Lithium battery external short circuit fault rapid diagnosis system and method based on PSO-Bayes parameter identification and thermoelectric coupling model

Abstract

The invention discloses a lithium battery external short circuit fault rapid diagnosis system and method based on PSO-Bayes parameter identification and a thermoelectric coupling model. According to the lithium battery external short circuit fault diagnosis method, the accuracy and the stability of lithium battery external short circuit fault diagnosis are remarkably improved by fusing thermoelectric coupling modeling and a PSO-Bayes parameter identification method. A bidirectional coupling mechanism of an electrical and thermal model can simultaneously capture a voltage dynamic response and a temperature change process, and an intelligent parameter identification algorithm is matched, so that modeling errors under complex working conditions are effectively reduced, and fault feature extraction is more accurate. The fault scoring module integrates multi-dimensional indexes to carry out grading judgment, the capability of distinguishing mild, moderate and severe short circuits is enhanced, the problem of false alarm or missing alarm possibly caused by single parameter judgment is avoided, good response speed and environmental adaptability are shown in practical application, fault detection and grade evaluation can be completed within 4 seconds, and the fault detection efficiency is improved. And the real-time safety protection requirement is met.

Description

Technical Field

[0001] This invention belongs to the field of battery safety management technology, specifically a rapid diagnosis system and method for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model. Background Technology

[0002] In recent years, lithium-ion batteries have been widely used in various vehicles, portable devices, and large-scale energy storage systems due to their high energy density and long cycle life. However, batteries are prone to safety problems under extreme operating conditions such as external short circuits, mechanical damage, or overcharging. Among these, external short circuits are particularly typical, as they cause high-current discharge, leading to a rapid temperature rise, which in extreme cases can result in thermal runaway, fire, or even explosion.

[0003] Traditional methods for diagnosing external short circuits in lithium batteries primarily rely on a combination of voltage and current sensors and empirical rules for monitoring and judgment. Some studies employ RC equivalent modeling methods combined with recursive least squares (RLS) algorithms for modeling and identification. However, due to limitations such as the RLS algorithm's sensitivity to initial conditions and its convergence dependence on data quality, the identification accuracy and stability during dynamic nonlinear temperature rise processes cannot meet the demands of stringent application scenarios.

[0004] Furthermore, some existing methods employ temperature sensor-based fault detection mechanisms but fail to model and identify the electrochemical model state, lacking a quantitative characterization of the fault root cause and thus unable to support accurate severity classification. Therefore, there is an urgent need for an external short-circuit diagnostic method that integrates multi-dimensional sensing modeling, global optimization parameter identification, and fault grading mechanisms, possessing rapid, accurate, and stable identification capabilities and adapting to complex environmental changes. Summary of the Invention

[0005] The purpose of this invention is to provide a rapid diagnosis system and method for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model in order to solve the problems mentioned above.

[0006] The technical solution adopted in this invention is as follows: A rapid diagnosis system for external short circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model, comprising: an electrothermal modeling module, a thermal modeling module, a parameter identification module, a model coupling module, a fault scoring module, and an output module;

[0007] The electrothermal modeling module is used to construct a dynamic voltage response model of a lithium battery under external short-circuit conditions, taking into account electrode polarization, ohmic internal resistance, and capacitive hysteresis effect.

[0008] The thermal modeling module is used to establish a battery temperature rise prediction model, which incorporates parameters such as thermal resistance, thermal capacity, thermal conduction and convective heat transfer to reflect the thermal diffusion process.

[0009] The parameter identification module uses a joint mechanism of particle swarm optimization algorithm and Bayesian estimation to perform global optimization and posterior correction on multiple key parameters in the electrical and thermal models.

[0010] The model coupling module embeds the effect of voltage change on heat generation power and the feedback of heat accumulation on electrical parameters into the same system, thus constructing a thermoelectric bidirectional coupling mechanism.

[0011] The fault scoring module outputs a grading criterion for the severity of external short-circuit faults based on the residual and dynamic indicators between the model prediction results and the measured data.

[0012] The output module displays the predicted voltage, temperature, and fault level results in real time, and supports edge computing and remote platform deployment.

[0013] The voltage response output terminal of the electrothermal modeling module is connected to the electrical parameter input terminal of the model coupling module.

[0014] The temperature rise prediction output of the thermal modeling module is connected to the thermal parameter input of the model coupling module. The model coupling module uses a two-way coupling mechanism to transmit changes in electrical parameters to the heat generation power calculation end of the thermal modeling module on the one hand, and to feed back heat accumulation to the parameter correction end of the electrothermal modeling module on the other hand, thus forming a thermoelectric closed-loop regulation.

[0015] The optimized parameter output terminal of the parameter identification module is connected to the parameter update interface of the electrothermal modeling module and the thermal modeling module, respectively. Its parameter acquisition terminal obtains the current key parameter values ​​such as resistance, capacitance, thermal resistance, and thermal capacity from the two modules to achieve dynamic optimization and correction.

