Evaluation method and evaluation system for high-rate thermal characteristics of power battery

Abstract

The embodiment of the invention provides a method and a system for evaluating high-rate thermal characteristics of a power battery. The evaluation method comprises the following steps: step 1, carrying out a low-rate charge-discharge experiment on a single battery to obtain experiment data; 2, based on experimental data verification, adjusting the NTGK model to obtain a target NTGK model; 3, obtaining a Gaussian process regression optimization model based on the Gaussian process regression model and the experimental data; 4, acquiring a voltage-current density curve according to the related data and the Gaussian process regression optimization model, and fitting to obtain a parameter U and a parameter Y; 5, establishing a battery pack model, importing the parameter U and the parameter Y into the target NTGK model, and outputting a thermal characteristic prediction result of the battery pack model; and step 6, performing verification and correction according to a thermal characteristic prediction result, and forming a Gaussian process regression optimization model adaptive correction closed loop. According to the embodiment of the invention, the high-magnification experiment dependence is reduced, the physical support is ensured, and the high-magnification thermal characteristic evaluation precision is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power battery thermal management technology, and in particular to a method and system for evaluating the high-rate thermal characteristics of power batteries. Background Technology

[0002] With the increasing demand for fast charging performance of power batteries from new energy vehicles and energy storage systems, high-rate charging and discharging has become a core application scenario for power batteries. However, under high-rate operating conditions, the rate of heat generation inside the battery increases sharply, which can easily lead to safety risks such as temperature rise runaway and thermal runaway. Therefore, accurate assessment of the high-rate thermal characteristics of power batteries has become an important requirement for the industry.

[0003] Existing technologies for predicting the high-rate thermal characteristics of power batteries have several problems: On the one hand, while mainstream electrochemical-thermal coupling models (such as the NTGK model) can describe the internal heat generation process of batteries based on physical mechanisms, their core parameters need to be calibrated through a large number of high-rate charge-discharge experiments. This not only poses safety risks such as electrolyte decomposition and cell bulging during the experiment, but also requires repeated investment in cell and equipment costs. Furthermore, nonlinear thermal effects at high rates can easily lead to parameter calibration deviations, ultimately limiting the accuracy of model predictions. On the other hand, while some data-driven algorithms can assist in prediction by learning patterns from data, they either focus on health indicators such as capacity and internal resistance changes without involving high-rate thermal behavior modeling, or only focus on static performance such as cycle life, failing to reflect transient thermal characteristics. There are also schemes that establish temperature field prediction models based on experimental data, but these still rely on a large amount of high-rate experimental data and lack the ability to predict across rates from low to high.

[0004] Therefore, there is an urgent need for a high-rate thermal characteristic evaluation method that can reduce reliance on high-rate experiments, improve the accuracy of high-rate predictions, and take into account physical interpretability. This method should utilize low-rate experimental data to achieve cross-rate mapping prediction of high-rate thermal characteristics, thereby solving the problems of high safety and cost risks, insufficient prediction accuracy, lack of physical support, and poor adaptability in existing technologies. This would reduce costs while improving the prediction accuracy and safety of high-rate thermal characteristics of power batteries. Summary of the Invention

[0005] Therefore, in order to overcome at least some of the defects and deficiencies in the prior art, embodiments of the present invention provide a method for evaluating the high-rate thermal characteristics of a power battery and a system for evaluating the high-rate thermal characteristics of a power battery.

[0006] Specifically, on one hand, this invention provides a method for evaluating the high-rate thermal characteristics of a power battery, comprising: Step 1: conducting low-rate charge-discharge experiments on individual battery cells to obtain experimental data, the experimental data including voltage data, depth of discharge data, and temperature rise data, wherein the rate range of the low-rate charge-discharge experiment is 0.3C-3C; Step 2: verifying and adjusting the NTGK model based on the voltage data and the depth of discharge data to obtain a target NTGK model; Step 3: obtaining a Gaussian process regression optimization model based on a Gaussian process regression model, the depth of discharge data, and the charge-discharge rate data corresponding to the depth of discharge data; Step 4: obtaining a voltage-current density curve based on the voltage data, the depth of discharge data, and the Gaussian process regression optimization model, and extracting feature parameters from the voltage-current density curve to fit parameters U and Y, wherein parameter U is the current density of the voltage-current density curve. The voltage intercept when the degree is 0, the parameter Y is the reciprocal of the slope of the voltage-current density curve, wherein the high rate range is greater than or equal to 5C; Step 5: Establish a battery pack model based on the battery cell, import the parameter U and the parameter Y into the target NTGK model according to the battery pack model and establish the energy equation, perform meshing on the battery pack model and set boundary conditions, and use the SIMPLE algorithm to output the thermal characteristic prediction result of the battery pack model; Step 6: Verify and correct according to the thermal characteristic prediction result. If the verification result meets the preset error requirement, output the final thermal characteristic evaluation result; if the verification result does not meet the preset error requirement, adjust the Gaussian process regression optimization model and the fitting coefficients corresponding to the parameter U and the parameter Y through error feedback, and re-execute the simulation calculation of steps 3 to 5 to form the adaptive correction closed loop of the Gaussian process regression optimization model.

[0007] In a specific embodiment of the present invention, step 3: obtaining a Gaussian process regression optimization model based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data, includes: outputting predicted voltage data based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data, and constructing a training dataset D = {(x1,y1),...,(x n ,y n )}, where x is the input feature and y is the output label (n≥500); the exponential covariance function is selected, and the logarithmic boundary likelihood is maximized by the conjugate gradient method hyperparameter optimization method to train the Gaussian process regression model and obtain the Gaussian process regression optimization model.

[0008] In a specific embodiment of the present invention, the formula for the exponential covariance function is: in, Let l be the signal variance and l be the length scale. x1 represents the noise variance, x2 represents the rate data of the low-rate charge-discharge test, and x2 represents the high-rate data to be tested.

[0009] In a specific embodiment of the present invention, the logarithmic boundary likelihood formula is:

[0010]

[0011] Where y is the output label vector, X is the input feature matrix, θ is the hyperparameter of the Gaussian process regression model, C is the covariance matrix, det(C) is the determinant of the covariance matrix C, and n is the number of samples.

