diagram of a multi - dimensional test device for a golf cart battery in an embodiment of the present invention.

[0018] The realization of the object, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Specific embodiments

[0019] In order to make the object, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] Refer to Figure 1 , this embodiment provides a multi - dimensional test method for a golf cart battery, including the following steps:

[0021] S1, collect the standard operating condition data of the golf cart battery, construct an electro - chemical - thermal - mechanical coupling theoretical model, and obtain a battery multi - dimensional coupling theoretical matrix;

[0022] Among them, the operation data of the golf cart battery under different charge and discharge rate conditions are collected, including the voltage, current, and temperature distribution data at 0.5C, 1C, and 2C rates. These data constitute the original dataset of the battery operation. The original dataset is preprocessed to remove outliers, perform interpolation and completion, and execute normalization processing to obtain the battery standardized dataset. Based on the standardized dataset, an electrochemical-thermal-mechanical coupling theoretical model of the battery is constructed. The battery structure is divided into five layers, namely the negative current collector, negative electrode, separator, positive electrode, and positive current collector. This hierarchical division can describe the electrochemical characteristics of each component inside the battery, and then establish the ion transport equations for each layer. The ion transport equation adopts the Poisson-Nernst-Planck equation or Fick's diffusion equation to describe the ion migration behavior in the electrode and electrolyte, and combines the charge conservation relationship of the battery to derive a complete electrochemistry reaction characteristic matrix, which is used to characterize the potential distribution, charge accumulation, and current density change inside the battery. According to the electrochemistry reaction characteristic matrix, a three-dimensional unsteady heat conduction equation is established to describe how the heat generated during the charge and discharge process of the battery is transferred and diffused inside. The heat sources inside the battery mainly include ohmic heating, polarization heat, and side reaction heat, and these heat source terms are directly related to the electrochemistry reaction. Therefore, a mapping relationship is established between the heat source terms and the electrochemistry reaction to obtain the heat conduction characteristic matrix. To achieve this goal, the battery is meshed, and the battery structure is discretized into a finite number of small units. The temperature change of each unit is controlled by the three-dimensional unsteady heat conduction equation. This equation is based on Fourier's law of heat conduction and combines the thermal physical properties of the battery materials, such as thermal conductivity, specific heat capacity, and density, to calculate the distribution of the temperature field. The equation is solved by numerical methods (such as the finite element method or finite difference method) to obtain the heat conduction characteristic matrix of each layer of the battery, which is used to describe the temperature distribution and heat flow path of the battery under different working conditions. Based on the heat conduction characteristic matrix, the volume change amount of the electrode material during the charge and discharge process is calculated, and a stress distribution function is established in combination with the mechanical constitutive relationship. This function is used to describe the mechanical stress distribution in different regions inside the battery, and thus the mechanical stress characteristic matrix is derived. The mechanical stress characteristic matrix reveals problems such as deformation, swelling, and even cracks that occur during the cyclic use of the battery. The electrochemistry reaction characteristic matrix, heat conduction characteristic matrix, and mechanical stress characteristic matrix are input into a multi-physics field simulation solver for multi-field coupling calculation to obtain the multi-dimensional coupling theoretical matrix of the battery. The multi-physics field simulation solver simultaneously solves the electrochemical, thermodynamics, and mechanics equations and considers their mutual influence. For example, the temperature distribution affects the electrochemistry reaction rate of the battery, and the change in the electrochemistry reaction rate will in turn affect the heat release of the battery. At the same time, the change in mechanical stress may lead to a decrease in material performance, which will further affect the electrochemical behavior of the battery.

[0023] S2. Set a temperature sensor and strain gauges on the surface of the golf cart battery and perform real-time data synchronization acquisition to obtain multi-dimensional measured data of the battery, and conduct an electrochemical impedance spectroscopy test on the battery to obtain the battery impedance characteristic vector;

[0024] Specifically, sensors are reasonably arranged on the battery surface to obtain temperature and stress data. During this process, the surface of the golf cart battery is divided into equally spaced grids to ensure the uniform distribution of measurement points. M temperature sensors are evenly arranged at the grid points after division to obtain the temperature distribution information on the battery surface. At the same time, since the battery is affected by mechanical stress caused by thermal expansion, material expansion, and current flow during charge and discharge, a stress analysis is carried out on the stress concentration area of the golf cart battery housing to determine the parts with larger mechanical stress. After completing the stress analysis, N strain gauges are arranged at the target structural positions to measure the mechanical stress and deformation of the battery during operation. The output signals of the temperature sensors and strain gauges are conditioned to eliminate the noise in the signals and improve the measurement accuracy. Since the temperature sensors and strain gauges output analog signals, these analog signals are standardized to ensure the comparability and consistency of the measurement data. The standardized analog signals can eliminate the deviation caused by individual differences of the sensors, thereby improving the reliability of the data. After completing the signal conditioning, the standardized analog signals are digitally converted using analog-to-digital conversion technology to obtain the digital measurement data for subsequent analysis. Based on the digital measurement data, synchronous control is performed on the sampling timing of the temperature sensors and strain gauges to obtain the multi-dimensional measured data of the battery. According to the battery multi-dimensional coupling theory matrix, state-of-charge test points are selected to ensure that the test points represent the characteristics of the battery under different working conditions. The selection of the state-of-charge test points is based on the SOC curve of the battery to cover low, medium, and high state-of-charge conditions, thereby characterizing the electrochemical behavior of the battery. An electrochemical impedance spectroscopy test is conducted on the battery to analyze its impedance characteristics. The electrochemical impedance spectroscopy test obtains the impedance change of the battery at different frequencies by applying a small-amplitude alternating current signal to the battery and measuring its voltage response. By fitting and analyzing the impedance data at different frequencies, the internal equivalent circuit parameters of the battery are extracted, including ohmic resistance, charge transfer resistance, and diffusion impedance, etc., thereby characterizing the electrochemical performance of the battery. Through the above steps, the multi-dimensional measured data of the battery and the battery impedance characteristic vector are obtained.

[0025] Based on the electrochemical reaction characteristic data in the battery multi-dimensional coupling theory matrix, optimize the test conditions and select appropriate state of charge (SOC) test points to ensure that the test of the battery impedance characteristics comprehensively reflects its internal state. Since the electrochemical behavior of the battery is affected by factors such as state of charge (SOC), temperature, charge and discharge rate, etc., when optimizing the test conditions, these factors are comprehensively considered, and by analyzing the equivalent circuit characteristics of the battery, select the state of charge test points that can represent different operating states. Set appropriate AC excitation signals for each state of charge test point to stimulate the electrochemical reaction of the battery in different frequency ranges and form a set of test condition parameters. The set of test condition parameters mainly includes parameters such as the amplitude of the excitation signal, the frequency range, and the measurement time. Based on the set of test condition parameters, apply an AC excitation signal to the golf cart battery for electrochemical impedance spectroscopy (EIS) scanning. Electrochemical impedance spectroscopy scanning analyzes the impedance characteristics of the battery by applying a small-amplitude AC signal to the battery and measuring its voltage response. Apply a sinusoidal current signal at different frequencies and measure the corresponding voltage changes to calculate the AC impedance of the battery. Electrochemical impedance spectroscopy scanning covers multiple frequency points from high frequency to low frequency to fully characterize the internal transport characteristics of the battery. Through this method, the impedance response data of the battery are obtained, including the impedance modulus value and phase angle information at different frequencies. According to the impedance response data, construct an equivalent circuit model to analyze the electrochemical characteristics of the battery. The battery equivalent circuit model consists of multiple resistors, capacitors, and impedance elements, which are used to describe the ohmic loss, charge transfer process, double-layer effect, and ion diffusion behavior inside the battery. The equivalent circuit model includes a series resistance (Rs), which is used to describe the bulk resistance of the battery and the current collector resistance; a charge transfer resistance (Rct), which is used to characterize the electrochemical reaction impedance at the electrode-electrolyte interface; a double-layer capacitance (Cdl), which is used to describe the charge accumulation effect on the electrode surface; and a Warburg impedance (Zw), which is used to characterize the ion diffusion behavior inside the battery. By reasonably constructing the equivalent circuit model, the electrochemical impedance characteristics of the battery are effectively analyzed. Input the impedance response data into the non-linear least squares fitting algorithm to solve the unknown parameters in the equivalent circuit model. The non-linear least squares fitting algorithm minimizes the error between the experimental data and the calculation results of the equivalent circuit model and iteratively optimizes the circuit parameters to obtain the best fitting result. This method accurately extracts the set of equivalent circuit parameters, including key parameters such as Rs, Rct, Cdl, and Zw, and provides a high-precision analysis of the battery impedance characteristics. Conduct a frequency domain characteristic analysis on the set of equivalent circuit parameters to extract the key impedance characteristic vectors. The impedance in the high-frequency region is mainly dominated by the ohmic impedance (Rs), which reflects the internal resistance of the battery and the current collection efficiency; the impedance in the middle-frequency region is mainly determined by the charge transfer impedance (Rct), which reflects the charge transfer kinetics characteristics at the electrode-electrolyte interface; while the impedance in the low-frequency region is mainly dominated by the concentration polarization impedance (Zw), which describes the ion diffusion and polarization behavior inside the battery.By analyzing these frequency-domain characteristic parameters, the impedance characteristic vector of the battery is extracted.

