A battery state of health estimation method and system based on parameter identification

Abstract

The application discloses a battery health state evaluation method and system based on parameter identification, and aims to solve the problem that there is no timely, reliable and accurate battery fault detection method. The method constructs a battery model; based on the Levenberg-Marquardt (LM) algorithm, the model parameters of different batteries are identified respectively by combining the actual measurement data of different batteries, and a plurality of characteristic parameters representing the health states of the batteries to be measured are obtained; for each characteristic parameter and the corresponding residual loss, the batteries to be measured are analyzed by using the quartile range method or the median absolute deviation method, and the corresponding abnormal battery in each characteristic parameter is obtained; and the health states of the batteries to be measured are evaluated according to the distribution of the characteristic parameters. The method realizes efficient evaluation of the health state of the lithium ion battery and accurate positioning of the abnormal battery, and provides theoretical support and technical optimization scheme for the battery safety management system.

Description

Technical Field

[0001] This application relates to the field of battery testing technology, and in particular to a method and system for assessing battery health status based on parameter identification. Background Technology

[0002] Common faults in lithium-ion batteries include internal and external short circuits, open circuits, and others. An internal short circuit in a lithium-ion battery refers to a sudden temperature rise due to heat buildup inside the battery, usually caused by dendrite growth or lithium deposition. This can damage the separator and lead to accidental contact between the positive and negative electrodes. Open circuit faults in lithium-ion batteries are usually caused by insufficient or incomplete connections between individual cells. When a lithium-ion battery fails, mechanical stress, corrosion, and terminal problems can exacerbate these faults, increasing circuit resistance, interfering with current, reducing the total battery capacity, and in severe cases, even causing complete system failure. Therefore, timely and effective battery fault detection is crucial for ensuring the safety and performance of electric vehicles using lithium-ion batteries. However, the complexity, nonlinearity, and dynamic nature of lithium-ion batteries pose significant challenges to accurate and reliable battery fault detection technologies.

[0003] Currently, battery fault detection technologies can be mainly divided into three categories: threshold-based methods, model-based methods, and data-driven methods.

[0004] Threshold-based methods are widely used in fault detection for battery management systems. This method continuously monitors key parameters such as system voltage, temperature, and current; when these parameters exceed predefined thresholds, a battery fault is identified. However, threshold setting has certain issues. If the threshold is set too low, it may lead to frequent false alarms, while if the threshold is set too high, it may result in missed fault detections.

[0005] Model-based methods primarily rely on mathematical or electrochemical models to estimate internal state parameters of the battery (such as voltage, current, and temperature) to determine whether a system fault has occurred. The effectiveness of these methods is highly dependent on model accuracy, particularly regarding key parameters such as lithium-ion diffusion coefficient, internal resistance, and polarization voltage. However, in practical applications, long-term operation of battery systems can lead to parameter drift due to aging, cycling, and environmental temperature fluctuations, causing model parameters to gradually deviate from the actual state. Since models are difficult to update dynamically in real time, this mismatch significantly reduces the reliability of fault detection, resulting in false alarms or missed alarms. Especially when identifying early or minor faults, the changes in external parameters such as voltage and temperature are not significant, and traditional threshold-based methods often lack sufficient sensitivity to promptly capture abnormal signals, further limiting their applicability in real-time fault detection.

[0006] Data-driven approaches, which do not rely on precise physical models but instead build fault diagnosis models by analyzing and training on large amounts of historical operational data, are among the most widely used fault detection technologies in battery management systems. However, these methods typically require significant computational resources and training time, making it difficult to meet the efficiency requirements of real-time fault diagnosis. Furthermore, data-driven methods are essentially "black box" models, lacking interpretability in their decision-making processes and making it difficult to establish causal relationships between fault characteristics and the battery's internal physicochemical mechanisms. High-precision models, especially deep neural networks, while offering excellent recognition performance, have high computational complexity and stringent hardware requirements, making them typically difficult to deploy directly in resource-constrained embedded BMS systems. Summary of the Invention

[0007] This application provides a battery health status assessment method and system based on parameter identification to address the problem of the lack of a timely, reliable, and accurate battery fault detection method.

[0008] This application provides a battery health status assessment method based on parameter identification, including:

[0009] Build a battery model;

[0010] Based on the LM algorithm, model parameters are identified for each battery under test, and several characteristic parameters representing the health status of each battery under test are obtained.

[0011] For each characteristic parameter and the corresponding residual loss, the test battery is analyzed by the interquartile range method or the median absolute deviation method to obtain the corresponding abnormal data in each characteristic parameter and the test battery corresponding to the abnormal data.

[0012] The health status of each battery under test is assessed based on the distribution of its characteristic parameters.

[0013] In one example, constructing battery models corresponding to several batteries under test includes:

[0014] Several characteristic equations for constructing the battery model were determined; among them,

[0015] Based on the battery's nominal capacity, applied current, and average state of charge (SOC) value, the first characteristic equation is constructed.

[0016] Based on the open-circuit voltage, battery voltage, and overpotential term, construct the second characteristic equation for charging and discharging;

[0017] Based on the applied current, the nominal battery capacity, the diffusion time constant, and the dimensionless spatial variable, a third characteristic equation is constructed.

[0018] Based on battery current, battery voltage, and open-circuit voltage, a fourth characteristic equation based on heat is constructed.