[0012] In a specific embodiment of the present invention, step 4, obtaining the voltage-current density curve based on the voltage data, the discharge depth data, and the Gaussian process regression optimization model, includes: obtaining voltage prediction data and discharge depth prediction data at high rates based on the Gaussian process regression model, the discharge depth data, and the charge-discharge rate data corresponding to the discharge depth data; obtaining the voltage-discharge depth prediction curve based on the voltage prediction data and the discharge depth prediction data; and converting the voltage-discharge depth prediction curve into a voltage-current density curve.

[0013] In one specific embodiment of the present invention, the parameter U and the parameter Y are associated with the discharge depth prediction data through a fifth-order polynomial; wherein, the formula of the fifth-order polynomial is:

[0014] U(DOD) = 3.36 - 0.86DOD + 5.35DOD 2 -15.36DOD 3 +19.04DOD 4 -8.46DOD 5

[0015] Y(DOD)=807.44-6209.15DOD+31990.13DOD 2 -68158.6DOD 3

[0016] -63608.28DOD 4 -22097.2DOD 5 Where DOD stands for Depth of Discharge data.

[0017] In a specific embodiment of the present invention, step 2: verifying and adjusting the NTGK model based on the voltage data and the depth of discharge data to obtain the target NTGK model includes: outputting simulation results through the NTGK model based on the voltage data and the depth of discharge data; verifying the temperature rise data through the simulation results; if the verification results meet the preset error requirements, the accuracy of the NTGK model is confirmed; if the verification results do not meet the preset error requirements, the parameters of the NTGK model are adjusted and step 2 simulation and verification are re-executed until the error meets the preset requirements.

[0018] On the other hand, embodiments of the present invention also provide an evaluation system for the high-rate thermal characteristics of a power battery, comprising: a processor and a memory connected to the processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the evaluation method for the high-rate thermal characteristics of a power battery as described above.

[0019] In one specific embodiment of the present invention, the system further includes a network connection module for communicating with the battery management system to perform real-time temperature monitoring and safety warnings.

[0020] As can be seen from the above, the method for evaluating the high-rate thermal characteristics of power batteries provided in this invention obtains basic data on voltage, depth of discharge, and temperature rise by conducting low-rate (0.3C-3C) charge-discharge experiments, thus avoiding the safety risks and cost of high-rate experiments. Furthermore, this invention verifies the low-rate reliability of the NTGK model based on experimental data, retains the physical interpretability of the model to address the lack of mechanism in purely data-driven algorithms, and then learns the low-rate data patterns through the GPR model (i.e., Gaussian process regression model) to achieve high-rate voltage prediction and extraction of U and parameter Y, thus solving the problems of traditional NTGK... To address the issues of insufficient high-rate accuracy and lack of cross-rate prediction capability in the model, the NTGK model was ultimately used to construct a battery pack simulation model by importing U and parameter Y. The thermal characteristic data was then output to adapt to the complex structure of the battery pack. After outputting the thermal characteristic prediction results, an adaptive closed loop of "simulation output - result verification - error feedback - parameter optimization - re-simulation" was further constructed through verification and correction. This process adjusted and corrected the deviations between the model and parameters until the thermal characteristic prediction error met the standards for practical applications. As a result, the reliance on high-rate experiments was reduced while ensuring physical support, and the accuracy of high-rate thermal characteristic evaluation was significantly improved. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a method for evaluating the high-rate thermal characteristics of a power battery, as provided in the first embodiment of the present invention.

[0023] Figure 2 for Figure 1 A flowchart illustrating step 2.

[0024] Figure 3 for Figure 1 A flowchart illustrating step 3.

[0025] Figure 4 for Figure 1 A flowchart illustrating step 4.

[0026] Figure 5 This is a schematic diagram showing the cross-sectional locations of battery packs A and B and the battery numbers.

[0027] Figure 6 This is a temperature contour map of section A of the battery pack at a 5C rate.

[0028] Figure 7 This is a comparison chart of the U-DOD curve predicted by the GPR model at a 5C ratio and the experimental curve.

[0029] Figure 8 This is a schematic diagram of the structure of an evaluation system for the high-rate thermal characteristics of a power battery, provided in the second embodiment of the present invention.

[0030] Figure 9 This is a schematic diagram of the structure of a storage medium provided in the third embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments described in the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0032] In the embodiments of this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0033] [First Embodiment]

[0034] See Figure 1The first embodiment of the present invention provides a method for evaluating the high-rate thermal characteristics of a power battery, which may specifically include, for example, the following steps:

[0035] S10, Step 1: Conduct a low-rate charge-discharge experiment on a single battery cell to obtain experimental data, including voltage data, depth of discharge data, and temperature rise data, wherein the low-rate charge-discharge experiment ranges from 0.3C to 3C.

[0036] S20, Step 2: Verify and adjust the NTGK model based on the voltage data and the depth of discharge data to obtain the target NTGK model;

[0037] S30, Step 3: Obtain the Gaussian process regression optimization model based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data;

[0038] S40, Step 4: Based on the voltage data, the discharge depth data, and the Gaussian process regression optimization model, obtain the voltage-current density curve, and extract feature parameters from the voltage-current density curve to fit parameters U and Y, wherein parameter U is the voltage intercept of the voltage-current density curve when the current density is 0, and parameter Y is the reciprocal of the slope of the voltage-current density curve, wherein the high-rate range is greater than or equal to 5C;

[0039] S50, Step 5: Establish a battery pack model based on the battery cell, import the parameters U and Y into the target NTGK model according to the battery pack model and establish an energy equation, perform mesh generation on the battery pack model and set boundary conditions, and use the SIMPLE algorithm to output the thermal characteristic prediction results of the battery pack model;

[0040] S60, Step 6: Verify and correct the predicted thermal characteristics. If the verification result meets the preset error requirement, output the final thermal characteristic evaluation result. If the verification result does not meet the preset error requirement, adjust the Gaussian process regression optimization model and the fitting coefficients corresponding to the parameters U and Y through error feedback, and re-execute the simulation calculations of steps 3 to 5 to form an adaptive correction closed loop for the Gaussian process regression optimization model. The preset error requirement is, for example, that the temperature prediction error is no higher than 2℃, and the heat generation rate or heat flux density prediction error is no higher than 5%.