[0026] S3. According to the multi-dimensional measured data of the battery and the multi-dimensional coupling theory matrix of the battery, the volume of the battery is discretized into control volume units and thermal-mechanical coupling iterative solution is carried out to obtain the thermal-mechanical coupling characteristic matrix of the battery;

[0027] It should be noted that three-dimensional grid division is performed on the battery volume using multi-dimensional measured data of the battery to ensure accurate description of the temperature field and stress field distributions during numerical calculations. During grid division, the geometric structure of the battery is discretized into multiple small control volume units, and a unique identification number is assigned to each control volume unit to construct a battery grid model. Based on the multi-dimensional coupling theory matrix of the battery, the initial distribution of the heat conduction field is established, and the measurement data of the temperature sensor is mapped to the grid model using a three-dimensional spline interpolation algorithm to obtain the initial values of the temperature field. The interpolation algorithm is used to perform spatial expansion on the measurement data to reconstruct the complete temperature distribution field inside the battery. The three-dimensional spline interpolation algorithm effectively guarantees the continuity and numerical stability of the temperature field by constructing a high-order smooth interpolation function, and obtains a high-precision initial distribution of the temperature field. Boundary condition processing and thermal-mechanical coupling relationship analysis are performed on the initial values of the temperature field to establish a thermal-mechanical system equation. The setting of boundary conditions needs to consider the actual usage environment of the battery, including factors such as natural convection heat dissipation, radiation heat transfer, and contact thermal resistance with the casing or cooling system. At the same time, since the temperature change inside the battery will cause thermal expansion of the material, and the electrode material will also undergo volume changes due to the insertion and extraction of lithium ions during charge and discharge, the interaction between temperature and mechanical stress is considered in the thermal-mechanical system equation to establish a thermal-mechanical coupling relationship model. This model describes the stress changes of the battery under different temperature field distributions by introducing a thermal stress equation, and couples the heat conduction equation with the stress field equation to enable synchronous calculation of heat transfer and mechanical response. The thermal-mechanical system equation is input into a finite volume method solver to perform alternating iterative calculations of the temperature field and stress field, and a convergence criterion is set during the calculation to ensure the stability and accuracy of the calculation. The finite volume method is a numerical calculation method used for computational fluid dynamics and heat conduction problems. It discretizes the heat conduction equation through the analysis of heat conservation in control volume units and performs iterative calculations using numerical solution methods. During the solution process, the temperature and stress field values of each control volume unit will gradually converge as the number of iterations increases. To determine whether the calculation has reached a convergent state, a convergence criterion is set, that is, when the relative error of the temperature field between adjacent iteration steps is less than a certain target value, the iteration is stopped, and the final thermal-mechanical coupling distribution is obtained. Temperature gradient stress analysis is performed based on the thermal-mechanical coupling distribution to calculate the thermal stress and strain of each control volume unit, and a heat source correction model under the influence of stress is established. Since the battery material will generate additional stress when heated and expanded, this stress will affect the heat conduction path inside the battery, thereby changing the heat distribution of the battery. Therefore, the correction effect of the stress field on the heat source term is calculated through thermal stress analysis, and the heat source term in the thermal-mechanical equation is adjusted to obtain a more realistic thermal-mechanical distribution correction model that reflects the thermal stress influence of the battery under different operating conditions. The temperature field data and stress field data of each control volume unit in the corrected thermal-mechanical distribution are combined and mapped to form a battery thermal-mechanical coupling characteristic matrix.

[0028] The control volume discretization is performed on the equations of the thermal system to transform the continuous partial differential equations into algebraic equation systems suitable for numerical calculation. In the control volume method, the computational domain of the battery is divided into multiple small control volume units, and the discretized representations of the heat conduction equation and the mechanical equilibrium equation are established within each unit. The discretization of the heat conduction equation adopts the finite volume method, which transforms the integral form of the heat flux on the control volume surface into an algebraic relationship, obtaining a discrete algebraic equation containing the thermal conductivity matrix of the temperature field. For the mechanical equilibrium equation, based on the finite element method concept, the deformation and stress of the materials inside the battery are discretized, and a system of equations containing the stiffness matrix of the stress field is established. By combining the control volume method and the finite element method, a complete control volume discretization equation is constructed to describe the relationship between heat conduction and stress distribution inside the battery. Based on the control volume discretization equation, a coupling matrix of the momentum equation and the energy equation is constructed to achieve the joint solution of the temperature field and the stress field. The momentum equation is mainly used to describe the deformation behavior of the materials inside the battery under the action of thermal expansion and mechanical stress, while the energy equation is used to describe the heat conduction and temperature distribution inside the battery. When constructing the coupling matrix, the boundary conditions of the temperature field and the stress field are reasonably processed to ensure the stability and accuracy of the calculation results. The way of processing the boundary conditions depends on the working environment of the battery. For example, if there is a forced cooling system outside the battery, the convective heat transfer coefficient is introduced in the boundary conditions of the energy equation, and if the battery casing is subject to fixed constraints, displacement constraints or stress boundary conditions are imposed in the boundary conditions of the momentum equation. By reasonably setting the boundary conditions, it is ensured that the coupled solution matrix accurately describes the thermodynamic behavior of the battery under actual working conditions. The coupled solution matrix is input into the SIMPLE algorithm iterative solver to calculate the thermal stress increment caused by the temperature field and obtain the temperature stress correction amount. The SIMPLE algorithm is an iterative solution method for computational fluid dynamics and heat transfer problems. It gradually approaches the stable solution by alternately solving the pressure, velocity, and temperature fields. In this embodiment, the SIMPLE algorithm is used to calculate the influence of the temperature field on the stress field and update the interaction between the temperature field and the stress field in each iteration. Specifically, the distribution of the temperature field is calculated by solving the energy equation, and then the mechanical equilibrium equation is solved using this temperature field to calculate the stress increment caused by the temperature change, and further update the stress field distribution. Through continuous iterative calculation, it gradually converges to a stable thermo-mechanical coupling state and obtains the temperature stress correction amount. The temperature field distribution of the control volume unit is corrected according to the temperature stress correction amount, and the velocity-pressure coupling relationship between the stress field and the temperature field is established to ensure the physical consistency of the calculation. Since the change in the temperature field will cause the expansion or contraction of the material, thus affecting the distribution of the stress field, in each iteration step, the temperature field is corrected and the new stress distribution is calculated to ensure the accuracy of the coupled solution.Therefore, a correction coefficient matrix is constructed using the data of the corrected temperature field and the corrected stress field. This matrix is used to describe the interaction between the temperature field and the stress field, and dynamically adjusts the temperature field in subsequent iterative calculations to ensure the convergence and stability of the calculation. Based on the correction coefficient matrix, the relative error of the temperature field between adjacent iterative steps is calculated, and this relative error is compared with a set target value to determine whether the iterative termination criterion is met. At the end of each iterative step, the relative error between the current temperature field and the temperature field of the previous iterative step is calculated, and it is judged whether this error is less than a preset convergence threshold. If the relative error is less than the target value, it is considered that the calculation has converged, the iteration is terminated, and the calculation result is output; otherwise, the iteration needs to continue until the convergence condition is met. By setting reasonable convergence criteria, the accuracy of the calculation result is ensured, and unnecessary waste of computing resources is avoided. After meeting the iterative termination criterion, post-processing is performed on the calculated temperature field and stress field data to extract the temperature distribution and stress distribution data of each control volume unit, and generate a thermo-mechanical coupling distribution. The main contents of post-processing include data visualization, statistical analysis, and joint analysis with other physical field data. For example, heat maps or contour maps are used to display the temperature distribution inside the battery to identify hot spots; at the same time, the stress distribution is analyzed to evaluate the risk of structural damage or fatigue failure that occurs during the long-term cyclic use of the battery.