[0019] In one example, the characteristic parameters include ohmic overpotential, dimensionless charge exchange current, and diffusion time constant;

[0020] The construction of the battery model includes:

[0021] Determine the relationship between the characteristic equations and the characteristic parameters used to construct the battery model.

[0022] In one example, the model parameter identification for each battery under test based on the LM algorithm includes:

[0023] Based on the LM algorithm, and combined with the voltage and current data of the battery under test during the test, the model parameters of each battery under test are identified.

[0024] In the parameter identification process, the parameters are iterated based on the Jacobian matrix, identity matrix and damping factor of the residual. The damping factor is adaptively adjusted until the change in residual loss is less than the first threshold, the change in parameter or gradient range norm is less than the second threshold, or the maximum number of iterations is reached, at which point the iteration stops.

[0025] In one example, when the characteristic parameter is an ohmic overpotential or a dimensionless charge exchange current, the test cell is analyzed using the interquartile range method or the median absolute deviation method for each characteristic parameter and the corresponding residual loss, to obtain the corresponding abnormal data in each characteristic parameter and the test cell corresponding to the abnormal data, including:

[0026] Based on the residual value loss corresponding to the characteristic parameters of each battery under test after parameter identification, the first quartile, the third quartile, and the interquartile range are determined by the interquartile range method.

[0027] Based on the first quartile, the third quartile, and the interquartile range, determine the upper and lower limits of the normal data range;

[0028] Data that exceeds the normal data range will be considered abnormal data.

[0029] In one example, when the characteristic parameter is a diffusion time constant, the test cell is analyzed using the interquartile range method or the median absolute deviation method for each characteristic parameter and the corresponding residual loss, to obtain the corresponding abnormal data in each characteristic parameter and the test cell corresponding to the abnormal data, including:

[0030] Based on the difference between the median diffusion time constant of each battery under test and the value of each characteristic parameter, multiple absolute deviation values ​​are determined, as well as the median absolute deviation value.

[0031] The threshold for the normal data range is determined based on the median, the absolute deviation of the median, and the set multiple.

[0032] Data that exceeds the normal data range will be considered abnormal data.

[0033] In one example, assessing the health status of each battery under test based on the distribution of characteristic parameters includes:

[0034] Based on the abnormal data in the characteristic parameters of the battery under test, determine the possible fault type of the corresponding battery under test.

[0035] In one example, assessing the health status of each battery under test based on the distribution of characteristic parameters includes:

[0036] The feature parameters are divided into independent parameters and associated parameters;

[0037] If the characteristic parameter corresponding to the abnormal data is determined to be an independent parameter, then the battery under test may be faulty.

[0038] If the feature parameter corresponding to the abnormal data is determined to be a correlation parameter, and other correlation parameters are also confirmed to have abnormal data, the battery under test may be faulty.

[0039] In one example, before performing model parameter identification on each battery under test, the method further includes:

[0040] Under urban operating conditions, the health status of each battery under test is assessed.

[0041] This application provides a battery health status assessment system based on parameter identification, comprising:

[0042] Modules for building battery models;

[0043] The identification module is used to identify the model parameters of each battery under test based on the LM algorithm, and obtain several feature parameters that characterize the health status of each battery under test.

[0044] The analysis module is used to analyze the battery under test for each characteristic parameter and the corresponding residual loss using the interquartile range method or the median absolute deviation method, and to obtain the corresponding abnormal data in each characteristic parameter and the battery under test corresponding to the abnormal data.

[0045] The evaluation module is used to assess the health status of each battery under test based on the distribution of its characteristic parameters.

[0046] This application provides a battery health status assessment method and system based on parameter identification, which can achieve the following beneficial effects:

[0047] (1) Compared with traditional model-based methods, the method proposed in this application has higher robustness. Since battery models are often affected by multiple factors, this method can more accurately judge the health status of the battery through multi-parameter comparison analysis, overcoming the limitations of traditional methods that rely solely on output voltage for judgment and are easily affected by external environmental interference.

[0048] (2) Compared with traditional deep learning-based methods, the method proposed in this application has better interpretability and thus performs better in terms of battery fault identification rate and identification accuracy.

[0049] (3) The battery model constructed in this application accurately extracts the key feature parameters for judging the health status of the battery. The selected parameters are typical in the battery operation process and are extremely important for the status assessment. The model effectively reduces the computational complexity while ensuring accuracy, and is suitable for real-time assessment of health status and fault diagnosis in battery management system. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The accompanying drawings described herein are used to provide a further understanding of this application and constitute a part of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the accompanying drawings:

[0051] Figure 1 A flowchart of a battery health status assessment method based on parameter identification provided in this application embodiment;

[0052] Figure 2 A schematic diagram illustrating the principle of the Levenberg-Marquardt algorithm provided in the embodiments of this application;

[0053] Figure 3 A flowchart of another battery health status assessment method based on parameter identification provided in this application embodiment;

[0054] Figures 4(a) to 4(h) are schematic diagrams of the simulation and experimental voltage-time curves of the eight lithium-ion batteries provided in the embodiments of this application;

[0055] Figure 5 A box plot diagram of model parameter residuals based on the interquartile range method provided in an embodiment of this application;

[0056] Figure 6A schematic diagram of the lumped parameters of the integrated model parameters of the dimensionless charge exchange current and the diffusion time constant of the eight lithium-ion batteries provided in the embodiments of this application;

[0057] Figure 7 A box plot diagram of the integrated model parameters for dimensionless charge exchange current based on the interquartile range method provided in the embodiments of this application;

[0058] Figure 8 A box plot diagram of lumped parameters for diffusion time constant based on the median absolute deviation method provided in the embodiments of this application;

[0059] Figure 9 This is a schematic diagram of the battery health status assessment system based on parameter identification provided in an embodiment of this application. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] Figure 1 The flowchart of the battery health status assessment method based on parameter identification provided in this application embodiment specifically includes the following steps:

[0062] S101: Construct a battery model.