[0041] Specifically, convective heat transfer boundary conditions are used to describe the heat exchange between the battery surface and the environment or cooling medium. The expression for convective heat transfer boundary conditions is:

[0042]

[0043] Where h is the convective heat transfer coefficient, T ∞ Let n be the ambient temperature and n be the normal vector of the outer surface.

[0044] The SIMPLE algorithm is a pressure-velocity coupled solution algorithm used to solve the interaction between the energy equation, momentum equation, and continuity equation. In FLUENT software, the SIMPLE algorithm can guarantee computational stability and accelerate convergence.

[0045] This embodiment obtains experimental data such as voltage, depth of discharge, and temperature rise by conducting low-rate (0.3C-3C) charge-discharge experiments. The experimental data is then used for NTGK model verification and Gaussian process regression (GPR) model training to obtain parameters U and Y at high rates. U and parameters Y are then imported into the NTGK model to construct a battery pack simulation model, and thermal simulation is performed on high-rate (≥5C) operating conditions to achieve safe and accurate evaluation of high-rate thermal characteristics.

[0046] Specifically, the low-rate charge-discharge experiment in step 1 is conducted in a constant temperature environment to eliminate the interference of initial temperature fluctuations on subsequent data acquisition.

[0047] In step 1, this embodiment of the invention obtains basic experimental data by conducting low-rate charge-discharge experiments. The experimental data includes, for example, voltage data, depth of discharge data, and temperature rise data. Specifically, a single power battery cell to be evaluated (e.g., an LF50K lithium iron phosphate battery cell) can be selected and tested using a charge-discharge testing system (e.g., a BTS20-5V / 4×200A) and a temperature control device (e.g., a HYF-TH-150DH constant temperature chamber). Then, the battery cell is pre-treated by placing it in a constant temperature chamber and allowing it to stand for a relatively long time in a suitable temperature environment to eliminate the interference of initial temperature fluctuations on the data. The battery is then charged to ensure it is fully charged. After charging, it is kept in a constant temperature environment for a period of time. In the process of charging, local heat accumulated during charging is eliminated. Then, discharge and data acquisition are performed. The discharge is carried out at a low rate (such as 0.3C, 1C, 2C, 3C, etc.) with constant current to 2.5V. Voltage data, depth of discharge data and temperature rise data are collected simultaneously. The experiment is repeated multiple times for each rate and the average value is taken. Finally, the data is preprocessed to remove abnormal data (such as abnormal data with sudden voltage drop). The experiment is repeated multiple times for each rate and the average value is taken. Finally, multiple sets of low rate experimental data are obtained and U-DOD (voltage-depth of discharge) curves are plotted.

[0048] The calculation of the depth of discharge is as follows: First, determine the nominal capacity of the battery cell to be tested. Then, during the discharge phase of the low-rate charge-discharge experiment, collect the discharge current and discharge time in real time through the charge-discharge test system. Calculate the current discharged capacity (in Ah) according to the formula "discharged capacity = discharge current × discharge time". Finally, divide the real-time collected discharged capacity by the nominal capacity of the battery cell and multiply by 100% to obtain the corresponding depth of discharge.

[0049] In step 1, under the low-rate operating conditions of 0.3C-3C, the internal heat generation rate of the battery is low, there is no safety risk of electrolyte decomposition or cell bulging, and the cell life decay in a single experiment is lower than that in high-rate experiments, which greatly reduces the experimental cost. On the other hand, by rigorously collecting experimental data of low-rate charging and discharging, a data foundation is provided for subsequent model verification and related training, avoiding subsequent evaluation bias caused by data distortion.

[0050] In step 2, the embodiments of the present invention verify the NTGK model based on experimental data to ensure the reliability of the model simulation accuracy. The NTGK model is a simple semi-empirical electrochemical model that can be used for efficient battery modeling. It derives battery characteristics from experimental values ​​and uses them as input values ​​to calculate battery performance and obtain heat distribution with low computational cost.

[0051] See Figure 2 Step S20 verifies and adjusts the NTGK model based on the voltage data and the depth of discharge data to obtain the target NTGK model, including:

[0052] S21, Step 2.1: Based on the voltage data and the depth of discharge data, output simulation results using the NTGK model. Verify the temperature rise data using the simulation results. If the verification results meet the preset error requirements, confirm the accuracy of the NTGK model; if the verification results do not meet the preset error requirements, adjust the parameters of the NTGK model and re-execute the simulation and verification in Step 2 until the error meets the preset requirements. The preset error requirements, for example, are a temperature prediction error of no more than 2℃ and a heat generation rate or heat flux density prediction error of no more than 5%.