[0029] S4. Based on the battery impedance characteristic vector, the battery thermo-mechanical coupling characteristic matrix, and the multi-dimensional measured data of the battery, battery performance evaluation is carried out to obtain the battery performance evaluation result.

[0030] Specifically, the battery impedance characteristic vector, the thermo-mechanical coupling characteristic matrix, and the multi-dimensional measured data of the battery are input into the dual-encoder network to fuse and weight-calculate the feature data from different sources. The dual-encoder network consists of two parallel encoder modules, which respectively process the electrochemical characteristics and the thermodynamic and mechanical characteristics of the battery, and learn the operating condition feature representation through a feature weighting mechanism to reduce the influence of the measurement conditions and improve the generalization ability of the model. In this process, the self-attention mechanism is used to dynamically adjust the importance of different features, so that the key features obtain higher weights in the calculation process, ensuring that the operating condition feature representation fully reflects the operating state of the battery. Based on the operating condition feature representation, an InfoNCE contrastive loss function is constructed, and the dual-encoder network is iteratively optimized using a batch normalization layer and an Adam optimizer. The InfoNCE contrastive loss function improves the robustness of the model under different operating conditions by maximizing the similarity of positive samples and minimizing the similarity of negative samples. The batch normalization layer is used to stabilize the network training process, reduce internal covariate shift, and improve the efficiency of feature extraction. The Adam optimizer combines momentum optimization and an adaptive learning rate adjustment mechanism, enabling the network to converge quickly during training and effectively suppressing gradient oscillation. After multiple rounds of iterative training, the model gradually removes the operating condition dependence and extracts stable operating condition-independent features. The operating condition-independent features are input into the discriminator network for feature adversarial training to minimize the accuracy of operating condition prediction, thereby enhancing the operating condition independence. By constructing a weighted combination of the operating condition classification cross-entropy loss function and the feature reconstruction loss function, gradient update optimization is performed to enable the model to effectively remove redundant information related to the operating condition. The role of the discriminator network is to attempt to predict the operating condition category based on the extracted features, while the goal of adversarial training is to make the features generated by the encoder unable to be accurately classified by the discriminator, thereby forcing the encoder to further remove the operating condition information and only retain the essential features related to the battery performance. Through the adversarial training strategy, it is ensured that the finally extracted performance feature vector has strong generalization ability and can maintain consistency under different operating condition conditions, thereby improving the accuracy of battery performance evaluation. Probability distribution constraints are imposed on the stable performance feature vector to ensure its numerical stability and interpretability in the high-dimensional space. The KLD loss function is calculated based on the performance feature vector, and by minimizing the relative entropy between the feature distribution and the target distribution, the features are mapped into the normal distribution space. The role of the KLD loss is to make the feature distribution learned by the model smoother, thereby reducing noise interference and improving the generalization ability of the model. In this process, the method of variational autoencoder is used to constrain the feature distribution to the standard normal distribution, and the reparameterization trick of the latent variable is used to ensure the transitivity of the gradient. By minimizing the KLD loss, a normalized battery feature set is obtained, which can remain stable under different test conditions. Performance metrics are calculated for the normalized battery feature set to quantify the health state of the battery.The performance indicators include the capacity attenuation rate, the internal resistance growth rate, the temperature consistency coefficient, and the stress concentration coefficient. Among them, the capacity attenuation rate is used to describe the degree of capacity loss of the battery during cyclic use and is calculated by comparing the current capacity with the initial capacity; the internal resistance growth rate reflects the increasing trend of the internal ohmic impedance and polarization impedance of the battery and is related to the charge-discharge efficiency and power performance of the battery; the temperature consistency coefficient is used to measure the uniformity of the internal temperature distribution of the battery, and a large temperature gradient will lead to a shortened battery life or an increased safety risk; while the stress concentration coefficient is used to analyze the mechanical stress distribution of the battery structure to predict structural damage or material fatigue. By calculating these key performance indicators, the health state of the battery is evaluated. The capacity attenuation rate, the internal resistance growth rate, the temperature consistency coefficient, and the stress concentration coefficient are input into the degradation mechanism analysis model for time series feature mapping and performance degradation warning. The degradation mechanism analysis model is based on time series modeling methods, such as long short-term memory networks or temporal convolutional networks, to learn the historical data of the battery and predict the future performance degradation trend. By performing time series analysis on indicators such as the capacity attenuation rate and the internal resistance growth rate, the degradation mode of the battery is identified, and its future health state is predicted. And combined with physical modeling methods, such as equivalent circuit models or pseudo-two-dimensional models, the internal chemical and physical processes of the battery are modeled to improve the accuracy of the degradation mechanism analysis. By comprehensively using data-driven and physical modeling methods, high-precision battery performance degradation warning results are generated.

[0031] In one example, the standard operating condition data of a golf cart battery is collected, and an electrochemistry-thermodynamics-mechanics coupling theory model is constructed to obtain a battery multi-dimensional coupling theory matrix, including:

[0032] Collect the voltage, current, and temperature distribution data of the golf cart battery under the charge-discharge conditions of 0.5C, 1C, and 2C rates to obtain the original battery operation dataset, and perform data preprocessing on the original battery operation dataset to obtain the battery standardized dataset;

[0033] Based on the battery standardized dataset, the battery structure is divided into five layers: the negative current collector, the negative electrode, the separator, the positive electrode, and the positive current collector, and ion transport equations for each layer are established to obtain the electrochemistry reaction characteristic matrix;

[0034] According to the electrochemistry reaction characteristic matrix, a three-dimensional unsteady heat conduction equation is established, the golf cart battery is meshed, and a mapping function between the heat source term and the electrochemistry reaction is established to obtain the heat conduction characteristic matrix;

[0035] Based on the heat conduction characteristic matrix, the volume change of the electrode material during the charge-discharge process is calculated, and a stress distribution function is established to obtain the mechanical stress characteristic matrix;

[0036] Input the electro-chemical reaction characteristic matrix, heat conduction characteristic matrix, and mechanical stress characteristic matrix into a multi-physics field simulation solver for coupled calculation to obtain the battery multi-dimensional coupling theory matrix.

[0037] In this example, the operating data of the battery is collected at different charge and discharge rates to ensure that the test results cover a variety of actual usage scenarios. When performing charge and discharge tests at 0.5C, 1C, and 2C rates, the voltage, current, and temperature distribution data of the battery are synchronously recorded to form the original data set of the battery operation. Data preprocessing is performed on the original data set to improve data quality and calculation accuracy. The data preprocessing process includes removing outliers, performing interpolation to complete data, aligning different measurement signals in time, and performing normalization processing to ensure that the data of different physical quantities have the same numerical scale. After data processing, a standardized battery data set is obtained. Based on the standardized data set, different functional layers are divided according to the internal structure of the battery, and corresponding ion transport equations are established to characterize the electrochemical behavior inside the battery. The basic structure of a lithium-ion battery includes five main layers: the negative electrode current collector, the negative electrode, the separator, the positive electrode, and the positive electrode current collector. Each layer plays a different role during the operation of the battery. In the negative and positive electrodes, the insertion and extraction of lithium ions are the main electrochemical reactions, while in the separator, lithium ions are mainly transported by diffusion. Therefore, ion transport equations applicable to different layers are established respectively. Assuming that the distribution of lithium ion concentration in the x direction is , its diffusion behavior is described by Fick's first law:

[0038]

[0039] where represents the ion flux, is the diffusion coefficient, represents the concentration gradient. According to the law of conservation of mass, Fick's second law is obtained:

[0040]

[0041] This equation describes the diffusion process of lithium ions in the electrode and the separator, and needs to be solved in combination with boundary conditions and initial conditions to obtain the electro-chemical reaction characteristic matrix of the battery. At the same time, at the electrode interface, the charge transfer process is described by the Butler-Volmer equation:

[0042]

[0043] where is the current density, is the exchange current density, is the electron transfer coefficient, is the Faraday constant, is the overpotential, is the gas constant, is the temperature. By solving this equation, the current distribution on the electrode surface is obtained, and the internal potential distribution of the battery is calculated to form a complete electro-chemical reaction characteristic matrix. After constructing the electro-chemical reaction characteristic matrix, a three-dimensional unsteady heat conduction equation of the battery is established to describe the temperature distribution during the charge and discharge process. The heat sources of the battery mainly include ohmic heat, polarization heat and side reaction heat, and these heats will conduct and diffuse inside the battery and affect its thermal stability. Assuming the temperature distribution inside the battery is , its evolution is described by the three-dimensional unsteady heat conduction equation:

[0044]

[0045] where, is the material density, is the specific heat capacity, is the thermal conductivity, is the heat source term per unit volume. In order to correlate the electro-chemical reaction and heat conduction, the relationship between the heat source term and the current density and potential distribution is established. The heat source term is expressed as:

[0046]

[0047] where, is the current, is the ohmic internal resistance of the battery, is the interfacial current density, is the overpotential of the electrode. By establishing this heat source term mapping function, the heat conduction characteristic matrix of the battery is constructed, so as to simulate the temperature distribution of the battery under different working conditions. Based on the heat conduction characteristic matrix, the volume change of the electrode material during the charge and discharge process is calculated, and a stress distribution function is established to describe the mechanical deformation that occurs during the cyclic use of the battery. Since the insertion and extraction of lithium ions will cause the volume expansion and contraction of the electrode material, the stress distribution is solved by the thermo-mechanical coupling equation. Assuming the strain field of the electrode is , and the stress field is , their relationship is described by the linear elastic constitutive equation:

[0048]

[0049] where, is the Young's modulus, is expressed as:

[0050]

[0051] where, is the thermal expansion coefficient, is the volume expansion coefficient caused by the change in lithium ion concentration, is the temperature change, is the change in lithium-ion concentration. By solving this stress equation, the mechanical stress characteristic matrix of the battery is obtained, and the risk of structural damage to the battery is predicted. The electro-chemical reaction characteristic matrix, the heat conduction characteristic matrix, and the mechanical stress characteristic matrix are input into a multi-physics field simulation solver for coupled calculation, and a complete multi-dimensional coupling theory matrix of the battery is obtained. During the simulation solution process, the finite volume method or the finite element method is used to discretize the calculation domain, and numerical solution methods are used for iterative calculation to obtain the steady-state or transient multi-dimensional field distribution of the battery. Through this method, the internal state of the battery under different charge and discharge rates is accurately simulated.

[0052] In one example, temperature sensors and strain gauges are set on the surface of a golf cart battery and real-time data synchronization acquisition is performed to obtain multi-dimensional measured data of the battery, and an electro-chemical impedance spectroscopy test of the battery is performed to obtain a battery impedance characteristic vector, including:

[0053] Equidistant grid division is performed on the surface of the golf cart battery, and M temperature sensors are evenly arranged at the grid points after division. At the same time, a force analysis is performed on the stress concentration area of the golf cart battery housing, and N strain gauges are arranged at the target structural positions;

[0054] The output signals of the temperature sensors and the strain gauges are conditioned to obtain a standardized analog signal, and the standardized analog signal is digitally converted to obtain digital measurement data;

[0055] Based on the digital measurement data, synchronous control of the sampling timing of the temperature sensors and the strain gauges is performed to obtain multi-dimensional measured data of the battery;

[0056] According to the multi-dimensional coupling theory matrix of the battery, state-of-charge test points are selected and an electro-chemical impedance spectroscopy test of the battery is performed to obtain a battery impedance characteristic vector.

[0057] In this example, equidistant grid division is performed on the battery surface to ensure uniform distribution of measurement points and obtain temperature and stress data with high spatial resolution. Based on the battery geometry and thermodynamic characteristics, the grid division technology in the finite element method is used to divide the battery surface into multiple regular or irregular units, and the center point of each unit is the installation position of the sensor. Assume the length of the battery surface is , and the width is , then it is divided into grids, and the size of each grid is and , so as to ensure uniform distribution of the temperature sensors at the grid centers, and a total of A temperature sensor. At the same time, in order to accurately measure the force on the battery case during operation, a stress analysis is performed on the stress concentration area of the battery case to determine the optimal placement location of the strain gauges. The identification of the stress concentration area uses the finite element analysis method. By establishing a mechanical model of the battery case and applying boundary conditions under actual operating conditions, such as the fixed support of the battery, external vibration, and internal thermal expansion, etc., to calculate the stress distribution of the battery case. Assume that the material of the battery case has an elastic modulus , Poisson's ratio , and is subjected to an external load in a certain direction, then its stress distribution is solved by the equations of elasticity:

[0058]

[0059] Among them, is the stress tensor, is the material stiffness tensor, is the strain tensor. By numerically solving this equation, the stress distribution of different parts of the battery case is obtained, and strain gauges are arranged in the stress concentration area to monitor the mechanical stress change of the battery. After completing the arrangement of the temperature sensor and the strain gauges, their output signals are conditioned to remove noise and converted into standardized analog signals. The original signals of the temperature sensor and the strain gauges are usually analog signals, containing high-frequency noise and drift, and are processed by filtering and amplification. Assume that the output signal of the temperature sensor is , and its relationship with the temperature is expressed as:

[0060]

[0061] Among them, is the sensitivity of the sensor, is the bias voltage. In order to remove noise, a low-pass filter is used, such as a first-order RC filter, and its transfer function is expressed as:

[0062]

[0063] Among them, and are the resistance and capacitance parameters of the filter respectively. After filtering, the signal amplitude is adjusted through an amplifier circuit to ensure that it is within an appropriate range. After completing the signal conditioning, the standardized analog signal is digitally converted to obtain digital measurement data. The digital conversion uses analog-to-digital conversion technology, and its core formula is:

[0064]

[0065] Among them, is the digitized data, is the input voltage, is the reference voltage, is the resolution of the ADC (e.g., 12-bit or 16-bit). After the ADC conversion is completed, digitized measurement data of temperature and stress are obtained. Based on the digitized measurement data, synchronous control of the sampling timing of the sensor is performed. Since the change rates of temperature and stress are different, a synchronous sampling strategy is adopted, that is, a unified sampling frequency is set, and the data of all sensors are read simultaneously within each sampling period. Suppose the temperature data at a certain moment is , and the stress data is , then after synchronous sampling, a battery multi-dimensional measured data matrix is constructed:

[0066]

[0067] This matrix is used to describe the temperature and stress states of the battery at different moments. According to the battery multi-dimensional coupling theory matrix, state-of-charge test points are selected, and electrochemical impedance spectroscopy (EIS) tests of the battery are performed to obtain the impedance characteristic vector of the battery. The EIS test applies a small-amplitude alternating current signal to the battery at different frequencies and measures its impedance response, and an equivalent circuit model is used for analysis. Suppose the equivalent circuit model of the battery consists of an ohmic resistance , a charge transfer resistance , a double-layer capacitance and a Warburg impedance , then its impedance expression is:

[0068]

[0069] Among them, is the angular frequency, is the imaginary unit, is mainly used to describe the ion diffusion behavior inside the battery. By measuring the impedance data at different frequencies and using the non-linear least squares method for parameter fitting, the impedance characteristic vector of the battery is extracted for evaluating the electrochemical performance of the battery.