[0063] In this embodiment, multiple batteries under test are simultaneously subjected to fault detection to assess their health status. Specifically, it is necessary to model the batteries under test and construct corresponding battery electrochemical models. By simulating the charge and discharge characteristics of the batteries through electrochemical models, the voltage response, capacity decay, and energy efficiency of the batteries under different operating conditions can be predicted, reflecting the relationship between the heat generated by the battery during operation and its internal temperature, as well as the electrochemical reaction rate and the battery's health status.

[0064] In one embodiment, the health status of each battery under test can be assessed under the Urban Dynamometer Driving Schedule (UDDS), which can well and comprehensively reflect the performance of the model.

[0065] In one embodiment, when constructing a battery model of the battery under test, it is necessary to determine several characteristic equations for constructing the battery model. The characteristic equations are mathematical expressions that describe the dynamic characteristics of the battery and are used to establish quantitative relationships between parameters such as voltage, current, and capacity and state variables (such as remaining charge SOC and temperature).

[0066] Specifically, a battery model is constructed using graphite as the negative electrode and LiFePO4 as the positive electrode. The battery model may include the following characteristic equations:

[0067] Based on the battery's nominal capacity, applied current, and average state of charge (SOC), the first characteristic equation is constructed, expressed as Formula 1: Where SOC represents the average state of charge of the battery, and t is time. I cell For the current applied to the battery, Q cell,0 This represents the battery's nominal capacity. This relationship indicates that during constant current charging and discharging, the SOC value is linearly proportional to the battery current.

[0068] Based on the open-circuit voltage, battery voltage, and overpotential term, a second characteristic equation for charging and discharging is constructed; expressed as Formula 2 based on charging: And Formula 3 based on discharge: ,in, It is the battery voltage value. V charge This is the battery's charging voltage, and OCV is the battery's open-circuit voltage. and These represent the ohmic overpotential, concentration overpotential, and activation overpotential, respectively. In Equations 2 and 3, the first term on the right-hand side gives the battery charging voltage and OCV as functions of SOC and battery temperature, respectively.

[0069] Overpotential defines the total voltage drop within a battery. In battery models, voltage loss is typically represented by ohmic overpotential, concentration overpotential, and activation overpotential.

[0070] Ohmic overpotential is the voltage loss caused by the ohmic resistance within the battery, which can be expressed as Equation 4: ,in, This represents the ohmic loss of the battery cell during 1C discharge, corresponding to the current required to discharge the battery to its cutoff voltage in one hour. I 1C This represents the battery discharge current during a 1C discharge process. I cellThis represents the battery current value. These values ​​are obtained experimentally and used for various SOC values ​​as input to the model. The battery discharge current at 1C can be expressed as Equation 5: ,in, This indicates the nominal capacity of the lithium-ion battery.

[0071] The activation overpotential is related to electrochemical kinetics and electron migration. The cell model can represent the activation overpotential during the process using Equation 6: Where R represents the gas constant, T represents the surface temperature of the battery, F represents the Faraday constant, and J0 represents the comprehensive model parameters of the dimensionless charge exchange current. η act This indicates the activation overpotential.

[0072] Concentration overpotential is related to mass transport limitations, and it significantly affects voltage drop as current density increases. The voltage drop caused by concentration overpotential during charge and discharge can be expressed as Equation 7: ,in, This indicates the voltage value at the surface of the open-circuit battery. η conc This indicates concentration overpotential.

[0073] At x=1, using the SOC value of the battery surface as a boundary condition, it can be expressed as Equation 8: Where x represents a dimensionless spatial variable, The lumped parameters representing the diffusion time constant are used to construct the third characteristic equation based on the applied current, nominal battery capacity, diffusion time constant, and dimensionless spatial variables. Simultaneously, a general energy conservation expression can be defined based on the average properties, expressed as Equation Nine: ,in, p Indicates density, c p Let represent the average specific heat, T represent the surface temperature, v represent the electrolyte velocity vector, k represent the thermal conductivity, and q represent the volumetric heat generation of the battery. If the battery pack composition is non-uniform, the thermal properties in Equation 9 can be considered anisotropic, with different values ​​in each direction. The first term of the equation defines the heat accumulation within the battery; this term involves the only unknown parameter in the equation, representing the temperature change of the battery. The second term on the left represents convection, which is crucial for batteries with flowing electrolytes but neglected for stationary batteries. The other terms represent conduction and heat generation, respectively.

[0074] The heat dissipation of the battery is stated as a boundary condition applied to the battery surface, which can be expressed as Equation 10: Where n represents the normal component, h represents the convective heat transfer coefficient, ε represents the surface emissivity, and σ represents the Stefan-Boltzmann constant. T surr The ambient temperature is represented by Equation 10. Equation 10 describes the rate of heat loss from the battery to the environment through convection and radiation. After applying the energy conservation equation to each part of the lumped battery model—the core, active cell section, and casing—Equation 10 is applied to the battery boundary, neglecting the effects of thermal radiation. It can be seen that the radiation effect becomes increasingly significant as the battery temperature increases.