[0053] Specifically, in step 2, based on the experimental data obtained in step 1, this embodiment of the invention constructs and verifies the NTGK model in the FLUENT software. Step 2 specifically involves, for example, initializing the model by importing the basic physical parameters of the power battery cell (such as electrode area, material thermal conductivity, etc.) and calling the FLUENT NTGK model; then transforming the data and calculating the current density J (J = discharge current / electrode area, unit A / m²) based on the discharge current and electrode area corresponding to each low-rate in step 1. 2The process involves converting the voltage-depth of discharge (U-DOD) curve into a voltage-current density (UJ) curve adapted to the NTGK model. The UJ curve is then input into the NTGK model, with the ambient temperature and discharge rate (0.3C-3C) set as in step 1. The model is run to output simulated temperature rise data. Finally, simulation and verification are performed, comparing the simulated temperature rise data with the measured temperature rise data from step 1, calculating the temperature difference error, and ensuring the error is less than the preset error range. If the error exceeds the limit, the initial values ​​of the material's thermal conductivity and polarization resistance are adjusted until the error meets the preset error range. Then, the UJ curve is linearly fitted, and parameters U and Y are extracted. Parameter U is the voltage intercept of the UJ curve when the current density is 0. The parameter Y is the reciprocal of the slope of UJ. Then, polynomial fitting is performed, and the fitting error is controlled to meet the preset requirements. For example, the fitting error range is controlled within ±5%. If it exceeds the fitting error range, the fitting function parameters are re-optimized. The relationship between U, Y and DOD is fitted using polynomials to obtain the polynomials of U and parameter Y under low-rate conditions (0.3C-3C). Finally, U and parameter Y are imported into the FLUENT NTGK model to construct a three-dimensional model of the battery cell, simulate the temperature rise under low-rate conditions (0.3C-3C), and output the simulation results. The simulation results are compared with the experimental data to ensure that the error is within an acceptable range, so as to ensure the high prediction accuracy of the NTGK model under low-rate conditions. This invention establishes a high-precision NTGK model under low-rate operating conditions, ensuring that the model accurately describes the actual correlation between internal electrochemical parameters, heat generation processes, and temperature changes within the battery. This preserves the physical mechanism-based calculation basis for subsequent high-rate thermal characteristic assessments of the battery. Simultaneously, low-rate verification proactively avoids deviations in basic model parameters (such as initial values ​​of material thermal conductivity and polarization resistance), preventing error amplification in high-rate scenarios and laying the foundation for the accuracy of subsequent high-rate thermal characteristic assessments. In step 3, based on the discharge depth data obtained in step 1, the corresponding charge / discharge rate data, and the voltage data obtained in step 1, this invention constructs a Gaussian process regression model. Through model training and parameter adjustment, this invention optimizes model performance, ultimately obtaining an optimized Gaussian process regression model. This ensures that the model can output corresponding prediction results (e.g., predicted voltage) based on the input charge / discharge rate and discharge depth data.

[0054] This invention addresses the problem of traditional NTGK models lacking cross-rate prediction capabilities by constructing and optimizing a Gaussian process regression model. This optimized model does not rely on high-rate experimental data; by learning the patterns of low-rate (0.3C-3C) data, it can predict high-rate (≥5C) voltages, avoiding the limitation of traditional models that "cannot be evaluated without high-rate data." Furthermore, the optimized Gaussian process regression model possesses good prediction accuracy, providing reliable data support for the subsequent extraction of parameters U and Y under high-rate operating conditions, ensuring the accuracy of subsequent evaluation steps.

[0055] In step 4, based on the voltage data and depth of discharge data obtained in step 1, and combined with the Gaussian process regression optimization model obtained in step 3, this embodiment of the invention generates a voltage-current density curve at high rate (≥5C). Then, characteristic parameters are extracted from this curve, including parameter U (voltage intercept when current density is 0) and parameter Y (reciprocal of the curve slope). The values ​​of the two fitted parameters are obtained through fitting. This embodiment of the invention, by combining the Gaussian process regression optimization model with the aforementioned relevant data, obtains the key characteristic parameters U and Y without relying on high-rate experiments. This solves the problem of traditional methods relying on high-rate experiments for parameter calibration, avoiding the safety risks and cost of high-rate experiments, and providing accurate input parameters for subsequent battery pack thermal characteristic simulation, ensuring the accuracy of thermal characteristic evaluation under high-rate conditions.

[0056] In step 5, this embodiment of the invention constructs a battery pack model based on the individual battery cells from step 1, and imports the parameters U and Y obtained in step 4 into the NTGK module. Simultaneously, an energy equation is established, the battery pack model is meshed, boundary conditions are set, and the SIMPLE algorithm is used for calculation. Finally, the temperature field distribution, heat generation power, and inter-cell temperature difference data of the battery pack model are output. This step, by constructing a battery pack model and importing high-rate parameters, directly achieves battery pack-level thermal characteristic evaluation, solving the problem that traditional models are difficult to adapt to the complex structure of battery packs. The output thermal characteristic prediction results, such as the output temperature rise rate, maximum temperature, temperature distribution, and thermal uniformity, can intuitively reflect the thermal distribution characteristics of the battery pack under high-rate operating conditions, providing a specific and reliable reference for battery pack thermal management design. At the same time, the application of the SIMPLE algorithm ensures evaluation efficiency.

[0057] Specifically, the energy equation describes the heat generation and transfer mechanism of a single battery cell or battery pack during operation, and is the core governing equation of the thermal prediction model. The equation is as follows:

[0058]

[0059] Where ρ is the material density, c pis the specific heat capacity, T is the temperature, k is the thermal conductivity, and q is the volumetric heat source term, which includes various heat generation mechanisms.

[0060] In step 6, this embodiment adjusts the hyperparameters of the Gaussian process regression model and the polynomial coefficients corresponding to parameters U and Y through error feedback, and re-executes the simulation calculation to form an adaptive correction closed loop for the model. The hyperparameters of the Gaussian process regression model, the parameter U, and the parameter Y are adjusted. In this embodiment, the verification and correction are used to ensure the accuracy of the thermal characteristic evaluation results. By adjusting the key parameters of the model (the hyperparameters of the Gaussian process regression model, the polynomial coefficients corresponding to U and Y) through the error feedback mechanism and re-executing the simulation calculation to form an adaptive correction closed loop, the accuracy of high-rate thermal characteristic evaluation can be continuously optimized.

[0061] In this embodiment, the verification and correction in step S60 is used to verify the reliability of the thermal characteristic prediction results. By adjusting the relevant parameters of the Gaussian process regression optimization model and the fitting coefficients corresponding to parameters U and Y through error feedback, steps 3 to 5 are re-executed to form an adaptive correction closed loop, which provides a guarantee for improving the accuracy of the high-rate thermal characteristic evaluation results.