[0070] In one example, according to the battery multi-dimensional coupling theory matrix, state-of-charge test points are selected and electrochemical impedance spectroscopy tests of the battery are performed to obtain the impedance characteristic vector of the battery, including:

[0071] According to the electrochemical reaction characteristic data in the battery multi-dimensional coupling theory matrix, the test conditions are optimized, state-of-charge test points are selected, and an alternating current excitation signal is set for each state-of-charge test point to obtain a set of test condition parameters;

[0072] Based on the test condition parameter set, an alternating current excitation signal is applied to the golf cart battery for electrochemical impedance spectroscopy scanning to obtain impedance response data;

[0073] According to the impedance response data, an equivalent circuit model including series resistance, charge transfer resistance, double-layer capacitance, and Warburg impedance is constructed;

[0074] The impedance response data is input into the nonlinear least squares fitting algorithm for parameter identification calculation of the equivalent circuit model to obtain the equivalent circuit parameter set;

[0075] Perform frequency domain characteristic analysis on the equivalent circuit parameter set to extract the battery impedance characteristic vector including ohmic impedance in the high frequency region, charge transfer impedance in the middle frequency region, and concentration polarization impedance in the low frequency region.

[0076] In this example, the test conditions are optimized based on the electrochemical reaction characteristic data in the battery multi-dimensional coupling theory matrix to ensure that the test accurately reflects the internal state of the battery. The electrochemical behavior of the battery is affected by factors such as state of charge (SOC), temperature, charge and discharge rate, etc. When optimizing the test conditions, these factors are comprehensively considered, and appropriate state of charge test points are selected by analyzing the equivalent circuit characteristics of the battery. Assume the state of charge range of the battery is , in order to comprehensively cover the operating characteristics of the battery, one or more test points are selected in the low, medium, and high SOC regions for testing. The selection of test points is based on the relationship between the open circuit voltage of the battery and the state of charge. By measuring the open circuit voltage of the battery and combining with the known SOC curve, the test points are accurately selected. After determining the test points, an alternating current excitation signal is set for each state of charge test point to stimulate the electrochemical reaction of the battery in different frequency ranges and obtain the test condition parameter set. The alternating current excitation signal is a small-amplitude sine wave current signal, and its mathematical expression is:

[0077]

[0078] where, is the applied alternating current, is the amplitude of the excitation current, is the angular frequency, is the frequency of the excitation signal. The EIS test covers multiple frequency points from millihertz (mHz) to kilohertz (kHz) to fully characterize the impedance characteristics of the battery. Multiple frequency points are set for scanning to form a complete test condition parameter set:

[0079]

[0080] Based on the test condition parameter set, an alternating current excitation signal is applied to the golf cart battery, and electrochemical impedance spectroscopy (EIS) scanning is performed to obtain the impedance response data of the battery. The principle of EIS testing is to apply a small-amplitude alternating current signal to the battery and measure the voltage response of the battery, and then calculate its impedance. :

[0081]

[0082] Among them, is the complex impedance, and are the alternating current voltage and alternating current of the battery at frequency respectively. The data of EIS testing is presented in a Nyquist plot (real part versus imaginary part ) or a Bode plot (frequency versus impedance modulus and phase angle) to reflect the internal dynamic characteristics of the battery. After obtaining the impedance response data, an equivalent circuit model is constructed to analyze the electrochemical characteristics of the battery. The equivalent circuit model of a lithium-ion battery consists of multiple resistors, capacitors, and impedance elements, which are used to describe the ohmic loss, charge transfer process, double-layer effect, and ion diffusion behavior inside the battery. A typical equivalent circuit model includes: series resistance , which represents the bulk resistance of the battery and the resistance of the current collector and electrolyte; charge transfer resistance , which is used to characterize the electrochemical reaction impedance at the electrode-electrolyte interface; double-layer capacitance , which describes the charge accumulation effect on the electrode surface; Warburg impedance , which describes the ion diffusion behavior inside the battery. The total impedance expression of the equivalent circuit is written as:

[0083]

[0084] Among them, is the imaginary unit, is the angular frequency. The Warburg impedance is expressed as:

[0085]

[0086] Among them, is the Warburg coefficient, which characterizes the ion diffusion rate. In order to extract the equivalent circuit parameters from the experimental data, the impedance response data is input into a non-linear least squares fitting algorithm to optimize the values of the circuit parameters. The goal of the non-linear least squares method is to minimize the error between the experimentally measured impedance value and the impedance value calculated by the equivalent circuit model:

[0087]

[0088] In this optimization process, the Levenberg-Marquardt algorithm (LM algorithm) is used for iterative solution to quickly converge to the optimal parameter values. After optimization, a complete set of equivalent circuit parameters is obtained:

[0089]

[0090] After obtaining the equivalent circuit parameters, frequency-domain characteristic analysis is carried out to extract the key impedance characteristic vectors of the battery. The impedance characteristics of the battery are divided into a high-frequency region, a mid-frequency region, and a low-frequency region, where: The impedance in the high-frequency region is mainly dominated by the ohmic impedance and reflects the internal resistance of the battery and the current collection efficiency:

[0091]

[0092] The impedance in the mid-frequency region is mainly determined by the charge transfer impedance and reflects the charge transfer kinetics characteristics at the electrode-electrolyte interface:

[0093]

[0094] The impedance in the low-frequency region is mainly dominated by the concentration polarization impedance and describes the ion diffusion and polarization behavior inside the battery:

[0095]

[0096] By extracting the high-frequency ohmic impedance , the mid-frequency charge transfer impedance , and the low-frequency concentration polarization impedance , a complete battery impedance characteristic vector is obtained:

[0097]

[0098] In one example, according to the multi-dimensional measured data of the battery and the multi-dimensional coupling theory matrix of the battery, the battery volume is discretized into control volume units and thermal-mechanical coupling iterative solution is carried out to obtain the thermal-mechanical coupling characteristic matrix of the battery, including:

[0099] According to the multi-dimensional measured data of the battery, three-dimensional grid division is carried out on the battery volume, the battery structure is discretized into control volume units, and a unique identification number is assigned to each control volume unit to obtain the battery grid model;

[0100] Based on the multi-dimensional coupling theory matrix of the battery, an initial field distribution of heat conduction is established, and the temperature sensor measurement data is mapped to the battery grid model through a three-dimensional spline interpolation algorithm to obtain the initial value of the temperature field;

[0101] Perform boundary condition processing on the initial temperature field values and analyze the thermal-mechanical coupling relationship to obtain the thermal-mechanical system equations;

[0102] Input the thermal-mechanical system equations into the finite volume method solver, perform alternating iterative calculations on the temperature field and stress field, and set the convergence criterion as the relative error of the temperature field between adjacent iterative steps being less than the target value to obtain the thermal-mechanical coupling distribution;

[0103] Based on the thermal-mechanical coupling distribution, conduct temperature gradient stress analysis, calculate the thermal stress and strain of each control volume unit, establish a heat source correction model under the influence of stress, and obtain the corrected thermal distribution;

[0104] Combine and map the temperature field data and stress field data of each control volume unit in the corrected thermal distribution to obtain the battery thermal-mechanical coupling characteristic matrix.

[0105] In this example, based on the multi-dimensional measured data of the battery, perform three-dimensional grid division on the battery volume to discretize the continuous physical domain and make it suitable for numerical calculations. The non-uniform grid division method is adopted, that is, finer grids are used in the temperature gradient and stress concentration regions, while coarser grids are used in the regions where the temperature and stress change more gently. Assume the three-dimensional geometric dimensions of the battery are , and divide it into control volume units, and the volume of each unit is:

[0106]

[0107] where are the number of grids along the directions respectively. Each control volume unit is assigned a unique identification number to ensure the traceability of data storage and calculation and form a complete battery grid model. Based on the battery multi-dimensional coupling theory matrix, establish the initial field distribution of heat conduction. Since the temperature data is measured by a finite number of sensors, map the temperature measurement data to the entire grid model through a three-dimensional spline interpolation algorithm to obtain the initial temperature field values. Assume the temperature values measured at sensor positions are , and the corresponding coordinates are , then the three-dimensional spline interpolation function is expressed as:

[0108]

[0109] where is the three-dimensional spline basis function, is the interpolation coefficient. By solving the interpolation equations, the interpolation coefficient is determined, and the initial distribution of the temperature field inside the entire battery is obtained. Boundary condition processing and thermal-mechanical coupling relationship analysis are performed on the initial values of the temperature field to establish a complete thermo-mechanical system equation. The heat conduction behavior of the battery is described by the three-dimensional unsteady heat conduction equation:

[0110]