[0016] The voltage and temperature prediction results of the model coupling module are output in two ways. One way is connected to the prediction data input of the fault scoring module. After comparing with the measured data, the fault level is generated through the multidimensional residual index and then transmitted from the result output of the fault scoring module to the fault information display of the output module. The other way is directly connected to the real-time data display of the output module to display the predicted voltage and temperature.

[0017] In a preferred embodiment, the thermal modeling module employs an improved second-order RC equivalent circuit modeling method, which includes polarization resistors, capacitors, and on-state voltage regulators, and supports dynamic modeling of short-circuit sudden operating conditions.

[0018] The electrothermal modeling module employs an improved second-order RC equivalent circuit modeling method. It simulates the charge transfer hysteresis effect at the electrode interface using polarization resistance (Rp) and polarization capacitance (Cp), reflects electrolyte ion conduction losses using the series ohmic resistance (R0), and introduces a state-of-charge (SOC) regulator to dynamically correct the open-circuit voltage (OCV). This module features a refined segmented modeling strategy designed for sudden short-circuit conditions, accurately capturing the dynamic characteristics of the entire short-circuit process: the initial voltage drop (dominated by ohmic resistance), the polarization decay during the intermediate transition phase (capacitor charging and discharging), and the subsequent stable gradual decrease (electrochemical polarization equilibrium). It supports a voltage response resolution of 10ms.

[0019] In a preferred embodiment, the thermal modeling module introduces a temperature nonlinearity adjustment factor (α(T)) during the modeling process. This factor is constructed based on the Arrhenius formula, and its expression is:

[0020] α(T) = exp[-Ea / (k(T-T0))];

[0021] Where Ea is the activation energy, k is the Boltzmann constant, T is the real-time temperature, and T0 is the reference temperature; when the battery temperature exceeds 45℃, the adjustment factor triggers nonlinear decay, and the physical phenomenon of heat dissipation efficiency decreases at high temperature is reflected by correcting the convective heat transfer coefficient (h=h0·α(T)). At the same time, the thermal resistance network (including the internal thermal resistance Rth1 of the cell, the conductive thermal resistance Rth2 of the shell, and the environmental convective thermal resistance Rth3) and the three-dimensional heat capacity distribution (Cth=ρ·V·cp, where ρ is the material density, V is the volume, and cp is the specific heat capacity) are coupled to realize the prediction of the spatiotemporal distribution of the heat diffusion process.

[0022] In a preferred embodiment, the parameter identification module employs a "two-stage optimization mechanism": the particle swarm optimization (PSO) algorithm first performs global optimization in the multidimensional parameter space, initializing the particle swarm size to 50 and iterating 30 times, with the weighted sum of the root mean square error of voltage prediction (RMSE < 0.04V) and the error of temperature prediction (RMSE < 1.8℃) as the objective function; the initial parameter value interval obtained by optimization is then input into the Bayesian estimation module, which constructs a posterior probability model p(θ|ε)∝p(ε|θ)p(θ) based on the Gaussian prior distribution and the observation residual (ε = measured value - predicted value), and realizes dynamic updating of parameter distribution through the Markov chain Monte Carlo (MCMC) method, outputting parameter estimation results containing a 95% confidence interval, significantly improving the identification robustness under extreme conditions.

[0023] In a preferred embodiment, the model coupling module constructs a thermoelectric bidirectional feedback channel: on the one hand, the real-time current (I) output by the electrical module and the resistance (R) are fed together using Joule's law (Q = I). 2The heat generation power is calculated by Rt and input into the thermal module as a heat source term. The temperature field distribution calculated by the thermal module is reversed by material property parameters (such as conductivity σ(T)=σ0[1+β(T-T0)]) to correct the electrical parameters, forming a closed-loop coupling mechanism of "current-heat generation-temperature rise-resistance change-current adjustment".

[0024] In a preferred embodiment, the fault scoring module constructs a weighted scoring function based on the following five-dimensional indicators:

[0025] (1) Maximum temperature rise rate (dT / dt_max, weight 30%), reflecting the risk of thermal runaway;

[0026] (2) Voltage drop duration (t_drop, weight 25%), characterizing short-circuit current stability;

[0027] (3) Mean squared error of fit (RMSE, weight 20%), to evaluate the model's prediction accuracy;

[0028] (4) Temperature settling time (t_stab, weight 15%), which measures the rate of thermal diffusion equilibrium;

[0029] (5) Parameter change slope (dRp / dt, weight 10%) to monitor the deterioration trend of polarization characteristics.

[0030] The scoring function expression is:

[0031] SeverityScore=0.3·(dT / dt_max / 10)+0.25·(t_drop / 5)+0.2·(RMSE / 0.04)+0.15·(t_stab / 20)+0.1·(dRp / dt / 0.5);

[0032] All indicators have been normalized (0-100 points).