[0062] The method for evaluating the high-rate thermal characteristics of power batteries provided in this invention obtains basic data on voltage, depth of discharge, and temperature rise by conducting low-rate (0.3C-3C) charge-discharge experiments, thus avoiding the safety risks and cost of high-rate experiments. Furthermore, this invention verifies the low-rate reliability of the NTGK model based on experimental data, retains the model's physical interpretability to address the lack of mechanism in purely data-driven algorithms, and then learns the patterns of low-rate data through a GPR model (Gaussian process regression model) to achieve high-rate voltage prediction and extraction of U and parameters Y. This solves the problems of insufficient high-rate accuracy and lack of cross-rate prediction capability in traditional NTGK models. Finally, U and parameters Y are imported into the NTGK model to construct a battery pack simulation model, outputting thermal characteristic data to adapt to the complex structure of the battery pack. After outputting the thermal characteristic prediction results, further... An adaptive closed loop of "simulation output - result verification - error feedback - parameter optimization - re-simulation" is constructed through verification and calibration: First, the matching degree between the thermal characteristic prediction results and the actual operating conditions is verified to determine whether the error meets the preset requirements; if the error exceeds the threshold, the hyperparameters of the Gaussian process regression model and the polynomial coefficients corresponding to parameters U and Y are precisely adjusted through the error feedback mechanism. After correcting the model and parameter deviation, the high-rate voltage prediction, parameter extraction and battery pack thermal characteristic simulation process are re-executed until the thermal characteristic prediction error meets the actual application standards. Thus, the embodiments of the present invention significantly improve the accuracy of high-rate thermal characteristic evaluation while reducing reliance on high-rate experiments and ensuring physical support, providing reliable technical support for the work in related fields such as battery pack safety design, thermal management strategy formulation and life optimization.

[0063] See Figure 3 Step S30, based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data, obtains the Gaussian process regression optimization model, including:

[0064] S31, Step 3.1: Based on the Gaussian process regression model, the discharge depth data, and the charge / discharge ratio data corresponding to the discharge depth data, output predicted voltage data, and construct a training dataset D = {(x1,y1),...,(x n ,y n )}, where x is the input feature and y is the output label (n≥500);

[0065] S32, Step 3.2: Select the exponential covariance function, maximize the logarithmic boundary likelihood using the conjugate gradient method for hyperparameter optimization, and train the Gaussian process regression model to obtain the Gaussian process regression optimization model.

[0066] In this embodiment, the training dataset is used to provide model learning samples, where x is a combination of discharge depth data and corresponding charge / discharge rate data, used to specifically describe the battery's operating state; y is the predicted voltage data corresponding to this combination, serving as the output label. This dataset ensures that the model learns the mapping relationship between discharge depth, charge / discharge rate, and voltage. Furthermore, this embodiment trains and optimizes a Gaussian process regression model, selects an exponential covariance function for the Gaussian process regression model, and then uses the conjugate gradient method to optimize the model's hyperparameters, maximizing the logarithmic boundary likelihood to determine the optimal values ​​of the hyperparameters.

[0067] Specifically, the formula for the exponential covariance function is:

[0068]

[0069] in, Let l be the signal variance and l be the length scale. x1 represents the noise variance, x2 represents the rate data of the low-rate charge-discharge test, and x2 represents the high-rate data to be tested.

[0070] Specifically, the log-boundary likelihood formula is:

[0071]

[0072] Where y is the output label vector (e.g., a vector composed of observed values ​​such as voltage of the power battery under different operating conditions), X is the input feature matrix (e.g., a matrix composed of input features such as the depth of discharge and charge / discharge rate of the power battery), θ is the hyperparameter of the Gaussian process regression model (e.g., signal variance, length scale, noise variance, etc. in the covariance function), C is the covariance matrix, det(C) is the determinant of the covariance matrix C, and n is the number of samples.

[0073] See Figure 4 Step S40, which involves obtaining the voltage-current density curve based on the voltage data, the depth of discharge data, and the Gaussian process regression optimization model, includes:

[0074] S41, Step 4.1, Based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data, obtain the voltage prediction data and discharge depth prediction data at high rates;

[0075] S42, Step 4.2, Obtain the voltage-discharge depth prediction curve based on the voltage prediction data and the discharge depth prediction data;

[0076] S43, Step 4.3, convert the voltage-discharge depth prediction curve into a voltage-current density curve.

[0077] In this embodiment, the voltage-depth of discharge prediction curve is used to intuitively reflect the relationship between voltage and depth of discharge at high rates. The voltage-current density curve obtained by conversion can further extract characteristic parameters U and Y, providing data support for subsequent evaluation of battery pack thermal characteristics.

[0078] In S40, the parameter U and the parameter Y are associated with the discharge depth prediction data through a fifth-order polynomial.

[0079] The fifth-degree polynomial is:

[0080] U(DOD) = 3.36 - 0.86DOD + 5.35DOD 2 -15.36DOD 3 +19.04DOD 4 -8.46DOD 5

[0081] Y(DOD)=807.44-6209.15DOD+31990.13DOD 2 -68158.6DOD 3 -63608.28DOD 4 -22097.2DOD 5

[0082] Here, DOD refers to Discharge Depth Data. The DOD includes both low-rate DOD data used to train a fifth-order polynomial to establish the correlation between DOD and parameters U and Y, and DOD prediction data applied to high-rate scenarios based on this correlation. This polynomial yields the parameters U and Y corresponding to different DODs under high-rate conditions, providing parameter support for subsequent high-rate thermal characteristic evaluation. In simpler terms, a fifth-order polynomial model is first trained using low-rate DODs, and then this model is applied to high-rate scenarios to predict the U and Y parameters corresponding to different DODs at high rates.

[0083] To facilitate understanding of the method for estimating the high-rate thermal characteristics of the power battery pack provided in the first embodiment of the present invention, specific experimental examples are given below.

[0084] Experimental Example 1

[0085] This experiment selected the LF50K lithium iron phosphate battery cell as the experimental object. Some parameters are shown in Tables 1.1 and 1.2. This model is widely used in power battery packs for new energy vehicles and is representative.