[0111] where, is the density of the battery material, is the specific heat capacity, are the thermal conductivities in different directions, is the heat source term. The heat source term is mainly composed of ohmic heating, polarization heat, and side reaction heat, and is expressed as:

[0112]

[0113] where, is the current, is the ohmic internal resistance, is the current density, is the overpotential of the electrode. At the same time, temperature changes will cause material expansion, resulting in internal stress, and this process is described by the thermal stress equation:

[0114]

[0115] where, is the stress tensor, is the material stiffness tensor, is the strain tensor, and the strain tensor includes a thermal expansion term:

[0116]

[0117] where, is the thermal expansion coefficient, is the expansion coefficient caused by the change in lithium-ion concentration, is the temperature change, is the change in lithium-ion concentration. By solving the coupled system of the heat conduction equation and the stress balance equation, a complete thermo-mechanical system equation is obtained. The thermo-mechanical system equation is input into the finite volume method solver to perform alternating iterative calculations of the temperature field and the stress field and ensure the convergence of the calculations. The finite volume method discretizes the partial differential equation into an algebraic equation by performing an energy conservation analysis on the control volume element. For any control volume element, the discretized form of the heat conduction equation is expressed as:

[0118]

[0119] where, and represent the temperature at the current and next time steps respectively, is the time step, is the area of the grid surface, is the thermal conductivity at the grid surface, is the internal heat source term of the element. During iterative calculation, the convergence criterion is set as the relative error of the temperature field between adjacent iterative steps being less than the target value, that is:

[0120]

[0121] where, is the convergence threshold. When all control volume elements satisfy this condition, the iteration terminates and the final thermo-mechanical coupling distribution is obtained. Based on the thermo-mechanical coupling distribution, temperature gradient stress analysis is carried out to calculate the thermal stress and strain of each control volume element, and a heat source correction model under the influence of stress is established. Based on the stress field, the heat conduction equation is corrected. Let be the effective thermal conductivity after stress correction, and its calculation formula is:

[0122]

[0123] where, is the initial thermal conductivity, is the material property parameter, is the local stress. By iteratively calculating the corrected heat conduction equation, the final corrected thermo-mechanical distribution is obtained. It is necessary to combine and map the temperature field data and stress field data in the corrected thermo-mechanical distribution to generate a complete battery thermo-mechanical coupling characteristic matrix.

[0124] In an example, the thermo-mechanical system equation is input into the finite volume method solver, and the temperature field and stress field are alternately iteratively calculated, and the convergence criterion is set as the relative error of the temperature field between adjacent iterative steps being less than the target value, and the thermo-mechanical coupling distribution is obtained, including:

[0125] Perform control volume discretization on the thermo-mechanical system equation, establish a discrete algebraic equation system including the temperature field thermal conductivity matrix and the stress field stiffness matrix, and obtain the control volume discrete equation;

[0126] Based on the control volume discrete equation, construct the coupling matrix of the momentum equation and the energy equation, and process the boundary conditions of the temperature field and the stress field to obtain the coupling solution matrix;

[0127] Input the coupling solution matrix into the SIMPLE algorithm iterative solver to calculate the thermal stress increment caused by the temperature field, and obtain the temperature stress correction amount;

[0128] According to the temperature stress correction amount, perform correction calculation on the temperature field distribution of the control volume element, establish the velocity-pressure coupling relationship between the stress field and the temperature field, and obtain the correction coefficient matrix;

[0129] Calculate the relative error of the temperature field between adjacent iterative steps based on the correction coefficient matrix, and compare the relative error with the set target value to obtain the iteration termination criterion;

[0130] Post-process the temperature field and stress field data that meet the iteration termination criterion, extract the temperature distribution and stress distribution data of each control volume unit, and obtain the thermo-mechanical coupling distribution.

[0131] In this example, the control volume discretization is performed on the thermo-mechanical system equations to accurately simulate the temperature and stress distributions inside the battery in numerical calculations. The thermo-mechanical system equations consist of a three-dimensional unsteady heat conduction equation and an elasticity equation. Among them, the heat conduction equation is used to describe the temperature change inside the battery, and the mechanical equation is used to describe the stress distribution caused by the temperature change. Assume the temperature field of the battery satisfies the three-dimensional heat conduction equation:

[0132]

[0133] where is the material density, is the specific heat capacity, are the thermal conductivities of the battery in the directions respectively, is the internal heat source term, mainly including ohmic heat, polarization heat and side reaction heat:

[0134]

[0135] where is the current, is the ohmic resistance of the battery, is the current density, is the overpotential of the electrode. For discretization, the computational domain is divided into multiple control volume units, and the volume of each unit is , and the temperature value is calculated at the center point of each unit. Applying the finite volume method, integrating the heat conduction equation over the control volume and using discrete approximation, the discrete equation of the temperature field is obtained:

[0136]

[0137] where and represent the temperatures at the current and next time steps respectively, is the grid face area, is the thermal conductivity on the grid face, is the heat source term inside the control volume unit. At the same time, the mechanical equation describes the stress change caused by thermal expansion. Assume the stress tensor Satisfy the linear elastic constitutive relation:

[0138]

[0139] where, is the material stiffness tensor, is the strain tensor, which is given by the following relation:

[0140]

[0141] where, is the coefficient of thermal expansion, is the coefficient of expansion caused by the change in lithium-ion concentration, is the change in temperature, is the change in lithium-ion concentration. By discretizing the mechanical equations, the discrete equations of the stress field are obtained, forming a discrete algebraic system of equations containing the thermal conductivity matrix of the temperature field and the stiffness matrix of the stress field. Based on the discrete equations of the control volume, the coupling matrices of the momentum equation and the energy equation are constructed, and the boundary conditions of the temperature field and the stress field are processed to ensure the stability and accuracy of the calculation. The boundary conditions include the convective heat transfer boundary, the fixed constraint boundary, and the free expansion boundary, etc. In the heat conduction equation, the boundary heat flux satisfies:

[0142]

[0143] where, is the convective heat transfer coefficient, is the ambient temperature, and in the mechanical equation, the boundary force satisfies:

[0144]

[0145] where, is the boundary normal vector, is the externally applied load. After processing the boundary conditions, the coupled solution matrix is obtained. The coupled solution matrix is input into the SIMPLE algorithm iterative solver to calculate the thermal stress increment caused by the temperature field, and the temperature stress correction amount is obtained. The SIMPLE algorithm adjusts the flow field through the pressure correction equation to make it converge to a physically reasonable state. The SIMPLE algorithm is used to simultaneously solve the temperature field and the stress field, so that the iterative processes of the two can converge stably. According to the temperature stress correction amount, the temperature field distribution of the control volume unit is corrected and calculated, and the velocity-pressure coupling relationship between the stress field and the temperature field is established. Since the temperature gradient will affect the expansion behavior of the material, thereby changing the local heat conduction characteristics, a correction coefficient matrix is introduced to make the solution processes of the temperature field and the stress field consistent. Assume is the corrected effective thermal conductivity, then its calculation formula is:

[0146]

[0147] Among them, is the initial thermal conductivity, is the material characteristic parameter, is the local stress. Through this correction model, the physical consistency of the thermo-mechanical coupling solution is ensured. Based on the correction coefficient matrix, the relative error of the temperature field between adjacent iteration steps is calculated and compared with the set target value to determine whether the iteration termination criterion is met. The method for calculating the relative error is as follows:

[0148]

[0149] When the relative error of all control volume elements is less than the preset threshold

[0150] the iteration terminates, and the final thermo-mechanical coupling distribution is obtained. After satisfying the iteration termination criterion, the calculated temperature field and stress field data are post-processed to extract the temperature distribution and stress distribution data of each control volume element and generate a complete thermo-mechanical coupling distribution. The post-processing includes data visualization, contour analysis, and stress concentration region identification, etc. For example, in practical applications, if the temperature gradient in a certain area exceeds the design threshold or the stress concentration coefficient is high, the battery structure is adjusted or the cooling strategy is optimized to reduce the degradation risk during long-term use.

