Lithium battery multi-fault real data set construction method based on physical coupling model

Abstract

The invention discloses a lithium battery multi-fault real data set construction method based on a physical coupling model, and relates to the technical field of lithium ion battery fault diagnosis and data modeling. High-fidelity modeling and data generation of complex multi-fault behaviors are realized by constructing a physical model integrating electric, thermal, internal short circuit (ISC) and aging modules. The modular framework can capture cross-domain coupling effects such as thermal stress accelerated aging and internal short circuit disturbance electro-thermal response, and the fault evolution process is truly restored. According to the method, open source battery data and verified typical fault features of the Wishball University are fused, multi-fault samples are reconstructed through modeling enhancement, and key electrochemical response is reserved. A constructed data set covers 312 operating conditions, including ambient temperature (-20 DEG C to 40 DEG C), charge-discharge rate (0.5 C to 2C), typical discharge conditions (such as WLTP and NEDC) and different initial SOCs (0% to 100%), and provides high-credibility data support for lithium battery fault diagnosis.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of lithium battery fault diagnosis, and is a lithium battery fault modeling and multi-fault multi-condition data set construction method based on physical coupling. BACKGROUND

[0002] Lithium-ion power batteries are the core energy units of electric vehicles, and their safe and reliable operation directly determines the performance of the vehicle and the safety of the user. A high-quality, accurately labeled, and widely covering working condition fault data set is the basis for developing advanced diagnostic algorithms. However, the existing public data sets generally have a sampling frequency lower than 1 Hz, making it difficult to capture millisecond-level transient signals. The running conditions are highly concentrated in room temperature or a single rate, lacking extreme scenarios such as low temperature, high temperature, and variable load. The fault sample ratio is usually less than 5%, and the label information is incomplete, limiting the robustness of supervised learning models. Laboratory data often only undergoes a single test at 25°C, lacking repeated design across temperatures and multiple cycles. In summary, there is a need for a high-resolution battery fault data set that can simultaneously collect voltage, temperature, and state of charge multi-dimensional signals in a controllable environment, covering full-temperature, multi-current, and multi-SOC conditions, and providing rich fault types and complete labels, to support the further development of lithium-ion battery fault diagnosis, aging prediction, and safety control technology.

[0003] Patent document CN118584349A discloses a lithium-ion battery zero-sample internal short-circuit fault diagnosis method and system based on joint dictionary learning. The battery's electrical and thermal performance characteristics are considered through a battery electrical-thermal coupling model. Based on the fault mechanism, simulated fault data sets of lithium batteries are obtained by fault injection into the model, including internal short circuit, overheating, low temperature environment, and contact resistance fault. The attribute knowledge base of lithium-ion battery faults is defined based on expert knowledge. The fault time series data and attribute knowledge of collectable samples are projected into a shared subspace based on joint dictionary learning theory. The attribute knowledge of difficult-to-collect samples is introduced into the model. The fault discriminative constraint is introduced to accurately distinguish different faults in the shared subspace. Even in the absence of some fault samples, the fault can still be accurately diagnosed, providing an efficient and reliable solution for lithium-ion battery fault diagnosis in energy storage systems. SUMMARY

[0004] The present application aims to solve the problems of the above prior art. A lithium battery fault modeling and multi-fault multi-condition data set construction method based on physical coupling is proposed. The technical solution of the present application is as follows:

[0005] A lithium battery multi-fault real data set construction method based on a physical coupling model, comprising the following steps:

[0006] S1, electrical model establishment: based on the experimental data of 18650 lithium ion batteries published by the University of Wisconsin-Madison, the open-circuit voltage, ohmic resistance R0 and two groups of RC networks (R1C1, R2C2) of the second-order equivalent circuit model are parameterized;

[0007] S2, parameter optimization: the parameters R1, R2, C1 and C2 are iteratively optimized by a particle swarm optimization algorithm, and the optimal parameter set is obtained by taking the experimental-simulation voltage error as the objective function;

[0008] S3, thermal model establishment: on the basis of the electrical model, the heat of the battery and the environmental convection heat transfer are introduced, the total heat of the battery is obtained, and the battery temperature is obtained, realizing the electrical-thermal coupling simulation;