[0086] Table 1.1

[0087] category name Chemical formula cathode materials Lithium iron phosphate <![CDATA[LiFePO4]]> Anode material carbon C Blue film material polyethylene PE

[0088] Table 1.2

[0089] category numerical values unit Energy density 130 w / kg Specific heat capacity 1296 J / (kg·K) nominal capacity 50 Ah

[0090] Experimental equipment: Hubei Depu Electric BTS20-5V / 4×200A charge and discharge test system (current accuracy ±0.1%FS, voltage accuracy ±0.05%FS), Dongguan Hongjin HYF-TH-150DH constant temperature chamber (temperature control accuracy ±0.5℃), PT100 thermocouple (measurement accuracy ±0.1℃) and data acquisition card.

[0091] Experimental procedure:

[0092] 1. First, conduct charge-discharge experiments under low-rate conditions and collect experimental data. The specific steps are as follows:

[0093] (1) Prepare 3 LF50K lithium iron phosphate battery cells and ensure that their initial SOC (i.e., state of charge, which reflects the proportion of the battery's current charge to its full charge state) are consistent (charged to full charge at 0.3C).

[0094] (2) Place the battery cells into the HYF-TH-150DH constant temperature chamber, set the temperature to 303.15K, and let stand for 10 hours;

[0095] (3) Using the BTS20-5V / 4×200A charging and discharging system, charge to 3.65V at a rate of 0.3C, and then charge at a constant voltage until the current is less than 0.05A;

[0096] (4) After standing for 3 hours, discharge to 2.5V at rates of 0.3C, 1C, 2C and 3C respectively. Repeat the experiment 3 times for each battery at each rate and collect U-DOD curves and temperature rise data.

[0097] (5) Preprocess the experimental data: remove outliers (such as data with a voltage drop of more than 0.1V) and take the average value as the final low-rate data.

[0098] 2. Based on the NTGK model, the voltage data, and the depth of discharge data, output simulation results. Verify the temperature rise data using the simulation results to ensure that the error between the simulation results and the temperature rise data is within a preset error range. The specific steps are as follows:

[0099] (1) Curve conversion: Combine each low-rate U-DOD curve with the corresponding current density to convert it into a voltage-current density (UJ) curve;

[0100] (2) Parameter extraction: Linear fitting of the UJ curve for each discharge depth (DOD, 5% interval) is performed to extract the intercept U and slope, and the reciprocal of the slope is calculated to obtain Y;

[0101] (3) Polynomial fitting: A 5th-order polynomial was used to fit the relationship between U, Y and DOD to obtain the polynomials of U and parameter Y at low magnification:

[0102] Wherein, the parameter U polynomial is:

[0103] U(DOD) = 3.36 - 0.86DOD + 5.23DOD 2 -15.36DOD 3 +32.04DOD 4 -5.92DOD 5

[0104] Parametric polynomial Y:

[0105] Y(DOD)=809.42-5132.23DOD+23612.41DOD 2 -72936.12DOD 3

[0106] +69860.4DOD 4 -16753.66DOD 5

[0107] (4) Model Validation: Parameters U and Y were imported into the FLUENT NTGK model to construct a three-dimensional model of the battery cell (size 185.0mm×135.0mm×29.3mm, tetrahedral mesh, 100,000 nodes). Temperature rise at 1C, 2C, and 3C rates was simulated. The results showed that the maximum temperature difference between the experiment and the simulation was 1.05℃, 1.7℃, and 1.26℃, respectively, and the error was within the preset error range, validating the low-rate accuracy of the NTGK model.

[0108] 3. Based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data, a Gaussian process regression optimization model is obtained. The specific steps are as follows:

[0109] (1) Input and output definitions: Using "charge / discharge ratio (C) + depth of discharge (DOD)" as the input feature vector x = [C, DOD], and the corresponding open-circuit voltage (U) as the output label y = U, a training dataset D = {(x1, y1), ..., (x...} is constructed. n ,y n (n≥500, covering 0.3C~3C, 0%~100% DOD);

[0110] (2) Gaussian process prior: Assume that the voltage output follows a Gaussian process y~GP(m(x),K(x,x′)), where the mean function m(x)=0 and the covariance function K(x,x′) describes the input similarity;

[0111] (3) Noise handling: Consideration of experimental noise The observed values ​​satisfy y = f(x) + εf(x) as the true voltage function.

[0112] The process and related formulas for kernel function selection and hyperparameter optimization are as follows:

[0113] (1) Kernel function: The exponential covariance function is selected to accurately track the local characteristics of the U-DOD curve (such as the discharge plateau and the sudden drop in voltage at the end). The expression is:

[0114] in Let l be the signal variance and l be the length scale.

[0115] (2) Hyperparameter optimization: The conjugate gradient algorithm is used to maximize the log-boundary likelihood to determine the hyperparameters. The log-boundary likelihood formula is:

[0116] ( I is the identity matrix), iterate until the likelihood value converges (threshold 10). -6 ).

[0117] 4. Based on the voltage data, the depth of discharge data, and the Gaussian process regression optimization model, obtain the voltage-current density curve, and extract feature parameters from the voltage-current density curve to fit parameters U and Y. The specific steps are as follows:

[0118] (1) Combine the high-magnification (5C) DOD sequence (0%~100%, 5% interval) with the magnification value to form the input x. * =[5C,DOD], input the trained GPR model, predict the 5CU-DOD curve and the 95% confidence interval. Validation shows that the predicted voltage corresponding to the 5% to 95% DOD interval has an error of <0.05V compared to the experiment, which meets the requirements for subsequent parameter fitting.

[0119] (2) The fitting process of parameters U and Y under high magnification conditions is as follows:

[0120] UJ curve conversion: At a 5C rate, the discharge current of the LF50K battery is 250A. The current density is calculated based on the electrode area, and the predicted 5CU-DOD curve is converted into the UJ curve.