[0033] In a preferred embodiment, the output module supports a cross-platform deployment architecture: during the development phase, it can be integrated into the Simulink environment and co-simulated with MATLAB through the S-Function module; in engineering applications, it can be ported to embedded microcontroller platforms such as STM32H743 or TITMS320F28379D, with a hardware resource utilization rate of ≤60%. The communication interface supports a dual-channel redundant design, with a CAN bus (500kbps baud rate) for real-time interaction with the vehicle BMS, and a Modbus RTU protocol (RS485 interface) for connecting to the energy storage system monitoring platform. The data refresh cycle is ≤200ms, and it has a diagnostic fault code (DTC) output function (compliant with ISO15031 standard).

[0034] In a preferred embodiment, the open-circuit voltage is defined as a binary function OCV = f(SOC, Ta) of state of charge and ambient temperature, and online correction is achieved using a dual interpolation mechanism: at a reference temperature of 25°C, the base open-circuit voltage is obtained by looking up a table; when the ambient temperature deviates from the reference, the temperature compensation coefficient matrix is ​​called, and cubic spline interpolation is used to calculate the temperature correction ΔOCV(Ta), and the final output is OCV_cal = OCV_base + ΔOCV(Ta). For high-rate operating conditions, an additional neural network compensation layer is introduced to control the OCV prediction error within ±8mV.

[0035] In a preferred embodiment, the fault scoring module classifies the external short-circuit level into three levels: a minor fault corresponds to a short-circuit current <5C and a temperature rise rate <2℃ / s, triggering a current-limiting protection strategy; a moderate fault corresponds to a short-circuit current of 5–10C and a temperature rise rate of 2–5℃ / s, executing power-off protection; and a severe fault corresponds to a short-circuit current >10C and a temperature rise rate >5℃ / s, initiating cooling linkage and triggering an audible and visual alarm. The system supports a confidence-based grayscale output mechanism, calculating the uncertainty interval of the fault level through a Bayesian posterior probability distribution, providing a quantitative basis for safety decisions.

[0036] In a preferred embodiment, a rapid diagnosis method for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model includes the following steps:

[0037] (1) Collect voltage, current and temperature data of the battery at different working stages;

[0038] (2) Input the collected data into the thermo-electric coupling model for prediction;

[0039] (3) Use particle swarm optimization and Bayesian estimation methods to jointly update the model parameters;

[0040] (4) Calculate the residual between the predicted value and the measured value, and extract multidimensional fault features;

[0041] (5) Determine the fault level based on the scoring function and output the diagnostic results.

[0042] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0043] 1. This invention significantly improves the accuracy and stability of lithium battery external short-circuit fault diagnosis by integrating thermoelectric coupling modeling with the PSO-Bayes parameter identification method. The bidirectional coupling mechanism of the electrical and thermal models can simultaneously capture the dynamic voltage response and temperature change process. Combined with the intelligent parameter identification algorithm, it effectively reduces modeling errors under complex operating conditions, making fault feature extraction more accurate. The fault scoring module comprehensively uses multi-dimensional indicators for graded judgment, enhancing the ability to distinguish between mild, moderate, and severe short circuits, and avoiding false alarms or missed alarms that may be caused by judging with a single parameter.

[0044] 2. In practical applications, this invention demonstrates excellent response speed and environmental adaptability, completing fault detection and severity assessment within 4 seconds to meet real-time safety protection requirements. It is highly portable, adaptable to different types of lithium batteries and various application scenarios, ensuring stable operation in new energy vehicles, energy storage devices, and unmanned equipment. The output module supports embedded platform deployment and multiple communication protocols, facilitating integration with existing battery management systems. Through hierarchical control strategies, it promptly triggers current limiting, power-off, or cooling measures, effectively reducing the risk of thermal runaway caused by external short circuits and improving the safety and reliability of the battery during use. Attached Figure Description

[0045] Figure 1 This is a schematic diagram illustrating the process principle of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0047] Example:

[0048] Reference Figure 1 ,

[0049] A rapid diagnosis system for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model includes: an electrothermal modeling module, a thermal modeling module, a parameter identification module, a model coupling module, a fault scoring module, and an output module.

[0050] The electrothermal modeling module is used to construct a dynamic voltage response model of a lithium battery under external short-circuit conditions, taking into account electrode polarization, ohmic internal resistance, and capacitive hysteresis effect.

[0051] The thermal modeling module is used to establish a battery temperature rise prediction model, which incorporates parameters such as thermal resistance, thermal capacity, thermal conduction and convective heat transfer to reflect the thermal diffusion process.

[0052] The parameter identification module uses a joint mechanism of particle swarm optimization algorithm and Bayesian estimation to perform global optimization and posterior correction on multiple key parameters in the electrical and thermal models.