[0009] S4, internal short circuit ISC fault modeling: by applying a controllable external short circuit, the short circuit current is calculated, and the electrical-thermal response under different short circuit levels is obtained;

[0010] S5, aging model introduction: a semi-empirical aging model is used to calculate the capacity attenuation, and the cycle aging effect is quantified;

[0011] S6, multi-working condition simulation and data acquisition: under the combination of temperature -20℃-40℃, charge-discharge rate 0.5C-2C, typical discharge working condition WLTP / NEDC and initial SOC 0%-100%, the multi-fault data is obtained by coupling steps S1-S5;

[0012] S7, data set generation: the voltage, temperature and SOC time series data obtained by simulation are normalized and labeled with fault types to form a data set that can be used for lithium battery fault diagnosis machine learning training.

[0013] Further, the second-order equivalent circuit model includes a series-connected open-circuit voltage source OCV with steady-state characteristics, an ohmic resistance R0 reflecting instantaneous voltage drop, and two groups of resistance-capacitance parallel branches R1C1 and R2C2 respectively used to represent short-time and long-time dynamic responses.

[0014] Further, the current-voltage relationship of the circuit model is described by the following equations:

[0015]

[0016] The two equations respectively describe the voltage and current relationship of the two RC circuits inside the battery, the first RC circuit is composed of capacitor C1 and resistor R1, the second RC circuit is composed of capacitor C2 and resistor R2, and I is the applied charging current.

[0017] Battery output voltage equation:

[0018] V mod =UOCV -U1-U2-IR0

[0019] where V mod represents the output voltage of the battery, U OCV is the open circuit voltage, U1, U2 represent the voltage of two RC circuits respectively, and R0 is the internal resistance of the battery.

[0020] Further, the particle swarm optimization algorithm takes the mean square error between the experimental voltage and the simulation voltage as the objective function, and under the working condition of 25℃, the maximum voltage error of the optimized model is less than 1%, and the average absolute error is less than 0.03%, including the following steps:

[0021] The objective function is used to measure the error between the simulation voltage and the experimental voltage, and the mean square error (MSE) is used,

[0022]

[0023] where V exp,i is the experimental voltage at the i th time point, V sim,i is the simulation voltage at the i th time point calculated by the model, and n is the total number of voltage data points,

[0024] Set the particle swarm optimization parameters: including the number of particles, the maximum number of iterations, and the parameter boundaries of R1, R2, C1 and C2, each particle represents a four-dimensional vector [R1, R2, C1, C2];

[0025] Initialize the particle swarm: randomly initialize the position and velocity of 100 particles, and calculate the objective function value of each particle, set the initial global optimal solution and individual optimal solution;

[0026] Calculate fitness:

[0027] Iterative update:

[0028] Update velocity:

[0029] Update position:

[0030] where is the velocity of the i th particle in the t th generation, is the position of the i th particle in the t th generation, is the historical optimal position of the particle, g best is the optimal position of all particles, w is the inertia weight, c1 and c2 are learning factors, and r1 and r2 are random numbers between 0 and 1;

[0031] Update individual optimal and global optimal: if the current error of the particle is less than its historical optimal, update pbest ; if current error < global history best, then update g best ;

[0032] Judge the termination condition: if the maximum number of iterations is reached or the error converges, stop the loop;

[0033] Output the final optimal parameter combination: R1, R2, C1, C2.

[0034] Further, the thermal model also considers Ohmic heat, polarization heat, reaction heat and internal short circuit heat at the same time, and the calculation formula is as follows:

[0035] Ohmic heat: Q ohm = I 2 R0Δt

[0036] Polarization heat: Q p = I 2 R1Δt+I 2 R2Δt

[0037] Reaction heat:

[0038] Internal resistance heat: Q ISC = I 2 R ISC Δt

[0039] Environmental convection heat dissipation: Q exchange = hA(T b -T Ambient )

[0040] Total heat generation: Q total = Q ohm + Q p + Q r + Q ISC -Q exchange

[0041] Where I is the current, R0 is the Ohmic resistance, Δt is the time increment, R1 and R2 represent the polarization resistance of the battery, T b is the temperature of the battery, ΔS is the entropy coefficient, R ISC represents the short circuit resistance, h is the heat transfer coefficient, A is the surface area, T Ambient is the ambient temperature, and V ocv represents the open circuit voltage.