[0075] For electrochemical heat sources, a fourth characteristic equation based on heat is constructed according to battery current, battery voltage, and open-circuit voltage to evaluate the heat generation within the battery during discharge. This equation can be expressed as Equation 11:

[0076] Where Q represents heat, defining the heat generation within the battery. Irreversible heat generation can be assessed through the first term on the right-hand side of the formula, while the second term provides the reversible heat source, which includes the derivative of the open-circuit potential with respect to temperature, also known as the entropy coefficient.

[0077] Using the above formulas, an electrochemical-thermal coupled lumped cell (i.e., the cell under test) can be modeled to obtain a cell model that characterizes the changes in various characteristic parameters of the cell under test.

[0078] S102: Based on the LM algorithm, model parameters are identified for each battery under test to obtain several characteristic parameters representing the health status of each battery under test.

[0079] This application's embodiments are based on the Levenberg-Marquardt (LM) algorithm, combined with the voltage and current data of the battery under test during testing, to identify model parameters for each battery under test, such as... Figure 2 As shown.

[0080] The LM algorithm combines the advantages of the Gauss-Newton method and gradient descent, and is a highly efficient numerical optimization method widely used in nonlinear least squares optimization problems. It can achieve a balance between convergence speed and stability by dynamically adjusting the step size, thus quickly and accurately minimizing the objective function. Its parameter iteration formula can be expressed as Equation XII: ,in, β k This represents the parameter vector at the k-th iteration. rLet J represent the residual vector, I represent the identity matrix, and λ represent the damping factor. Decreasing λ can accelerate convergence when processing speed is slow, while increasing λ can reduce the step size and improve accuracy to enhance stability. During iteration, the LM algorithm solves for the parameter increment ∆x using a modified Gauss-Newton equation. Iteration stops when the residual loss change is less than a threshold, the parameter change or gradient range norm is sufficiently small, or the maximum number of iterations is reached. As a nonlinear least squares optimization method, the LM algorithm demonstrates significant advantages in battery modeling and simulation parameter identification. Battery models typically exhibit highly nonlinear characteristics. The LM algorithm can adaptively adjust the damping factor during iteration, allowing it to quickly approach the optimal solution in the initial stage and accelerate convergence when close to the optimal solution. This characteristic makes it excellent at handling nonlinear parameters such as battery dynamic response and polarization resistance. Furthermore, the algorithm optimizes parameters by minimizing the sum of squared residuals, exhibiting strong robustness to noise in experimental data. Furthermore, compared to stochastic optimization methods such as Monte Carlo methods and genetic algorithms, LM calculations require less computation and do not involve complex parameter tuning. In lithium-ion battery models, it uses analytical Jacobian matrices or numerical difference for rapid iteration, avoiding the time-consuming issues of simulated annealing algorithms, making it suitable for real-time or online parameter estimation scenarios.

[0081] This method allows for parameter identification of time-voltage-current data from different lithium-ion batteries, thereby deriving characteristic parameter values ​​and residual losses for batteries in different states.

[0082] In one embodiment, the characteristic parameters to be identified include ohmic overpotential, dimensionless charge exchange current, and diffusion time constant. Changes in the values ​​of these characteristic parameters are used to detect battery malfunctions and assess the battery's health status. Specifically, when a battery model is pre-built, the correlation between the model's characteristic equations and the characteristic parameters to be identified can be obtained.

[0083] Ohmic overpotential ( η ohmic The integrated model parameters (J0) of the dimensionless charge exchange current and the lumped parameters (τ) of the diffusion time constant are typical and extremely important characteristic parameters in the operation of lithium-ion batteries.

[0084] Ohmic overpotential is the voltage drop caused by ohmic resistance when current flows through the internal materials of a battery (such as electrodes, electrolyte, and current collector). This parameter is a core parameter of the battery's instantaneous response, directly affecting the battery's charge and discharge voltage plateau and energy efficiency, especially under high-rate operating conditions. Furthermore, the Joule heat generated by ohmic resistance is one of the main sources of battery temperature rise. Accurately characterizing this parameter in a thermal model helps in the precise detection of battery thermal management.

[0085] Dimensionless charge-exchange current density is a dimensionless parameter characterizing the rate of electrode reaction kinetics and reflecting the ease of charge transfer at the interface. In lumped models, dimensionless treatment simplifies the complexity of multi-physics coupling, facilitates parameter calibration through experiments (such as electrochemical impedance spectroscopy), and helps improve the model's generalization ability.

[0086] The diffusion time constant characterizes the diffusion kinetics of lithium ions in electrode particles or electrolytes, capturing the macroscopic effects of lithium ion transport while ensuring simulation efficiency. Furthermore, diffusion performance degradation (such as electrode material structure collapse and SEI thickening) increases the time constant, making it one of the core parameters reflecting battery aging models. Simultaneously, the lumped model established in this application simplifies the distributed diffusion process into a lumped kinetic equation through the diffusion time constant, reducing computational complexity while maintaining accuracy, making it suitable for real-time state estimation of battery management systems (BMS).