[0053] The model coupling module embeds the effect of voltage change on heat generation power and the feedback of heat accumulation on electrical parameters into the same system, thus constructing a thermoelectric bidirectional coupling mechanism.

[0054] The fault scoring module outputs a grading criterion for the severity of external short-circuit faults based on the residual and dynamic indicators between the model prediction results and the measured data.

[0055] The output module displays the predicted voltage, temperature, and fault level results in real time, and supports edge computing and remote platform deployment.

[0056] The voltage response output of the electrothermal modeling module is connected to the electrical parameter input of the model coupling module.

[0057] The temperature rise prediction output of the thermal modeling module is connected to the thermal parameter input of the model coupling module. The model coupling module uses a two-way coupling mechanism to transmit changes in electrical parameters to the heat generation power calculation end of the thermal modeling module on the one hand, and to feed back heat accumulation to the parameter correction end of the electrothermal modeling module on the other hand, thus forming a thermoelectric closed-loop regulation.

[0058] The optimized parameter output terminal of the parameter identification module is connected to the parameter update interface of the electrothermal modeling module and the thermal modeling module respectively. Its parameter acquisition terminal obtains the current key parameter values ​​such as resistance, capacitance, thermal resistance, and thermal capacity from the two modules to achieve dynamic optimization and correction.

[0059] The voltage and temperature prediction results of the model coupling module are output in two ways. One way is connected to the prediction data input of the fault scoring module. After comparing with the measured data, the fault level is generated through the multidimensional residual index and then transmitted from the result output of the fault scoring module to the fault information display of the output module. The other way is directly connected to the real-time data display of the output module to display the predicted voltage and temperature.

[0060] The thermal modeling module adopts an improved second-order RC equivalent circuit modeling method, which includes polarization resistors, capacitors and on-state voltage regulators, and supports dynamic modeling of short-circuit sudden operating conditions.

[0061] The electrothermal modeling module employs an improved second-order RC equivalent circuit modeling method. It simulates the charge transfer hysteresis effect at the electrode interface using polarization resistance (Rp) and polarization capacitance (Cp), reflects electrolyte ion conduction losses using the series ohmic resistance (R0), and introduces a state-of-charge (SOC) regulator to dynamically correct the open-circuit voltage (OCV). This module features a refined segmented modeling strategy designed for abrupt short-circuit conditions, accurately capturing the dynamic characteristics of the entire short-circuit process: the initial voltage drop (dominated by ohmic resistance), the polarization decay during the intermediate transition phase (capacitor charging and discharging), and the subsequent stable gradual decrease (electrochemical polarization equilibrium). It supports a voltage response resolution of 10ms.

[0062] The thermal modeling module introduces a temperature nonlinearity adjustment factor (α(T)) during the modeling process. This factor is constructed based on the Arrhenius formula, and its expression is:

[0063] α(T) = exp[-Ea / (k(T-T0))];

[0064] Where Ea is the activation energy, k is the Boltzmann constant, T is the real-time temperature, and T0 is the reference temperature; when the battery temperature exceeds 45℃, the adjustment factor triggers nonlinear decay, and the physical phenomenon of heat dissipation efficiency decreases at high temperature is reflected by correcting the convective heat transfer coefficient (h=h0·α(T)). At the same time, the thermal resistance network (including the internal thermal resistance Rth1 of the cell, the conductive thermal resistance Rth2 of the shell, and the environmental convective thermal resistance Rth3) and the three-dimensional heat capacity distribution (Cth=ρ·V·cp, where ρ is the material density, V is the volume, and cp is the specific heat capacity) are coupled to realize the prediction of the spatiotemporal distribution of the heat diffusion process.

[0065] The parameter identification module adopts a "two-stage optimization mechanism": the particle swarm optimization (PSO) algorithm first performs global optimization in the multi-dimensional parameter space, initializes the particle swarm size to 50, and iterates 30 times, with the weighted sum of the root mean square error of voltage prediction (RMSE < 0.04V) and the error of temperature prediction (RMSE < 1.8℃) as the objective function; the initial parameter value interval obtained by optimization is then input into the Bayesian estimation module, which constructs a posterior probability model p(θ|ε)∝p(ε|θ)p(θ) based on the Gaussian prior distribution and the observation residual (ε = measured value - predicted value). The parameter distribution is dynamically updated through the Markov chain Monte Carlo (MCMC) method, and the output parameter estimation results containing a 95% confidence interval are significantly improved, which significantly improves the identification robustness under extreme conditions.

[0066] The model coupling module constructs a thermoelectric bidirectional feedback channel: on the one hand, the real-time current (I) output by the electrical module is fed back to the resistance (R) through Joule's law (Q = I). 2The heat generation power is calculated by Rt and input into the thermal module as a heat source term. The temperature field distribution calculated by the thermal module is reversed by material property parameters (such as conductivity σ(T)=σ0[1+β(T-T0)]) to correct the electrical parameters, forming a closed-loop coupling mechanism of "current-heat generation-temperature rise-resistance change-current adjustment".