[0042] Further, the internal short circuit fault current is calculated as follows:

[0043]

[0044] Where R ISC represents the equivalent fault resistance, is the current flowing through the short circuit path, U t is the terminal voltage of the lithium battery.

[0045] Further, the aging model adopts a semi-empirical aging model to calculate the capacity decay, which can effectively capture two main aging acceleration factors: working current (I) and battery temperature (T), and the relative capacity loss Q loss is:

[0046]

[0047] where B is the pre-exponential factor, E a is the activation energy of cycle aging, alpha is the aging acceleration coefficient caused by current, I is the applied charging current, R represents the ideal gas constant, Ah is the cumulative charge and discharge amount, and z represents the aging model parameter.

[0048] The application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the lithium battery multi-fault real data set construction method based on the physical coupling model.

[0049] The application finally provides an electronic device, which comprises:

[0050] a memory for storing executable instructions;

[0051] a processor for executing the executable instructions stored in the memory to realize the lithium battery multi-fault real data set construction method based on the physical coupling model.

[0052] The advantages and beneficial effects of the application are as follows:

[0053] The application synchronously collects multi-dimensional time sequence signals such as voltage (U), temperature (T) and state of charge (SOC) at intervals of 1 second, covers eight typical fault modes including internal short circuit and aging, and can truly and comprehensively reflect the battery failure characteristics. The test process includes environmental temperature (-20℃ to 40℃), charge and discharge rate (0.5C-2C), typical discharge working conditions (such as WLTP, NEDC) and different initial SOC (0%-100%), realizes fault triggering and recording under multiple working conditions. The collected data are accompanied by detailed metadata such as fault type, severity and test conditions, and the proportion of fault samples is more than 80%, which provides a high-quality, publicly available benchmark resource for the development and verification of fault diagnosis models.

[0054] To evaluate the physical consistency and physical credibility of the dataset, the invention compared the internal short circuit (ISC) fault characteristics with real experimental data for benchmarking. The results show that the invention not only accurately reproduces the electrochemical behavior of real ISC, but also retains the charge rate-dependent fault characteristics, providing highly credible real labels for fault diagnosis model development.

[0055] Compared with patent document CN118584349A, the data set construction method of the invention has significant advantages in physical modeling rigor, simulation result authenticity verification, working condition combination coverage, composite fault modeling capability, and data set annotation integrity. The data set construction method in patent document CN118584349A mainly relies on model injection and artificial definition of attribute labels to generate simulation data, lacks consistency verification with real experimental data, has limited working condition coverage and fault types, and does not support composite fault modeling. While the invention optimizes the construction of physical coupling models based on real experimental data, systematically integrates electro-thermal-aging multi-domain coupling mechanisms, can accurately reproduce the evolution process of internal short circuit and other faults, supports large-scale multi-working condition simulation, and generates high-credibility data sets with complete metadata, significantly improving the authenticity, richness and application promotion value of the data. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A flowchart of a lithium battery fault modeling and multi-fault multi-working condition data set construction method based on physical coupling is provided for the embodiments of the invention.

[0057] Figure 2 A lithium battery fault data set research framework diagram is provided.

[0058] Figure 3 A comparison and analysis diagram of internal short circuit fault characteristic experimental data and simulation results under different charging rates is provided.

[0059] Figure 4 A voltage and SOC change diagram between real experimental battery faults and the data set of the invention under the same cycle conditions is provided.

[0060] Figure 5 A charging cycle and discharging cycle SOC change diagram is provided. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the invention will be described clearly and completely below with reference to the drawings in the embodiments of the invention. The described embodiments are only a part of the embodiments of the invention, not all.