[0151] Input the battery impedance characteristic vector, the battery thermo-mechanical coupling characteristic matrix, and the battery multi-dimensional measured data into a dual encoder network for feature weighting to obtain a working condition feature representation;

[0152] Construct an InfoNCE contrast loss function based on the working condition feature representation, and iteratively optimize the dual encoder network based on the batch normalization layer and the Adam optimizer to obtain working condition-independent features;

[0153] Input the working condition-independent features into the discriminator network, minimize the working condition prediction accuracy through adversarial training, and construct a weighted combination of the working condition classification cross-entropy loss function and the feature reconstruction loss function for gradient update to obtain a performance feature vector;

[0154] Calculate the KLD loss function based on the performance feature vector, constrain the feature distribution in the normal distribution space, and obtain a normalized battery feature set by minimizing the relative entropy between the feature distribution and the target distribution;

[0155] Calculate the performance indicators for the normalized battery feature set to obtain the capacity attenuation rate, internal resistance growth rate, temperature consistency coefficient, and stress concentration coefficient;

[0156] The capacity attenuation rate, internal resistance growth rate, temperature consistency coefficient, and stress concentration coefficient are input into the degradation mechanism analysis model for time series feature mapping and performance degradation warning to obtain the battery performance evaluation result.

[0157] In this example, the impedance characteristic vector of the battery, the thermal-mechanical coupling characteristic matrix of the battery, and the multi-dimensional measured data of the battery are input into the dual encoder network for feature weighting and obtaining the operating condition feature representation. Let the impedance characteristic vector of the battery be , which includes the high-frequency ohmic impedance , the intermediate-frequency charge transfer impedance , and the low-frequency concentration polarization impedance , that is, ; Let the thermal-mechanical coupling characteristic matrix of the battery be , used to characterize the temperature and stress distributions, that is, , where is the temperature distribution, is the stress distribution; Let the multi-dimensional measured data of the battery be , which includes information such as current, voltage, and ambient temperature, that is, . The dual encoder network consists of two independent encoders, which are used to process the impedance and thermal-mechanical coupling characteristics, and the multi-dimensional measured data respectively. Assume the parameters of the encoders are and , then the encoded feature representation is:

[0158]

[0159] The encoded features are weighted and fused to obtain the final operating condition feature representation :

[0160]

[0161] Among them, and are trainable feature weighting coefficients to ensure reasonable distribution of the contributions from different data sources. According to the operating condition feature representation, an InfoNCE contrastive loss function is constructed to optimize the dual encoder network and obtain the operating condition-independent features. The goal of the hfoNCE loss is to maximize the similarity of the positive samples while minimizing the similarity of the negative samples, ensuring that the features learned by the model can remain stable under different operating conditions. Assume the positive sample is , and the negative sample is , then the InfoNCE loss function is defined as:

[0162]

[0163] Among them, represents the cosine similarity, is the total number of negative samples. To improve training stability, batch normalization layers are used to normalize the features during training, and the Adam optimizer is adopted for gradient optimization, enabling the model to converge rapidly. The condition-independent features are input into the discriminator network to minimize the condition prediction accuracy through adversarial training and extract the core performance feature vectors of the battery. The discriminator predicts the condition category of the battery based on the input features, while the goal of the dual encoder network is to generate condition-independent features that are difficult to be correctly classified. This adversarial training mechanism makes the model focus more on the physical characteristics of the battery itself rather than the condition-related information. The loss function of the discriminator consists of the condition classification cross-entropy loss and the feature reconstruction loss and is expressed mathematically as:

[0164]

[0165]

[0166] where is the number of condition categories, is the true condition label, is the predicted probability, is the reconstructed feature representation. The final adversarial loss function is a weighted combination of the two:

[0167]

[0168] where and are the weight coefficients. Through adversarial training, the performance feature vector is obtained as:

[0169]

[0170] where is the mapping function of the discriminator network. Based on the performance feature vector, the KLD loss is calculated to constrain the feature distribution within the normal distribution space to ensure the consistency of the features under different batteries and conditions. The KLD loss function is used to measure the relative entropy between two probability distributions and and its expression is:

[0171]

[0172] where is the distribution of the current feature, is the target normal distribution. By minimizing the KLD loss, it is ensured that the final battery features meet the standardization requirements, and a normalized battery feature set is obtained. Performance indicators are calculated for the normalized battery feature set to extract key health state parameters. Among them, the capacity attenuation rate By comparing the current capacity with the initial capacity It is calculated that:

[0173]

[0174] Internal resistance growth rate By measuring the current internal resistance and the initial internal resistance The change is calculated as:

[0175]

[0176] Temperature consistency coefficient Characterized by calculating the standard deviation of the internal temperature distribution of the battery:

[0177]

[0178] Wherein, is the average temperature of the battery. Stress concentration coefficient Obtained by calculating the maximum gradient of the stress field:

[0179]

[0180] These performance indicators are input into the degradation mechanism analysis model for time series feature mapping and performance degradation warning. The degradation mechanism analysis model uses a long short-term memory network (LSTM) or a temporal convolutional network (TCN) to learn the temporal change trend of battery performance and predict the future health state. Assuming that the performance of the battery changes over time Satisfies the following recurrence relation:

[0181]

[0182] Wherein, is the amount of performance change at the current moment. By training this model, the future trend of battery performance degradation is predicted, and a warning is issued before the battery reaches a certain safety threshold.

[0183] Referring to Figure 2 , this embodiment provides a multi-dimensional test device for a golf cart battery, including:

[0184] Data acquisition unit 1, used to collect the standard operating condition data of the golf cart battery, construct an electrochemical-thermal-mechanical coupling theory model, and obtain a battery multi-dimensional coupling theory matrix;

[0185] Test unit 2, used to set temperature sensors and strain gauges on the surface of the golf cart battery and perform real-time data synchronous acquisition to obtain battery multi-dimensional measured data, and perform an electrochemical impedance spectroscopy test on the battery to obtain a battery impedance characteristic vector;

[0186] An iterative solution unit 3, configured to discretize the battery volume into control volume units according to the multi-dimensional measured data of the battery and the multi-dimensional coupling theory matrix of the battery, and perform thermal-mechanical coupling iterative solution to obtain a battery thermal-mechanical coupling characteristic matrix;

[0187] A performance evaluation unit 4, configured to perform battery performance evaluation based on the battery impedance characteristic vector, the battery thermal-mechanical coupling characteristic matrix, and the multi-dimensional measured data of the battery to obtain a battery performance evaluation result.

[0188] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the description in the above method embodiment, and details are not described herein again.

[0189] It should be noted that in this article, the terms "including", "comprising" or any other variation thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, device, article or method including that element.

[0190] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A multi-dimensional testing method for golf cart batteries, characterized in that: The following steps are involved: Collect data of golf cart battery under standard working conditions, build electrochemical-thermal-mechanical coupling theoretical model, and obtain multi-dimensional coupling theoretical matrix of battery; A temperature sensor and a strain gauge are arranged on the surface of the golf cart battery to perform real-time data synchronous acquisition to obtain multi-dimensional measured data of the battery, and an electrochemical impedance spectrum test of the battery is performed to obtain a battery impedance characteristic vector; According to the multi-dimensional measured data of the battery and the multi-dimensional coupling theoretical matrix of the battery, the battery volume is discretized into control volume units and thermal-mechanical coupling iterative solution is performed to obtain the battery thermal-mechanical coupling characteristic matrix; Based on the battery impedance characteristic vector, the battery thermal-mechanical coupling characteristic matrix and the battery multi-dimensional measured data, a battery performance evaluation is performed to obtain a battery performance evaluation result; specifically comprising: inputting the battery impedance characteristic vector, the battery thermal-mechanical coupling characteristic matrix and the battery multi-dimensional measured data into a dual encoder network for feature weighting to obtain a working condition feature representation; constructing an InfoNCE contrast loss function according to the working condition feature representation, iteratively optimizing the dual encoder network based on a batch normalization layer and an Adam optimizer to obtain working condition-independent features; inputting the working condition-independent features into a discriminator network, minimizing the working condition prediction accuracy through adversarial training, and constructing A weighted combination of the working condition classification cross entropy loss function and the feature reconstruction loss function is gradient updated to obtain a performance feature vector; the KLD loss function is calculated based on the performance feature vector, the feature distribution is constrained within the normal distribution space, and the normalized battery feature set is obtained by minimizing the relative entropy between the feature distribution and the target distribution; performance indicators of the normalized battery feature set are calculated to obtain capacity decay rate, internal resistance growth rate, temperature consistency coefficient and stress concentration coefficient; the capacity decay rate, internal resistance growth rate, temperature consistency coefficient and stress concentration coefficient are input into the degradation mechanism analysis model for time series feature mapping and performance degradation warning to obtain the battery performance evaluation result.