[0067] To address the voltage prediction error (ΔU = U_pred - U_meas), the system dynamically corrects the estimated heat source intensity using a proportional-integral (PI) controller with correction coefficients Kp = 0.8 and Ki = 0.2, ensuring that the thermoelectric coupling error is ≤3%.

[0068] The fault scoring module constructs a weighted scoring function based on the following five dimensions:

[0069] (1) Maximum temperature rise rate (dT / dt_max, weight 30%), reflecting the risk of thermal runaway;

[0070] (2) Voltage drop duration (t_drop, weight 25%), characterizing short-circuit current stability;

[0071] (3) Mean squared error of fit (RMSE, weight 20%), to evaluate the model's prediction accuracy;

[0072] (4) Temperature settling time (t_stab, weight 15%), which measures the rate of thermal diffusion equilibrium;

[0073] (5) Parameter change slope (dRp / dt, weight 10%) to monitor the deterioration trend of polarization characteristics.

[0074] The scoring function expression is:

[0075] SeverityScore=0.3·(dT / dt_max / 10)+0.25·(t_drop / 5)+0.2·(RMSE / 0.04)+0.15·(t_stab / 20)+0.1·(dRp / dt / 0.5);

[0076] All indicators have been normalized (0-100 points).

[0077] The output module supports cross-platform deployment architecture: during the development phase, it can be integrated into the Simulink environment and co-simulated with MATLAB through the S-Function module; in engineering applications, it can be ported to embedded microcontroller platforms such as STM32H743 or TITMS320F28379D, with hardware resource utilization ≤60%. The communication interface supports dual-channel redundancy design, with CAN bus (500kbps baud rate) for real-time interaction with the vehicle BMS, and Modbus RTU protocol (RS485 interface) for connecting to the energy storage system monitoring platform, with a data refresh cycle ≤200ms, and has a diagnostic fault code (DTC) output function (compliant with ISO15031 standard).

[0078] The open-circuit voltage is defined as a binary function of state of charge and ambient temperature, OCV = f(SOC, Ta), and online correction is achieved using a dual interpolation mechanism: at a reference temperature of 25℃, the base open-circuit voltage is obtained by looking up a table; when the ambient temperature deviates from the reference, the temperature compensation coefficient matrix is ​​called, and cubic spline interpolation is used to calculate the temperature correction ΔOCV(Ta), ultimately outputting OCV_cal = OCV_base + ΔOCV(Ta). For high-rate operating conditions, an additional neural network compensation layer is introduced to control the OCV prediction error within ±8mV.

[0079] The fault scoring module classifies external short-circuit levels into three levels: minor faults correspond to short-circuit current <5°C and temperature rise rate <2°C / s, triggering current-limiting protection; moderate faults correspond to short-circuit current 5–10°C and temperature rise rate 2–5°C / s, executing power-off protection; severe faults correspond to short-circuit current >10°C and temperature rise rate >5°C / s, initiating cooling linkage and triggering audible and visual alarms. The system supports a confidence-based grayscale output mechanism, calculating the uncertainty interval of the fault level through a Bayesian posterior probability distribution, providing a quantitative basis for safety decisions.

[0080] A rapid diagnosis method for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model includes the following steps:

[0081] (1) Collect voltage, current and temperature data of the battery at different working stages;

[0082] (2) Input the collected data into the thermo-electric coupling model for prediction;

[0083] (3) Use particle swarm optimization and Bayesian estimation methods to jointly update the model parameters;

[0084] (4) Calculate the residual between the predicted value and the measured value, and extract multidimensional fault features;

[0085] (5) Determine the fault level based on the scoring function and output the diagnostic results.

[0086] Example 1:

[0087] Experimental Platform Setup and Data Acquisition: This embodiment uses lithium iron phosphate and ternary lithium 18650 cells as research objects to construct a standard test platform. The platform includes: a programmable power supply, electronic load, environmental temperature control chamber, CAN interface recorder, high-speed data acquisition card, industrial-grade embedded PC, and multi-channel voltage, current, and thermocouple temperature acquisition modules. The system is coordinated and controlled through a LabVIEW host computer program and supports automatic recording and export of test data.

[0088] The collected parameters include multi-dimensional information such as cell terminal voltage, discharge current, casing temperature, ambient temperature, and energy release per unit time (obtained through voltage-current integration). The sampling frequency is set to 60Hz, and at least 10 seconds of complete data are recorded before and after each short-circuit cycle.