[0062] As shown in Figure 1 A lithium battery fault modeling and multi-fault multi-working condition data set construction method based on physical coupling, which comprises the following steps:

[0063] S1, electrical model establishment: based on the experimental data of 18650 lithium-ion batteries published by the University of Wisconsin-Madison, the open-circuit voltage, ohmic resistance R0 and two groups of RC networks (R1C1, R2C2) of the second-order equivalent circuit model are parameterized;

[0064] S2, parameter optimization: the parameters R1, R2, C1 and C2 are iteratively optimized by a particle swarm optimization algorithm, and the optimal parameter set is obtained by taking the experimental-simulation voltage error as the objective function;

[0065] S3, thermal model establishment: on the basis of the electrical model, the battery heat and environmental convection heat transfer are introduced to obtain the total heat of the battery and thus the battery temperature, realizing the electrical-thermal coupling simulation;

[0066] S4, internal short circuit (ISC) fault modeling: by applying a controllable external short circuit, the short circuit current is calculated to obtain the electrical-thermal response under different short circuit levels;

[0067] S5, introduction of aging model: a semi-empirical aging model is used to calculate the capacity decay to quantify the cycle aging effect;

[0068] S6, multi-working condition simulation and data acquisition: under the combination of temperature -20℃-40℃, charge-discharge rate 0.5C-2C, typical discharge working condition WLTP / NEDC and initial SOC 0%-100%, the multi-fault data are obtained by coupling steps S1-S5;

[0069] S7, data set generation: the voltage, temperature and SOC time series data obtained by simulation are normalized and labeled with fault types to form a data set that can be used for lithium battery fault diagnosis machine learning training.

[0070] In step S1, the electrical model adopts a second-order equivalent circuit structure, which specifically includes: an open-circuit voltage source (OCV) representing steady-state characteristics, an ohmic resistance R1 reflecting instantaneous voltage drop, and two groups of resistance-capacitance parallel branches R1C1 and R2C2 respectively used to represent short-time and long-time dynamic responses. The current-voltage relationship of the model can be described by the following equations:

[0071] Current equation:

[0072]

[0073] These two equations describe the voltage and current relationship of the two RC circuits inside the battery respectively. The first RC circuit is composed of capacitor C1 and resistor R1, and the second RC circuit is composed of capacitor C2 and resistor R2, and I is the applied charging current. They jointly affect the dynamic change of the battery voltage.

[0074] Battery output voltage equation:

[0075] V mod = U OCV - U1 - U2 - IR0

[0076] The equation represents the output voltage V mod of the battery, where U OCV is the open circuit voltage, U1 and U2 represent the voltages of the two RC circuits, respectively, and R0 is the internal resistance of the battery.

[0077] In step S2, the particle swarm optimization algorithm uses a population size of 100 and a maximum number of iterations of 600, with the mean square error between the experimental voltage and the simulated voltage as the objective function. Under the working condition of 25°C, the maximum voltage error of the optimized model is less than 1%, and the average absolute error is less than 0.03%. The optimization steps are as follows:

[0078] Step 1, define the objective function

[0079] The objective function is used to measure the error between the simulated voltage and the experimental voltage, and the mean square error (MSE) is used.

[0080]

[0081] where V exp,i is the experimental voltage at the i-th time point, V sim,i is the simulated voltage at the i-th time point calculated by the model, and n is the total number of voltage data points.

[0082] Step 2, set the particle swarm optimization parameters

[0083] Number of particles: 100, maximum number of iterations: 600, parameter boundary setting: R1 ∈ [0.001, 0.05] Ω, R2 ∈ [0.001, 0.03] Ω, C1 ∈ [10, 1,000,000] F, C2 ∈ [10, 1,000,000] F. Each particle represents a four-dimensional vector [R1, R2, C1, C2].

[0084] Step 3, initialize the particle swarm

[0085] Randomly initialize the position and velocity of 100 particles and calculate the objective function value of each particle. Set the initial global optimal solution and individual optimal solution.