[0121] Parameter extraction: For each point in the DOD 5% to 95% range, the UJ curve is linearly fitted to extract U and the slope, and the reciprocal of the slope is calculated to obtain Y;

[0122] Polynomial Fitting: The least squares method was used to fit U, Y, and DOD using a 5th-order polynomial, with the fitting error controlled within 5%, yielding the polynomials for parameters U and Y under 5C:

[0123] U(DOD) = 3.36 - 0.86DOD + 5.35DOD 2 -15.36DOD 3 +19.04DOD 4 -8.46DOD 5

[0124] Y(DOD)=807.44-6209.15DOD+31990.13DOD 2 -68158.6DOD 3

[0125] -63608.28DOD 4 -22097.2DOD 5

[0126] (3) Import the fitted parameters U and Y into the FLUENT NTGK model and enable the energy equation (considering reversible heat, ohmic heat, polarization heat, and convective heat transfer; the heat generation model is q = q re +q ohm +q act +q exq rev It is a reversible heat (generated by the entropy change of an electrochemical reaction), q ohm Ohmic heat (generated by internal resistance), q act The heat of polarization (caused by the overpotential of the electrochemical reaction), q ax To account for the additional heat term generated by convective heat transfer or axial heat conduction of the current collector, the ambient temperature was set to 303.15K, the external computational domain was 0.8m×0.8m×0.8m (consistent with the constant temperature chamber), and convective heat transfer boundary conditions were set (radiation was ignored). The SIMPLE algorithm was used for the calculation.

[0127] (4) Model accuracy verification: Comparing the highest temperature of the battery cell in the fusion model and the experiment at 5C rate: the original NTGK model predicted a difference of 5.82℃, while the GPR-NTGK model predicted only 0.96℃, and the difference in the highest temperature prediction was reduced by 4.86℃, proving that the accuracy of the fusion model was significantly improved.

[0128] 5. Prediction of high-rate thermal characteristics of power battery packs: In this stage, the GPR-NTGK model is applied to power battery packs to predict high-rate thermal characteristics, supporting engineering applications. The specific steps are as follows:

[0129] (1) Battery pack model construction:

[0130] A battery pack consisting of five LF50K cells connected in series was constructed, with the following parameters:

[0131] Electrical parameters: Rated capacity 250Ah, rated voltage 16V, operating voltage 12.5V~18.25V;

[0132] Geometric dimensions: Individual cell 185.0mm×135.0mm×29.3mm, battery pack 185.0mm×135.0mm×166.5mm (individual cells stacked along the thickness with a 2mm gap);

[0133] Mesh generation: Tetrahedral mesh, 2.2 million elements, 197,000 nodes, with finer meshing in critical areas (interstellar gaps, tabs) to ensure mesh independence (mesh increase of 10% results in temperature change <0.1℃).

[0134] Specifically, see Figure 5 , Figure 5 This diagram shows the cross-sectional locations of battery packs A and B, as well as the battery serial numbers. Figure 5 The numbers 1 to 5 are battery numbers, arranged sequentially from left to right along the thickness direction. Section A is the central section perpendicular to the width direction of the battery (passing through the center of batteries 2 and 3); Section B is the central section perpendicular to the length direction of the battery (passing through the center of all batteries).

[0135] (2) Thermal property prediction results:

[0136] Simulation of the battery pack discharge process at 5C rate yielded the following key results:

[0137] Temperature rise rate: The average temperature rise rate is 2.97℃ / min, which is 24.07 times that of 1C (0.123℃ / min) and 2.39 times that of 3C (1.243℃ / min). At high rates, the positive feedback of "heating-temperature rise-increased internal resistance" leads to nonlinear growth of temperature rise.

[0138] Maximum temperature: The maximum temperature at the end of discharge was 61.17℃, which exceeds the safe operating range of lithium iron phosphate batteries (-20℃~55℃) and poses a risk of thermal runaway.

[0139] Temperature distribution: High temperatures are concentrated in batteries 2, 3, and 4 at the bottom center of the module. The temperature difference between the center and outer batteries is 0.73℃, and the temperature difference between the bottom and top of the batteries is 0.5℃. The temperature standard deviation is 0.35℃, indicating poor uniformity.

[0140] Thermal runaway risk: The temperature of the central battery is close to the electrolyte decomposition temperature (about 65°C), requiring a thermal management system to cool it down.

[0141] Specifically, see Figure 6 , Figure 6 This is a temperature cloud map of the cross-section of battery pack A at a 5C rate. The darker the color, the higher the temperature. The central area (batteries 2, 3, and 4) is the darkest, corresponding to a temperature of 61.17℃; the outer area (batteries 1 and 5) is lighter, corresponding to temperatures of 60.44℃ to 60.67℃.

[0142] 6. Model applicability verification:

[0143] Temperature rise experiments of the battery pack were conducted at 1C, 2C, and 3C rates. The maximum temperature differences between the simulation and the experiment at 12 monitoring points were 1.12℃, 1.35℃, and 1.44℃, respectively. The errors were within the preset error range, verifying the applicability of the model to the battery pack.

[0144] Specifically, see Figure 7 , Figure 7 This is a comparison chart of the U-DOD curve predicted by the GPR model and the experimental curve under 5C operating conditions. The green line is the GPR predicted curve, the red line is the experimental curve, and the shaded area is the 95% confidence interval. In the DOD 5% to 95% range, the two curves have a high degree of overlap and the error is less than 0.05V.

[0145] [Second Embodiment]

[0146] See Figure 8The second embodiment of the present invention provides a high-rate thermal characteristic evaluation system 20 for power batteries. The high-rate thermal characteristic evaluation system 20 includes a processor 400 and a memory 500 connected to the processor 400. The memory 500 may be, for example, a non-volatile memory cavity, and stores a computer program 510. The processor 400 may be, for example, an embedded processor. When the processor 400 executes the computer program 510, it executes the high-rate thermal characteristic evaluation method for power batteries described in the first embodiment.

[0147] For the specific working process and technical effects of the high-rate thermal characteristic evaluation system 20 for power batteries in this embodiment, please refer to the description of the first embodiment above.

[0148] [Third Embodiment]

[0149] See Figure 9 The third embodiment of the present invention provides a storage medium 30 storing a computer program 510.