[0087] S103: For each characteristic parameter and the corresponding residual loss, the test battery is analyzed by the interquartile range method or the median absolute deviation method to obtain the corresponding abnormal data in each characteristic parameter and the test battery corresponding to the abnormal data.

[0088] After identifying the characteristic parameter values ​​and residual losses of each battery under test, it is necessary to analyze these data using certain data analysis methods to identify any abnormal data in order to detect whether the battery is faulty.

[0089] The interquartile range (IQR) is a nonparametric statistical method based on the quantiles of data. It is mainly used to detect outliers in univariate data. Its core idea is to define a reasonable range of normal data by using the middle region of the data distribution (i.e., the IQR), and values ​​outside this range are considered potential anomalies.

[0090] The median absolute deviation method is a robust statistic for measuring the dispersion of data. It can effectively resist the interference of outliers in the data. By observing the dispersion of data around the median, it can identify outliers. It can remain stable even when there are a few extreme values ​​in the data, without being greatly affected by extreme values. Its principle is to use the median and median absolute deviation to replace the median and range in the upper and lower limit calculation formulas, and to calculate the median absolute deviation to avoid the influence of outliers on the threshold.

[0091] In this embodiment, the interquartile range (ICR) method can be used for data analysis of the ohmic overpotential or dimensionless charge exchange current among the characteristic parameters. However, for the diffusion time constant among the characteristic parameters, due to its large range and small sample size, the ICR method cannot accurately distinguish outliers. Therefore, the median absolute deviation method can be used for data analysis.

[0092] Furthermore, when using the interquartile range method to identify parameters of ohmic overpotential or dimensionless charge exchange current, it is necessary to determine the first quartile, third quartile, and interquartile range based on the residual value loss corresponding to the characteristic parameters of each battery under test after parameter identification. Based on the first quartile, third quartile, and interquartile range, the upper and lower limits of the normal data range are determined, and data exceeding the normal data range are regarded as abnormal data.

[0093] Specifically, the 25% and 75% quantiles of the identified residual loss data set are defined as the first quartile (IQR_divation.Q1) and the third quartile (IQR_divation.Q3), respectively, and the interquartile range (IQR_divation) is defined, which can be expressed as Equation Thirteen: Therefore, the lower limit of the normal data range can be expressed as Formula Fourteen: The upper limit can be expressed as formula fifteen: .

[0094] When using the median absolute deviation method to identify the diffusion time constant, multiple absolute deviation values ​​need to be determined based on the difference between the median of the diffusion time constant of each battery under test and the value of each characteristic parameter, as well as the median absolute deviation value. Based on the median, the median absolute deviation value, and the set multiplier, a threshold for the normal data range is determined, and data that exceeds the normal data range is regarded as abnormal data.

[0095] S104: Evaluate the health status of each battery under test based on the distribution of its characteristic parameters.

[0096] In this embodiment of the application, based on the abnormal data in the characteristic parameters of the battery under test, the possible fault type of the corresponding battery under test can be determined based on the characteristics of the characteristic parameters.

[0097] In addition, the feature parameters are divided into independent parameters and associated parameters. Independent parameters are used to independently evaluate the health status of the battery under test, while associated parameters are used to evaluate the health status of the battery under test in combination with other parameters.

[0098] Therefore, after identifying abnormal data, it is necessary to determine whether the feature parameter corresponding to the abnormal data is an independent parameter or a related parameter. If the feature parameter corresponding to the abnormal data is determined to be an independent parameter, the battery under test may be faulty. If the feature parameter corresponding to the abnormal data is determined to be a related parameter, the battery's faulty condition cannot be determined based on it alone. It is necessary to confirm that other related parameters also have abnormal data before assessing whether the battery under test may be faulty.

[0099] In this embodiment, by constructing a lumped electrochemical model, optimizing the parameter identification process using the LM algorithm, and introducing the interquartile range statistical method and the median absolute deviation method to analyze residual characteristics, a fault diagnosis strategy that integrates mechanism analysis and statistical discrimination is provided. This achieves efficient detection and accurate location of lithium-ion battery faults, improves the sensitivity and reliability of battery thermal runaway early warning, and provides theoretical support and technical optimization solutions for battery safety management systems.

[0100] The battery model constructed in this application accurately grasps the characteristic parameters used to judge the battery health status. It selects characteristic parameters that are typical and extremely important during battery operation, and reduces computational complexity while ensuring accuracy. It is suitable for real-time state estimation of battery management systems.

[0101] In one embodiment, Figure 3 This is a flowchart of another battery health status assessment method based on parameter identification proposed in this application. Figure 3 As shown, the battery is first modeled using the Levenberg-Marquardt (LM) algorithm, and model parameters are identified for different batteries using battery test data. After selecting the characteristic model parameters for battery health state estimation, the residual characteristics are analyzed using the interquartile range statistical method to determine whether the parameters are within the normal range. If so, the battery health state is normal; otherwise, abnormal batteries are identified based on the abnormal parameter results.

[0102] Based on the aforementioned parameter-based battery health status assessment method, this application provides a detailed illustration using eight lithium-ion batteries (numbered 1 to 8) as an example. Battery number 2 is designated as a faulty battery with an external short circuit, while the other batteries are considered normal.

[0103] Simulation and experimental analysis were performed on a lumped model of eight lithium-ion batteries within a time frame of 0 to 240 seconds. Voltage sampling analysis results with a sampling period of 1 second were obtained, and voltage-time data for different batteries were acquired.