[0086] Step 4, calculate the fitness:

[0087] Step 5, iterative update

[0088] Update the velocity:

[0089] Update the position:

[0090] where is the velocity of the ith particle in the tth generation, is the position of the ith particle in the tth generation, is the historical optimal position of the particle, g best is the global optimal position of all particles. w is the inertia weight, controlling the convergence speed and global search ability, c1, c2 are learning factors, generally taken as 2, r1, r2 are random numbers between (0, 1), increasing randomness.

[0091] Step 6, update individual optimal and global optimal

[0092] If the current error of the particle < its historical optimal, update p best ;

[0093] If the current error < global historical optimal, update g best .

[0094] Step 7, judge termination condition

[0095] If the maximum number of iterations or error convergence is reached, stop the loop.

[0096] Step 8, output the final optimal parameter combination: R1, R2, C1, C2.

[0097] In step S3, the established thermal model simultaneously considers Ohmic heat, polarization heat, reaction heat and internal short circuit heat, which are calculated according to the following formulas respectively:

[0098] Ohmic heat: Q ohm = I 2 R0Δt

[0099] Polarization heat: Q p = I 2 R1Δt+I 2 R2Δt

[0100] Reaction heat:

[0101] Internal resistance heat: Q ISC = I 2 R ISC Δt

[0102] Environmental convection heat dissipation: Q exchange = hA(T b -T Ambient )

[0103] Total heat generation: Q total = Q ohm + Q p + Qr +Q ISC -Q exchange

[0104] where I is the current, R0 is the ohmic resistance, Δt is the time increment, R1 and R2 represent the polarization resistance of the battery, T b is the temperature of the battery, ΔS is the entropy coefficient, R ISC represents the short-circuit resistance, h is the heat transfer coefficient, A is the surface area, T Ambient is the ambient temperature, V ocv represents the open-circuit voltage.

[0105] In step S4, the internal short-circuit fault model adopts a controlled external resistance to systematically reproduce the internal short-circuit failure mode, characterize the voltage / current characteristics after the occurrence of the internal short-circuit fault, and analyze the dynamic process of fault evolution. The internal short-circuit fault current is calculated as follows:

[0106]

[0107] where R ISC represents the equivalent fault resistance, is the current flowing through the short-circuit path, U t is the terminal voltage of the lithium battery. The current distribution between the short-circuit path and the normal output terminal follows Kirchhoff's law, thereby providing an important basis for judging the severity of the fault.

[0108] In step S5, the aging model mainly focuses on the degradation of lithium-ion batteries caused by cycling during charging and discharging. This model uses a well-verified semi-empirical method to quantitatively evaluate the battery's aging process, effectively capturing two main aging acceleration factors: working current (I) and battery temperature (T). The relative capacity loss Q loss of the battery is:

[0109]

[0110] where B is the pre-exponential factor, E a is the activation energy for cycle aging, a is the aging acceleration coefficient caused by the current, I is the applied charging current, R represents the ideal gas constant, Ah is the cumulative charge and discharge amount, and z represents the aging model parameter.

[0111] In step S6, according to the framework diagram as shown in Figure 2 , under the combination of temperature -20℃-40℃, charge and discharge rate 0.5C-2C, typical discharge working conditions WLTP / NEDC, and initial SOC 0%-100%, the multi-fault data is obtained by coupling steps S1-S5.

[0112] Experimental analysis verification:

[0113] To evaluate the physical consistency and reliability of the dataset, the internal short circuit (ISC) failure signatures were benchmarked against real experimental results. Figure 3 The comparison of the literature experimental data and the corresponding charge voltage curves in this dataset under different charge rates and different ISC severity conditions showed that the voltage drop patterns were highly consistent, the failure signatures were sensitive to the charge rate, and the failure severity was positively correlated with the voltage offset. Figure 4 Further, it was found that in the first charge-discharge cycle, the ISC failure battery (Cell 2) had a voltage drop of about 35 ± 5 mV compared to the normal battery (Cell 1). As the number of cycles increased, the voltage difference continued to widen, reflecting the increasing severity of the ISC effect. At the same time, the SOC of the ISC battery continued to decrease in continuous cycles, while the SOCs of the normal battery and the slightly aged batteries (Cells 1, 3, and 4) remained stable, with only slight fluctuations. From the SOC difference between the normal battery and the ISC battery, it can be seen that the SOC difference between the normal battery and the ISC battery increased with the number of cycles, which further confirmed the increasing severity of the ISC effect. Figure 5 It can be seen that in the discharge phase, the SOC of the ISC battery decreased faster than that of the normal battery (a: 8.5%-17.8%; b: 7.0%-13.7%), and the SOC of the aged battery also decreased faster than that of the normal battery (a: 5.7%-10.1%; b: 6.0%-11.3%). In the case of multiple failures (SOH 90%+ISC 50), the SOC decrease rate was more than 15% higher than that of the normal battery, reflecting the failure superposition effect of the battery. In the charging phase, the SOC of the ISC battery increased slower than that of the normal battery (c: 7.2%-11.9%; d: 8.3%-10.3%), while the aged battery showed a faster upward trend (c: 3.0%-7.5%; d: 2.8%-6.4%). The results confirmed that the dataset accurately captured the subtle but failure-distinguishable SOC dynamics, maintained the electrochemical consistency under different working conditions, and accurately modeled the real failure progression.

[0114] In summary, the dataset of the present application can accurately reproduce:

[0115] (1) typical voltage / SOC characteristics induced by internal short circuit and aging;

[0116] (2) the SOC divergence pattern between aged batteries, failure batteries, and healthy batteries;

[0117] (3) the observed trends of the batteries under the same cycling conditions.

[0118] The above results confirmed the physical reliability of the dataset in failure characterization, provided reliable real labels for diagnostic model development, and had application value as a public benchmark resource.

[0119] The above examples further illustrate the objects, technical solutions and advantages of the present application. It should be understood that the above examples are only preferred embodiments of the present application and are not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made to the present application within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for constructing a real data set of multiple faults of lithium batteries based on a physical coupling model, characterized in that: The following steps are involved: S1. Electrical model establishment: Based on the 18650 lithium-ion battery experimental data published by the University of Wisconsin-Madison, the open-circuit voltage, ohmic internal resistance R0, and two RC networks (R1C1 and R2C2) of the second-order equivalent circuit model were parameterized; S2. Parameter optimization: The parameters R1, R2, C1, and C2 are iteratively optimized using the particle swarm optimization algorithm, and the optimal parameter set is obtained using the experimental-simulation voltage error as the objective function; S3. Thermal model establishment: Based on the electrical model, the battery heat and ambient convection heat transfer are introduced to obtain the total battery heat and thus the battery temperature, realizing electrical-thermal coupling simulation; S4. Internal short circuit (ISC) fault modeling: By applying a controllable external short circuit voltage, the short circuit current is calculated to obtain the electro-thermal response under different short circuit levels; S5. Introduction of aging model: A semi-empirical aging model is used to calculate capacity decay and quantify the cycle aging effect; S6. Multi-operating condition simulation and data acquisition: Steps S1-S5 are coupled and executed under the following conditions: -20°C-40°C temperature, 0.5C-2C charge / discharge rate, typical WLTP / NEDC discharge operating conditions, and 0%-100% initial SOC combination to obtain multi-fault data. S7. Dataset generation: Normalize the voltage, temperature, and SOC time series data obtained from the simulation and label the fault types to form a dataset that can be used for machine learning training of lithium battery fault diagnosis.

2. The method for constructing a real lithium battery multi-fault dataset based on a physical coupling model according to claim 1, characterized in that: The second-order equivalent circuit model includes a series-connected open-circuit voltage source OCV with steady-state characteristics, an ohmic internal resistance R0 reflecting an instantaneous voltage drop, and two sets of resistor-capacitor parallel branches R1C1 and R2C2 for characterizing short-term and long-term dynamic responses, respectively.

3. The method for constructing a real data set of multiple faults of lithium batteries based on a physical coupling model according to claim 2, characterized in that: The current and voltage relationship of the circuit model is described by the following equation: These two equations describe the relationship between the voltage and current of the two RC circuits inside the battery. The first RC circuit consists of capacitor C1 and resistor R1, and the second RC circuit consists of capacitor C2 and resistor R2. I is the applied charging current. Battery output voltage equation: V mod =U OCV -U1-U2-IR0 Where V mod Indicates the output voltage of the battery, U OCV is the open circuit voltage, U1 and U2 represent the voltages of the two RC circuits, where R0 is the internal resistance of the battery.