[0089] Short-circuit triggering was simulated using a high-power relay. The short-circuit duration was set to four levels: 5s, 10s, 15s, and 20s. Each experiment was repeated three times to improve statistical reliability. During the experiment, the changes in temperature rise onset time, peak temperature rise, temperature rise rate, and voltage drop pattern were closely monitored.

[0090] Experiments revealed that the longer the short circuit duration, the near-linear increase in temperature rise rate, especially accelerating significantly above 15 seconds. In external short circuit conditions exceeding 10 seconds, some cells exhibited casing deformation and localized melting of the tabs, suggesting that this method has the potential to differentiate between different fault levels.

[0091] Example 2:

[0092] Thermoelectric Coupling Model Construction: This embodiment introduces a thermoelectric coupling mechanism based on the traditional equivalent circuit model to achieve joint modeling of voltage response and temperature change during the external short circuit process of a lithium battery. Specifically, the model is divided into two sub-modules: electrical and thermal. The electrical part describes the voltage response, while the thermal part is used to predict the temperature rise process, realizing the linkage and feedback between multiple physics fields.

[0093] The electrical section focuses on characterizing the dynamic decay process of the cell's terminal voltage under external short-circuit conditions, considering factors such as polarization resistance and capacitor charging / discharging hysteresis. Through a refined segmented modeling strategy, it accurately reflects the typical voltage change trends of the initial instantaneous drop, the mid-term transition phase, and the subsequent stable gradual decrease. The thermal section introduces multiple structural parameters such as thermal capacity and thermal resistance to simulate the generation, accumulation, and conduction of heat energy within the cell. The model considers convective heat dissipation between the casing and the environment, material non-uniformity along the heat conduction path, and variations in cooling efficiency under different operating conditions. Simultaneously, by setting up a temperature feedback channel, a two-way thermal-electric coupling mechanism is constructed, allowing voltage changes to modulate heat generation, and heat accumulation to in turn affect the material's electrical properties, forming a closed-loop dynamic system.

[0094] The model parameters encompass multiple resistances, capacitances, thermal constants, and coupling coefficients, covering physical quantities such as material thermal conductivity, specific heat capacity, convective heat transfer coefficient, and dynamic conductivity rate of change. To achieve accurate parameter identification and dynamic updates, this embodiment introduces a parameter identification mechanism combining particle swarm optimization (PSO) and Bayesian estimation methods. First, PSO is used for global search to quickly obtain the initial value range and optimal solution region of each parameter in the thermo-electric coupling model. Subsequently, Bayesian estimation methods are used to perform distribution modeling and gradual updates of each parameter, giving the parameter estimation results higher robustness and confidence interval representation capabilities.

[0095] Specifically, the PSO module initializes the distribution of all parameters to be estimated in a multidimensional space. By setting an objective function (such as a weighted sum of temperature and voltage prediction errors), it continuously updates the particle positions to find the global optimum. Meanwhile, the Bayes module performs posterior correction on the particle swarm search results based on measurement errors and prior parameter distributions, ensuring that parameter changes better reflect the dynamic characteristics of the actual system. Combining this dual optimization strategy, the model can still achieve high-precision prediction capabilities and exhibit good generalization performance under conditions of strong dynamic changes in external short circuits.

[0096] During the system identification process, based on the known input current, ambient temperature, and historical status, the predicted results of the cell terminal voltage and casing temperature are output in real time. To improve prediction accuracy, this embodiment introduces a temperature nonlinear adjustment coefficient to address the problem of significantly reduced heat dissipation efficiency in high-temperature ranges. Simultaneously, considering the characteristic that open-circuit voltage changes with both state of charge (SOC) and temperature, empirical mapping relationships are introduced or dynamic updates are performed through interpolation modeling to ensure a high consistency between model predictions and measured values. Furthermore, to verify the model's adaptability and stability, this paper builds a co-simulation environment based on the Simulink platform and conducts multiple sets of model effectiveness verification experiments, including operating condition switching tests, long-term operation tests, and boundary parameter perturbation tests. Under different initial SOC levels, temperature backgrounds, and aging conditions, the model can stably output reasonable results, with root mean square errors controlled within 0.04V for voltage and 1.8℃ for temperature, demonstrating good versatility and accuracy.

[0097] This model not only establishes a joint voltage-temperature prediction capability but also provides essential dynamic indicators for subsequent fault scoring and classification algorithms, such as key intermediate quantities like maximum temperature rise rate, peak heat accumulation time, and voltage drop duration, enabling multi-dimensional diagnostic feature extraction. The diagnostic system based on this model possesses advantages such as sensitive response, stable modeling, and low computational complexity, demonstrating good portability, real-time performance, and promising engineering applications.