[0104] Based on the voltage-time data of different batteries, the model parameters were identified, resulting in eight sets of voltage-time curves as shown in Figures 4(a) to (h). Statistical analysis of the model data and experimental data for the eight batteries (cell 1 to cell 8) shown in Figures 4(a) to (h) revealed that, for all tested batteries, the mean absolute error (MAE) between the simulated output voltage data and the experimentally measured data of the model constructed in this application was distributed in the range of 0.0031 to 0.00545, and the coefficient of determination (R-squared) of the fit was consistently within the range of 0.8743 to 0.9103. Furthermore, the model exhibited a small average error band width, demonstrating excellent simulation performance.

[0105] Specifically, Table 1 shows the values ​​of the mean absolute error and the coefficient of determination for the eight batteries.

[0106] Table 1: Correspondence between mean absolute error and coefficient of determination for different batteries

[0107]

[0108] The data in Table 1 show that the model has reached an advanced level in terms of global simulation accuracy. It accurately captures the voltage and current changes during the use of lithium-ion batteries, and the voltage-time data simulated by the model is in high agreement with the experimental data. It has good visualization features, and the model can be used to identify key internal parameters of the battery and to assess its health status based on the identification results.

[0109] As can be seen from the voltage-time plots in Figures 4(a) to (h), the lumped chemical model constructed in this application exhibits good simulation capabilities and robustness, with excellent fitting performance. Furthermore, the model successfully reproduces the key characteristics of the charge-discharge plateau and phase transition region. This demonstrates that the model accurately captures the ohmic polarization and concentration polarization characteristics of different batteries during the constant current charge-discharge stage, further validating the rationality of the model parameter identification and electrochemical-thermal coupling mechanism. In particular, when analyzing the transient behavior of the battery, it was found that the constructed model still possesses strong data tracking capabilities even in the face of dynamic processes with drastic voltage fluctuations, and can fully simulate the rate of voltage change. This demonstrates that the lumped model accurately grasps the characteristic parameters and characteristic equations of lithium-ion batteries. Simultaneously, the model can also capture small voltage fluctuations caused by temperature fluctuations or material phase transitions at different stages. This reflects that although the constructed lumped model does not introduce parameters for microscopic analysis, it can still clearly simulate and analyze parameter changes caused by microscopic mechanisms.

[0110] Considering a unique external short-circuit fault condition among the eight batteries, despite its unconventional state differences, the simulation accuracy showed no significant difference compared to the analysis results of the other seven normal batteries. This indicates that the model successfully achieved accurate characterization of battery voltage and internal state parameters under fault conditions, demonstrating special adaptability. Analysis of the fit diagrams for the faulty battery revealed that the fault model successfully reproduced the deviation of the experimental voltage from the standard curve. This demonstrates that the model has high accuracy in describing the kinetics of restricted ion diffusion and interfacial side reactions under extreme fault conditions.

[0111] Based on this, the model has strong simulation performance and provides a reliable simulation platform for the development of fault diagnosis algorithms for battery management systems. Its generalization ability under different health states and fault scenarios has been verified by multiple types of batteries and has the potential to be promoted in more market applications.

[0112] After parameter identification, Table 2 below shows the results obtained by the model based on the input "current-voltage-time" data and the LM algorithm. , J 0、 Parameter identification results for residual loss.

[0113] Table 2: Identification of Lumped Cell Parameters under Different States

[0114]

[0115] As shown in Table 2 and Figure 5 As shown, analysis of the residuals of the model parameters clearly reveals that the residual loss data of battery cell 2 far exceeds the upper limit, constituting a high-end outlier, i.e., an outlier. Using this statistical principle, abnormal batteries can be easily identified. Figure 5 In this context, IQR_lowest represents the minimum interquartile range, IQR_medium represents the median interquartile range, IQR_highest represents the maximum interquartile range, IQR_divation.Q1 represents the first interquartile range, and IQR_divation.Q3 represents the third interquartile range.

[0116] The following section analyzes the data for each of the three specified feature parameters.

[0117] First of all, Numerical analysis was performed. Ohmic overpotential is the additional voltage drop in an electrochemical system caused by the presence of ohmic resistance when current flows through it. The main factors that may cause this phenomenon include: impaired ion migration in the electrolyte, insufficient electronic conductivity or poor contact of the electrode itself, degradation of the electrode or electrolyte to form an insulating layer, and poor contact between the electrode and the current collector. These are all common failure modes of lithium-ion batteries. Furthermore, the design of batteries of the same model (such as material formulation, electrode thickness, electrolyte composition, etc.) is standardized, and their ohmic resistances are highly similar. If these batteries have almost identical SOC, temperature, and aging levels, and have not undergone any differential cycling or damage, their ohmic resistance differences are negligible. Therefore, by observing the numerical differences and changes, the battery failure situation can be well analyzed. Table 3 below shows the ohmic overpotential parameters of eight batteries.

[0118] Table 3: Ohmic overpotential parameters for different batteries

[0119]

[0120] As shown in Table 3, only cell 2 has a value different from the other cells. Although the difference is only about 0.9%, for a conservative parameter that has a significant impact on the battery, a small change may mean that the battery has suffered a serious failure, and it can be assessed that cell 2 has failed.