4. The method for constructing a real lithium battery multi-fault dataset based on a physical coupling model according to claim 1, characterized in that: The particle swarm optimization algorithm uses the mean square error between the experimental voltage and the simulated voltage as the objective function. Under the working condition of 25°C, the maximum voltage error of the optimized model is less than 1%, and the mean absolute error is less than 0.03%. It includes the following steps: The objective function is used to measure the error between the simulated voltage and the experimental voltage, using the mean square error (MSE). Where V exp,i is the experimental voltage at the i-th time point, V sim,i is the simulated voltage at the i-th time point calculated by the model, n is the total number of voltage data points, Set the particle swarm optimization parameters: including the number of particles, the maximum number of iterations, and the parameter boundaries of R1, R2, C1, and C2. Each particle represents a four-dimensional vector [R1, R2, C1, C2]. Initialize the particle swarm: Randomly initialize the positions and velocities of 100 particles, calculate the objective function value of each particle, and set the initial global optimal solution and individual optimal solution; Calculate fitness: Iterative updates: Update speed: Update location: in is the velocity of the ith particle of generation t, is the position of the i-th particle in the t-th generation, is the historical optimal position of the particle, g best is the optimal position of all particles, w is the inertia weight, c1 and c2 are learning factors, and r1 and r2 are random numbers between (0,1); Update individual optimal and global optimal: If the particle's current error is less than its historical optimal, update p best ; If the current error < global historical optimal, update g best ; Determine the termination condition: if the maximum number of iterations is reached or the error converges, stop the loop; Output the final optimal parameter combination: R1, R2, C1, C2.

5. The method for constructing a real data set of multiple faults of lithium batteries based on a physical coupling model according to claim 1, characterized in that: The thermal model also takes into account ohmic heat, polarization heat, reaction heat, and internal short-circuit heat. The calculation formula is as follows: Ohmic heating: Q ohm =I 2 R0Δt Polarization loss: Q p =I 2 R1Δt+I 2 R2Δt Reaction heat: Internal thermal resistance: Q ISC =I 2 R ISC Δt Ambient convection heat dissipation: Q exchange =hA(T b -T Ambient ) Total calorific value: Q total =Q ohm +Q p +Q r +Q ISC -Q exchange Where I is the current, R0 is the ohmic resistance, Δt is the time increment, R1 and R2 represent the polarization resistance of the battery, T b is the battery temperature, ΔS is the entropy coefficient, R ISC represents the short-circuit resistance, h is the heat transfer coefficient, A is the surface area, T Ambient is the ambient temperature, V ocv Indicates the open circuit voltage.

6. The method for constructing a real data set of multiple faults of lithium batteries based on a physical coupling model according to claim 1, characterized in that: The internal short-circuit fault current is calculated as follows: where R ISC represents the equivalent fault resistance, is the current flowing through the short-circuit path, U t is the terminal voltage of the lithium battery.

7. The method for constructing a real data set of multiple faults of lithium batteries based on a physical coupling model according to claim 1, characterized in that: The aging model uses a semi-empirical aging model to calculate capacity decay, which can effectively capture the two main aging acceleration factors: operating current (I) and battery temperature (T), and the relative capacity loss Q of the battery. loss for: Where B is the pre-exponential factor, E a is the activation energy of cyclic aging, α is the aging acceleration coefficient caused by current, I is the applied charging current, R is the ideal gas constant, Ah is the accumulated charge and discharge capacity, and z is the aging model parameter.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a real data set of multiple faults of lithium batteries based on a physical coupling model according to any one of claims 1 to 7 are implemented.

9. An electronic device, characterized in that: include: a memory for storing executable instructions; The processor is configured to implement the method for constructing a real lithium battery multi-fault data set based on a physical coupling model according to any one of claims 1 to 7 when executing the executable instructions stored in the memory.

Images