[0098] From the above, we can conclude that:

[0099] This invention significantly improves the accuracy and stability of lithium battery external short-circuit fault diagnosis by integrating thermoelectric coupling modeling with the PSO-Bayes parameter identification method. The bidirectional coupling mechanism of the electrical and thermal models can simultaneously capture the dynamic voltage response and temperature change process. Combined with the intelligent parameter identification algorithm, this effectively reduces modeling errors under complex operating conditions, making fault feature extraction more accurate. The fault scoring module comprehensively assesses faults based on multi-dimensional indicators, enhancing the ability to distinguish between mild, moderate, and severe short circuits and avoiding false alarms or missed alarms that may result from relying on a single parameter.

[0100] In practical applications, this invention demonstrates excellent response speed and environmental adaptability, completing fault detection and severity assessment within 4 seconds to meet real-time safety protection requirements. It exhibits strong portability, adapting to different types of lithium batteries and various application scenarios, ensuring stable operation in new energy vehicles, energy storage devices, and unmanned equipment. The output module supports embedded platform deployment and multiple communication protocols, facilitating integration with existing battery management systems. Through hierarchical control strategies, it promptly triggers current limiting, power-off, or cooling measures, effectively reducing the risk of thermal runaway caused by external short circuits and improving the safety and reliability of the battery during use.

[0101] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is 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 limitations, 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 the element.

[0102] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid diagnostic system for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model, characterized in that: include: The module includes an electrothermal modeling module, a thermal modeling module, a parameter identification module, a model coupling module, a fault scoring module, and an output module. The electrothermal modeling module is used to construct a dynamic voltage response model of a lithium battery under external short-circuit conditions, taking into account electrode polarization, ohmic internal resistance, and capacitive hysteresis effect. The thermal modeling module is used to establish a battery temperature rise prediction model, which incorporates thermal resistance, thermal capacity, thermal conduction and convective heat transfer parameters to reflect the thermal diffusion process. The parameter identification module uses a joint mechanism of particle swarm optimization algorithm and Bayesian estimation to perform global optimization and posterior correction on multiple key parameters in the electrical and thermal models. The model coupling module embeds the effect of voltage change on heat generation power and the feedback of heat accumulation on electrical parameters into the same system, thus constructing a thermoelectric bidirectional coupling mechanism. The fault scoring module outputs a grading criterion for the severity of external short-circuit faults based on the residual and dynamic indicators between the model prediction results and the measured data. The output module displays the predicted voltage, temperature, and fault level results in real time, and supports edge computing and remote platform deployment. The voltage response output terminal of the electrothermal modeling module is connected to the electrical parameter input terminal of the model coupling module. The temperature rise prediction output of the thermal modeling module is connected to the thermal parameter input of the model coupling module. The model coupling module transmits the changes in electrical parameters to the heat generation power calculation end of the thermal modeling module through a two-way coupling mechanism, and feeds back the heat accumulation to the parameter correction end of the electrothermal modeling module, forming a thermoelectric closed-loop regulation. The optimized parameter output terminal of the parameter identification module is connected to the parameter update interface of the electrothermal modeling module and the thermal modeling module respectively, while its parameter acquisition terminal obtains the current key parameter values ​​of resistance, capacitance, thermal resistance and thermal capacity from the two modules to realize dynamic optimization and correction. The voltage and temperature prediction results of the model coupling module are output in two ways. One way is connected to the prediction data input of the fault scoring module. After comparing with the measured data, the fault level is generated through the multidimensional residual index and then transmitted from the result output of the fault scoring module to the fault information display of the output module. The other way is directly connected to the real-time data display of the output module to display the predicted voltage and temperature.

2. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model as described in claim 1, characterized in that: The thermal modeling module adopts an improved second-order RC equivalent circuit modeling method, which includes polarization resistors, capacitors and on-state voltage regulators, and supports dynamic modeling of short-circuit sudden operating conditions. The electrothermal modeling module adopts an improved second-order RC equivalent circuit modeling method. It simulates the charge transfer hysteresis effect at the electrode interface through polarization resistance and polarization capacitance, reflects the electrolyte ion conduction loss through series ohmic internal resistance, and introduces a state voltage regulator to realize dynamic correction of the open-circuit voltage by the state of charge. This module designs a refined segmented modeling strategy for short-circuit sudden change conditions, which can accurately capture the dynamic characteristics of the entire stage, including the instantaneous voltage drop at the beginning of the short circuit, the polarization decay in the middle transition stage, and the stable gradual drop in the later stage, and supports a voltage response resolution of 10ms.

3. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The thermal modeling module introduces a temperature nonlinearity adjustment factor (α(T)) during the modeling process. This factor is constructed based on the Arrhenius formula, and its expression is: α(T) = exp[-Ea / (k(T-T0))]; Where Ea is the activation energy, k is the Boltzmann constant, T is the real-time temperature, and T0 is the reference temperature; when the battery temperature exceeds 45℃, the adjustment factor triggers nonlinear decay, which reflects the physical phenomenon of reduced heat dissipation efficiency at high temperature by correcting the convective heat transfer coefficient, and at the same time couples the thermal resistance network with the three-dimensional heat capacity distribution to realize the prediction of the spatiotemporal distribution of the heat diffusion process.

4. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The parameter identification module adopts a "two-stage optimization mechanism": the particle swarm optimization algorithm first performs global optimization in the multi-dimensional parameter space, initializes the particle swarm size to 50, and iterates 30 times, with the weighted sum of the root mean square error of voltage prediction and the error of temperature prediction as the objective function; the initial parameter value interval obtained by optimization is then input into the Bayesian estimation module, which constructs a posterior probability model p(θ|ε)∝p(ε|θ)p(θ) based on the Gaussian prior distribution and observation residuals, and realizes dynamic updating of parameter distribution through the Markov chain Monte Carlo method, outputting parameter estimation results containing a 95% confidence interval, which significantly improves the identification robustness under extreme conditions.

5. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The model coupling module constructs a thermoelectric bidirectional feedback channel: on the one hand, it connects the real-time current (I) output by the electrical module with the resistance (R) using Joule's law (Q = I). 2 The heat generation power is calculated by Rt and input as a heat source term into the thermal module; the temperature field distribution calculated by the thermal module is reversed by material property parameters (such as electrical conductivity σ(T)=σ0[1+β(T-T0)]) to correct the electrical parameters, forming a closed-loop coupling mechanism of "current-heat generation-temperature rise-resistance change-current adjustment".

6. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The fault scoring module constructs a weighted scoring function based on the following five-dimensional indicators: (1) Maximum temperature rise rate (dT / dt_max, weight 30%), reflecting the risk of thermal runaway; (2) Voltage drop duration (t_drop, weight 25%), characterizing short-circuit current stability; (3) Mean squared error of fit (RMSE, weight 20%), to evaluate the model's prediction accuracy; (4) Temperature settling time (t_stab, weight 15%), which measures the rate of thermal diffusion equilibrium; (5) Parameter change slope (dRp / dt, weight 10%), to monitor the deterioration trend of polarization characteristics; The scoring function expression is: SeverityScore=0.3·(dT / dt_max / 10)+0.25·(t_drop / 5)+0.2·(RM SE / 0.04)+0.15·(t_stab / 20)+0.1·(dRp / dt / 0.5); All indicators have been normalized (0-100 points).

7. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The output module supports a cross-platform deployment architecture: during the development phase, it can be integrated into the Simulink environment and co-simulated with MATLAB through the S-Function module; in engineering applications, it can be ported to the STM32H743 or TITMS320F28379D embedded microcontroller platform with a hardware resource utilization rate of ≤60%; the communication interface supports a dual-channel redundant design, with a CAN bus (500kbps baud rate) for real-time interaction with the vehicle BMS, and a Modbus RTU protocol (RS485 interface) for connecting to the energy storage system monitoring platform, with a data refresh cycle of ≤200ms and a diagnostic fault code (DTC) output function.

8. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The open-circuit voltage is defined as a binary function of state of charge and ambient temperature, OCV = f(SOC, Ta), and online correction is achieved using a dual interpolation mechanism: at a reference temperature of 25℃, the base open-circuit voltage is obtained by looking up a table; when the ambient temperature deviates from the reference, the temperature compensation coefficient matrix is ​​called, and cubic spline interpolation is used to calculate the temperature correction ΔOCV(Ta), and the final output is OCV_cal = OCV_base + ΔOCV(Ta); for high-rate operating conditions, an additional neural network compensation layer is introduced to control the OCV prediction error within ±8mV.

9. The rapid diagnosis system for external short-circuit faults of lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model according to claim 1, characterized in that, The fault scoring module classifies the external short circuit level into three levels: minor faults correspond to short circuit current <5C and temperature rise rate <2℃ / s, triggering current limiting protection strategy; A moderate fault corresponds to a short-circuit current of 5–10°C and a temperature rise rate of 2–5°C / s, triggering power-off protection; a severe fault corresponds to a short-circuit current >10°C and a temperature rise rate >5°C / s, triggering cooling linkage and an audible and visual alarm; the system supports a confidence-based grayscale output mechanism, calculating the uncertainty range of the fault level through a Bayesian posterior probability distribution, providing a quantitative basis for safety decisions.

10. A rapid diagnosis method for external short-circuit faults in lithium batteries based on PSO-Bayes parameter identification and thermoelectric coupling model, characterized in that: The method, based on any one of claims 1 to 9, includes the following steps: (1) Collect voltage, current and temperature data of the battery at different working stages; (2) Input the collected data into the thermo-electric coupling model for prediction; (3) Use particle swarm optimization and Bayesian estimation methods to jointly update the model parameters; (4) Calculate the residual between the predicted value and the measured value, and extract multidimensional fault features; (5) Determine the fault level based on the scoring function and output the diagnostic results.

Images