[0121] Secondly, such as Figure 6 The figure shows the integrated model parameters (J0) of the dimensionless charge exchange current and the lumped parameters of the diffusion time constant after parameter identification of the 8 battery cells. ).

[0122] When analyzing the parameters, it was found that the combined model parameter J0 for the dimensionless charge exchange current of cells 2 and 8 showed significant outlier values. Using the IQR statistical method described above, a range of normal data can be defined. The box plot obtained from the IRQ statistical method for J0 is shown below. Figure 7 As shown. Among them, Figure 7 The meaning of the parameters in Figure 5 The same applies, so I won't repeat it here.

[0123] based on Figure 7 It is known that the normal range of the comprehensive model parameters for the dimensionless charge exchange current is 1.07665~1.72505. However, the residual loss data of cell 2 is below the lower limit, which is a low-end outlier, i.e., an outlier. The parameters of cell 8, however, are within a reasonable range. Therefore, cell 2 can be preliminarily estimated as a faulty cell. However, many factors can cause an abnormal decrease in J0, such as: reduced electrode material activity, decreased temperature, increased interfacial impedance, increased diffusion rate, low SOC or high concentration gradient, phase separation or concentration polarization, loss of active material, and parameter coupling effects. These factors encompass many types of lithium-ion battery faults. Therefore, it is generally not possible to determine whether an external short-circuit fault has occurred by simply judging the difference in J0 values. Therefore, further analysis of the results after identifying other parameters is necessary.

[0124] Among the identified parameters, the lumped parameter of the diffusion time constant is... It is also an important indicator for describing the dynamic response of the diffusion process in lithium-ion battery materials, and can usually reflect the degree of dynamic response of the battery. Its classic definition can be expressed as Formula Sixteen: Where L represents the characteristic diffusion length such as electrode particle radius or electrode thickness, and D represents the effective diffusion coefficient of lithium ions, which mainly reflects the migration ability of lithium ions in the material. Therefore, A decrease in L could indicate an increase in D or a decrease in L. In this experiment, the batteries were of the same type, and the differences in L were minimal. Therefore, the difference between faulty and normal batteries is primarily determined by parameter D. A decrease in the rate indicates a stronger charge / discharge capability of the battery. Abnormal charge / discharge rates are generally caused by battery malfunctions, which lead to increased heat generation and potentially thermal runaway.

[0125] Median absolute deviation (MAD) method was used to... Difference detection was performed. The median was set to Median = 3891, and the median absolute deviation (MAD) was set to 70.1. An n value of 3 was selected, and the adjusted threshold ranged from 3680.7 to 4101.3. The corresponding MAD statistical box plot is shown below. Figure 8 As shown in the figure. Here, Median represents the median value, MAD_highest represents the maximum median deviation, and MAD_lowest represents the minimum median deviation.

[0126] Except for cell 2, cell 8 has An anomaly also occurred, but its J0 value remained within the normal range. Based on comprehensive analysis, cell 8 did not have an external short circuit fault, while cell 2 did.

[0127] Fault diagnosis of lithium-ion batteries exhibits significant multi-dimensional characteristics. The complexity of their failure mechanisms means that isolated analysis of single physical / chemical parameters is insufficient for accurate identification. The nonlinear degradation behavior of battery systems typically manifests as coupled correlations of multi-source heterogeneous data, necessitating the construction of a systematic analysis framework based on multi-dimensional characteristic parameters. Detailed analysis of different types of parameters is crucial to ensure accurate fault identification. Generally, system debugging with preset parameters is used to assess abnormal values ​​such as excessive residual value loss after identification, abnormal increases in ohmic overpotential, and anomalies in dimensionless charge exchange current and diffusion time constant. If two or more anomalies are present, a fault check of the battery is required to ensure safety.

[0128] It should be noted that the battery health status assessment method proposed in this application is not limited to lithium-ion batteries, but can also be used for fault detection and health status assessment of other battery types with the same principle. This application does not limit this application.

[0129] The above describes a battery health status assessment method based on parameter identification, as provided in the embodiments of this application. Based on the same inventive concept, the embodiments of this application also provide a corresponding battery health status assessment device based on parameter identification, such as... Figure 9 As shown.

[0130] Figure 9A schematic diagram of the battery health status assessment device based on parameter identification provided in this application embodiment specifically includes:

[0131] Module 901 is used to build the battery model;

[0132] The identification module 902 is used to identify the model parameters of each battery under test based on the LM algorithm, and obtain several feature parameters that characterize the health status of each battery under test.

[0133] Analysis module 903 is used to analyze the battery under test for each characteristic parameter and the corresponding residual loss using the interquartile range method or the median absolute deviation method, and to obtain the corresponding abnormal data in each characteristic parameter and the battery under test corresponding to the abnormal data.

[0134] Evaluation module 904 is used to evaluate the health status of each battery under test based on the distribution of characteristic parameters of each battery under test.

[0135] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0136] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0137] The systems and methods provided in this application are one-to-one correspondences. Therefore, the system also has similar beneficial technical effects as its corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system will not be repeated here.

[0138] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using dedicated hardware combined with computer instructions.