2. The multi-dimensional testing method for golf cart batteries according to claim 1, characterized in that: The standard operating data of the golf cart battery is collected, and an electrochemical-thermal-mechanical coupling theoretical model is constructed to obtain a battery multi-dimensional coupling theoretical matrix, including: Collecting voltage, current and temperature distribution data of a golf cart battery under 0.5C, 1C and 2C rate charge and discharge conditions to obtain a battery operation raw data set, and performing data preprocessing on the battery operation raw data set to obtain a battery standardized data set; Based on the battery standardized data set, the battery structure is divided into five layers: negative electrode current collector, negative electrode, separator, positive electrode, and positive electrode current collector, and the ion transport equation of each layer is established to obtain the electrochemical reaction characteristic matrix; A three-dimensional non-steady-state heat conduction equation is established according to the electrochemical reaction characteristic matrix, the golf cart battery is meshed, a mapping function between a heat source term and an electrochemical reaction is established, and a heat conduction characteristic matrix is ​​obtained; Based on the thermal conductivity characteristic matrix, the volume change of the electrode material during the charge and discharge process is calculated, and a stress distribution function is established to obtain a mechanical stress characteristic matrix; The electrochemical reaction characteristic matrix, the heat conduction characteristic matrix and the mechanical stress characteristic matrix are input into a multi-physics field simulation solver for coupling calculation to obtain a battery multi-dimensional coupling theoretical matrix.

3. The multi-dimensional testing method of golf cart batteries according to claim 1, characterized in that: The temperature sensor and strain gauge are arranged on the surface of the golf cart battery and real-time data synchronous acquisition is performed to obtain multi-dimensional measured data of the battery, and the battery electrochemical impedance spectrum test is performed to obtain the battery impedance characteristic vector, including: Divide the surface of the golf cart battery into equally spaced grids, evenly arrange M temperature sensors on the divided grid points, and at the same time, perform a force analysis on the stress concentration area of ​​the golf cart battery housing, and arrange N strain gauges at the target structural position; Conditioning the output signals of the temperature sensor and the strain gauge to obtain standardized analog signals, and digitally converting the standardized analog signals to obtain digital measurement data; Based on the digital measurement data, synchronously control the sampling timing of the temperature sensor and the strain gauge to obtain multi-dimensional measured data of the battery; According to the battery multi-dimensional coupling theory matrix, a state of charge test point is selected and a battery electrochemical impedance spectrum test is performed to obtain a battery impedance characteristic vector.

4. The multi-dimensional testing method for golf cart batteries according to claim 3, characterized in that: The step of selecting a state of charge test point according to the battery multi-dimensional coupling theory matrix and performing a battery electrochemical impedance spectrum test to obtain a battery impedance characteristic vector includes: Optimizing the test conditions according to the electrochemical reaction characteristic data in the multi-dimensional coupling theory matrix of the battery, selecting the state of charge test points, and setting an AC excitation signal for each state of charge test point to obtain a test condition parameter set; Based on the test condition parameter set, applying an AC excitation signal to the golf cart battery to perform an electrochemical impedance spectrum scan to obtain impedance response data; According to the impedance response data, an equivalent circuit model including series resistance, charge transfer resistance, double layer capacitance and Warburg impedance is constructed; Inputting the impedance response data into a nonlinear least squares fitting algorithm, performing parameter identification calculation on the equivalent circuit model, and obtaining an equivalent circuit parameter set; The frequency domain characteristic analysis is performed on the equivalent circuit parameter set to extract the battery impedance characteristic vector including the ohmic impedance in the high frequency region, the charge transfer impedance in the medium frequency region and the concentration polarization impedance in the low frequency region.

5. The multi-dimensional testing method for golf cart batteries according to claim 1, characterized in that: According to the multi-dimensional measured data of the battery and the multi-dimensional coupling theoretical matrix of the battery, the battery volume is discretized into control volume units and thermal-mechanical coupling iterative solution is performed to obtain the battery thermal-mechanical coupling characteristic matrix, including: Dividing the battery volume into three-dimensional grids according to the multi-dimensional measured data of the battery, discretizing the battery structure into control volume units, and assigning a unique identification number to each control volume unit to obtain a battery grid model; Establishing the initial field distribution of heat conduction based on the multi-dimensional coupling theory matrix of the battery, and mapping the temperature sensor measurement data to the battery grid model through a three-dimensional spline interpolation algorithm to obtain the initial value of the temperature field; Performing boundary condition processing and thermal-mechanical coupling relationship analysis on the initial value of the temperature field to obtain a thermal system equation; The thermodynamic system equation is input into the finite volume method solver, the temperature field and the stress field are alternately iterated and calculated, and the convergence criterion is set as the relative error of the temperature field in adjacent iteration steps is less than the target value, so as to obtain the thermomechanical coupling distribution; Based on the thermal-mechanical coupling distribution, a temperature gradient stress analysis is performed to calculate the thermal stress and strain of each control volume unit, and a heat source correction model under the influence of stress is established to obtain a corrected thermal-mechanical distribution; The temperature field data and stress field data of each control volume unit in the modified thermal distribution are combined and mapped to obtain a battery thermal coupling characteristic matrix.

6. The multi-dimensional testing method for golf cart batteries according to claim 5, characterized in that: The thermodynamic system equation is input into the finite volume method solver, the temperature field and the stress field are alternately iterated and calculated, and the convergence criterion is set as the relative error of the temperature field in adjacent iteration steps is less than the target value, and the thermomechanical coupling distribution is obtained, including: The control volume discretization process is performed on the thermodynamic system equations to establish a discrete algebraic equation group including a temperature field thermal conductivity matrix and a stress field stiffness matrix to obtain a control volume discretization equation; Based on the control volume discrete equation, a coupling matrix of the momentum equation and the energy equation is constructed, and the boundary conditions of the temperature field and the stress field are processed to obtain a coupling solution matrix; Inputting the coupling solution matrix into the SIMPLE algorithm iterative solver, calculating the thermal stress increment caused by the temperature field, and obtaining the temperature stress correction amount; According to the temperature stress correction amount, the temperature field distribution of the control volume unit is corrected and calculated, the velocity-pressure coupling relationship between the stress field and the temperature field is established, and a correction coefficient matrix is ​​obtained; Calculate the relative error of the temperature field in adjacent iteration steps based on the correction coefficient matrix, and compare the relative error with the set target value to obtain an iteration termination criterion; The temperature field and stress field data satisfying the iteration termination criterion are post-processed, and the temperature distribution and stress distribution data of each control volume unit are extracted to obtain the thermal-mechanical coupling distribution.

7. A multi-dimensional testing device for golf cart batteries, characterized in that: The steps for implementing the multi-dimensional testing method of a golf cart battery according to any one of claims 1 to 6, the multi-dimensional testing device of a golf cart battery comprises: A data acquisition unit is used to collect data of golf cart batteries under standard operating conditions, build an electrochemical-thermal-mechanical coupling theoretical model, and obtain a multi-dimensional coupling theoretical matrix of the battery; A testing unit, used to set a temperature sensor and a strain gauge on the surface of the golf cart battery and perform real-time data synchronous acquisition to obtain multi-dimensional measured data of the battery, and perform a battery electrochemical impedance spectrum test to obtain a battery impedance characteristic vector; An iterative solution unit, used to discretize the battery volume into control volume units and perform thermal-mechanical coupling iterative solution according to the multi-dimensional measured data of the battery and the multi-dimensional coupling theoretical matrix of the battery, so as to obtain a battery thermal-mechanical coupling characteristic matrix; The performance evaluation unit is used to evaluate the battery performance based on the battery impedance characteristic vector, the battery thermal-mechanical coupling characteristic matrix and the multi-dimensional measured data of the battery to obtain a battery performance evaluation result.

Images