[0150] Storage medium 30 is, for example, a non-volatile memory, such as magnetic media (e.g., hard disks, floppy disks, and magnetic tapes), optical media (e.g., CD-ROMs and DVDs), magneto-optical media (e.g., optical discs), and hardware devices specifically configured for storing and executing computer program 510 (e.g., read-only memory (ROM), random access memory (RAM), flash memory, etc.). The computer program 510 is stored on storage medium 30 and can be executed by a processor of the device where storage medium 30 is located. Storage medium 30 can be executed by one or more processors or processing devices to implement the high-rate thermal characteristic evaluation method for power batteries described in the first embodiment above.

[0151] Furthermore, it is understood that the foregoing embodiments are merely illustrative examples of the present invention. Provided that the technical features do not conflict, the structure is not contradictory, and the purpose of the invention is not violated, the technical solutions of the various embodiments can be arbitrarily combined and used.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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; and these 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 method for evaluating the high-rate thermal characteristics of a power battery, characterized in that, include: Step 1: Conduct low-rate charge-discharge experiments on individual battery cells to obtain experimental data, including voltage data, depth of discharge data, and temperature rise data. The low-rate charge-discharge experiment ranges from 0.3C to 3C. Step 2: Verify and adjust the NTGK model based on the voltage data and the depth of discharge data to obtain the target NTGK model; Step 3: Obtain the Gaussian process regression optimization model based on the Gaussian process regression model, the discharge depth data, and the charge / discharge rate data corresponding to the discharge depth data; Step 4: Based on the voltage data, the discharge depth data, and the Gaussian process regression optimization model, obtain the voltage-current density curve, and extract feature parameters from the voltage-current density curve to fit parameters U and Y, wherein parameter U is the voltage intercept of the voltage-current density curve when the current density is 0, and parameter Y is the reciprocal of the slope of the voltage-current density curve, wherein the high-rate range is greater than or equal to 5C; Step 5: Establish a battery pack model based on the individual battery cells, import the parameters U and Y into the target NTGK model according to the battery pack model and establish the energy equation, perform mesh generation on the battery pack model and set boundary conditions, and use the SIMPLE algorithm to output the thermal characteristic prediction results of the battery pack model; Step 6: Verify and correct the predicted thermal characteristics. If the verification results meet the preset error requirements, output the final thermal characteristic evaluation results. If the verification result does not meet the preset error requirements, the Gaussian process regression optimization model and the fitting coefficients corresponding to the parameters U and Y are adjusted through error feedback, and the simulation calculations of steps 3 to 5 are re-executed to form an adaptive correction closed loop of the Gaussian process regression optimization model.

2. The method for evaluating the high-rate thermal characteristics of a power battery as described in claim 1, characterized in that, Step 3: Based on the Gaussian process regression model, the discharge depth data, and the charge / discharge ratio data corresponding to the discharge depth data, a Gaussian process regression optimization model is obtained, including: Based on the Gaussian process regression model, the discharge depth data, and the corresponding charge / discharge ratio data, predicted voltage data is output, and a training dataset D = {(x1, y1), ..., (x n ,y n )}, where x is the input feature and y is the output label (n≥500); By selecting the exponential covariance function and maximizing the logarithmic boundary likelihood using the conjugate gradient method for hyperparameter optimization, the Gaussian process regression model is trained to obtain the optimized Gaussian process regression model.

3. The method for evaluating the high-rate thermal characteristics of a power battery as described in claim 2, characterized in that, The formula for the exponential covariance function is: in, Let l be the signal variance and l be the length scale. x1 represents the noise variance, x2 represents the rate data of the low-rate charge-discharge test, and x2 represents the high-rate data to be tested.

4. The method for evaluating the high-rate thermal characteristics of a power battery as described in claim 2, characterized in that, The log-boundary likelihood formula is: Where y is the output label vector, X is the input feature matrix, θ is the hyperparameter of the Gaussian process regression model, C is the covariance matrix, det(C) is the determinant of the covariance matrix C, and n is the number of samples.

5. The method for evaluating the high-rate thermal characteristics of a power battery as described in claim 1, characterized in that, Step 4 involves obtaining the voltage-current density curve based on the voltage data, the depth of discharge data, and the Gaussian process regression optimization model, including: Based on the Gaussian process regression model, the discharge depth data, and the charge / discharge ratio data corresponding to the discharge depth data, voltage prediction data and discharge depth prediction data at high rates are obtained. A voltage-discharge depth prediction curve is obtained based on the voltage prediction data and the discharge depth prediction data; The voltage-discharge depth prediction curve is converted into a voltage-current density curve.

6. The method for evaluating the high-rate thermal characteristics of a power battery as described in claim 5, characterized in that, The parameter U and the parameter Y are associated with the discharge depth prediction data through a fifth-order polynomial; The formula for the fifth-degree polynomial is as follows: U(DOD)=3.36-0.86DOD+5.35DOD 2 -15.36DOD 3 +19.04DOD 4 -8.46DOD 5 Y(DOD)=807.44-6209.15DOD+31990.13DOD 2 -68158.6DOD 3 -63608.28DOD 4 -22097.2DOD 5 Where DOD stands for Depth of Discharge data.

7. The method for evaluating the high-rate thermal characteristics of a power battery according to claim 1, characterized in that, Step 2: Based on the voltage data and the depth of discharge data, verify and adjust the NTGK model to obtain the target NTGK model, including: Based on the voltage data and the depth of discharge data, simulation results are output through the NTGK model. The temperature rise data is verified through the simulation results. If the verification results meet the preset error requirements, the accuracy of the NTGK model is confirmed. If the verification results do not meet the preset error requirements, the parameters of the NTGK model are adjusted and step 2 of the simulation and verification is repeated until the error meets the preset requirements.

8. A system for evaluating the high-rate thermal characteristics of a power battery, characterized in that, include: A processor and a memory connected to the processor, the memory storing a computer program, wherein when the processor executes the computer program, it performs the evaluation method for high-rate thermal characteristics of a power battery as described in any one of claims 1 to 7.

9. The evaluation system for high-rate thermal characteristics of power batteries according to claim 8, characterized in that, The system also includes a network connection module for communicating with the battery management system to perform real-time temperature monitoring and safety warnings.

Images