[0139] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0140] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0141] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A battery health status assessment method based on parameter identification, characterized in that, include: Build a battery model; Based on the LM algorithm, model parameters are identified for each battery under test, and several characteristic parameters representing the health status of each battery under test are obtained. For each characteristic parameter and the corresponding residual loss, the test battery is analyzed by the interquartile range method or the median absolute deviation method to obtain the corresponding abnormal data in each characteristic parameter and the test battery corresponding to the abnormal data. The health status of each battery under test is assessed based on the distribution of its characteristic parameters. The construction of the battery model includes: Several characteristic equations are determined to construct the battery model; these characteristic equations characterize the changes in various characteristic parameters of the battery under test, wherein, Based on the battery's nominal capacity, applied current, and average state of charge (SOC), a first characteristic equation is constructed; the first characteristic equation is expressed as follows: Where SOC represents the average state of charge of the battery, t represents time, Icell represents the current applied to the battery, and Qcell,0 represents the nominal capacity of the battery. Based on the open-circuit voltage, battery voltage, and overpotential term, a second characteristic equation for charging and discharging is constructed; this second characteristic equation is expressed as a charging-based equation. and discharge-based ,in, The values ​​represent the battery voltage: Vcharge represents the battery charging voltage, OCV represents the battery open-circuit voltage, and T represents the surface temperature. and These represent ohmic overpotential, concentration overpotential, and activation overpotential, respectively. Based on the applied current, battery nominal capacity, diffusion time constant, and dimensionless spatial variables, a third characteristic equation is constructed; the third characteristic equation is expressed as follows: Where x represents a dimensionless spatial variable, Lumped parameters representing the diffusion time constant; Based on battery current, battery voltage, and open-circuit voltage, a fourth characteristic equation based on heat is constructed.

2. The battery health status assessment method based on parameter identification according to claim 1, characterized in that, The characteristic parameters include ohmic overpotential, dimensionless charge exchange current, and diffusion time constant; The construction of the battery model includes: Determine the relationship between the characteristic equations and the characteristic parameters used to construct the battery model.

3. The battery health status assessment method based on parameter identification according to claim 1, characterized in that, The LM algorithm-based model parameter identification for each battery under test includes: Based on the LM algorithm, and combined with the voltage and current data of the battery under test during the test, the model parameters of each battery under test are identified. In the parameter identification process, the parameters are iterated based on the Jacobian matrix, identity matrix and damping factor of the residual. The damping factor is adaptively adjusted until the change in residual loss is less than the first threshold, the change in parameter or gradient range norm is less than the second threshold, or the maximum number of iterations is reached, at which point the iteration stops.

4. The battery health status assessment method based on parameter identification according to claim 2, characterized in that, When the characteristic parameter is an ohmic overpotential or a dimensionless charge exchange current, the test cell is analyzed using the interquartile range method or the median absolute deviation method for each characteristic parameter and the corresponding residual loss, to obtain the corresponding abnormal data in each characteristic parameter and the test cell corresponding to the abnormal data, including: Based on the residual value loss corresponding to the characteristic parameters of each battery under test after parameter identification, the first quartile, the third quartile, and the interquartile range are determined by the interquartile range method. Based on the first quartile, the third quartile, and the interquartile range, determine the upper and lower limits of the normal data range; Data that exceeds the normal data range will be considered abnormal data.

5. The battery health status assessment method based on parameter identification according to claim 2, characterized in that, When the characteristic parameter is a diffusion time constant, the test cell is analyzed using the interquartile range method or the median absolute deviation method for each characteristic parameter and the corresponding residual loss, to obtain the corresponding abnormal data in each characteristic parameter and the test cell corresponding to the abnormal data, including: Based on the difference between the median diffusion time constant of each battery under test and the value of each characteristic parameter, multiple absolute deviation values ​​are determined, as well as the median absolute deviation value. The threshold for the normal data range is determined based on the median, the absolute deviation of the median, and the set multiple. Data that exceeds the normal data range will be considered abnormal data.

6. The battery health status assessment method based on parameter identification according to claim 1, characterized in that, The assessment of the health status of each battery under test based on the distribution of its characteristic parameters includes: Based on the abnormal data in the characteristic parameters of the battery under test, determine the possible fault type of the corresponding battery under test.

7. The battery health status assessment method based on parameter identification according to claim 1, characterized in that, The assessment of the health status of each battery under test based on the distribution of its characteristic parameters includes: The feature parameters are divided into independent parameters and associated parameters; If the characteristic parameter corresponding to the abnormal data is determined to be an independent parameter, then the battery under test may be faulty. If the feature parameter corresponding to the abnormal data is determined to be a correlation parameter, and other correlation parameters are also confirmed to have abnormal data, the battery under test may be faulty.

8. The battery health status assessment method based on parameter identification according to claim 1, characterized in that, Before identifying the model parameters for each battery under test, the method further includes: Under urban operating conditions, the health status of each battery under test is assessed.

9. A battery health status assessment system based on parameter identification, characterized in that, The battery health status assessment method based on parameter identification as described in claim 1 is implemented, wherein the system comprises: Modules for building battery models; The identification module is used to identify the model parameters of each battery under test based on the LM algorithm, and obtain several feature parameters that characterize the health status of each battery under test. The analysis module is used to analyze the battery under test for each characteristic parameter and the corresponding residual loss using the interquartile range method or the median absolute deviation method, and to obtain the corresponding abnormal data in each characteristic parameter and the battery under test corresponding to the abnormal data. The evaluation module is used to assess the health status of each battery under test based on the distribution of its characteristic parameters.

Images