Method and system for detecting gas production state after thermal runaway of battery

Abstract

The invention provides a method for detecting a gas production state after thermal runaway of a battery, which comprises the following steps: placing a lithium battery in a closed anti-explosion tank, heating to trigger runaway, constructing a coupling matrix through pressure and temperature data acquired by a dynamic pressure compensation model, and correcting pressure distribution by using a nonlinear thermodynamic equation; analyzing the concentration of gas components by adopting a Fourier transform infrared spectrometer, and calculating the total gas production amount by combining multiple regression analysis; and quantizing energy dissipation based on a second law of thermodynamics to generate an energy release characteristic matrix, mapping the energy release characteristic matrix to a high-dimensional space to generate a risk index, and finally predicting thermal runaway probability distribution through a support vector machine. According to the method, high-precision measurement and analysis of the gas production rate in the thermal runaway process of the battery are realized through dynamic pressure compensation, multi-dimensional gas component analysis and nonlinear thermodynamic modeling, and the testing efficiency and the early warning capability of the battery are improved.

Description

Technical Field

[0001] This invention relates to the field of battery safety monitoring technology, and in particular to a method and system for detecting gas generation after battery thermal runaway. Background Technology

[0002] Lithium-ion batteries have been widely used in electric vehicles, energy storage systems, and other fields due to their significant advantages such as high energy density, long cycle life, and low self-discharge rate. However, thermal runaway, a battery safety issue, is a key factor restricting the further development of lithium-ion batteries. During thermal runaway, a large amount of gas, such as hydrogen, carbon monoxide, and carbon dioxide, is generated and released inside the battery. The generation of these gases not only directly affects the pressure relief design of the battery pack but is also closely related to the total amount and concentration of harmful substances in the flue gas.

[0003] Currently, testing methods for the gas generation characteristics of lithium-ion batteries during thermal runaway typically involve placing the battery in a sealed pressure vessel and triggering thermal runaway through heating, overcharging, or needle penetration. Temperature and pressure changes are recorded, and gas is collected to analyze the gas generation rate, gas composition, and pressure variation patterns. However, during the cooling phase after thermal runaway, a prolonged settling period is usually required to allow temperature and pressure values ​​to stabilize before calculating the gas generation rate using the ideal gas law. This process lacks clear judgment boundaries; too short a settling time may lead to inaccurate data, while too long a settling time increases time costs, reduces testing efficiency, and may result in underestimated gas generation rates due to potential gas leakage. Furthermore, the applicability of the ideal gas law in thermal runaway gas generation analysis has not been fully validated, introducing uncertainty into the accurate calculation of battery gas generation. Therefore, designing a method and system for detecting gas generation after battery thermal runaway is essential. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for detecting gas production after battery thermal runaway, so as to achieve high-precision measurement and analysis of gas production during battery thermal runaway through dynamic pressure compensation, multi-dimensional gas component analysis and nonlinear thermodynamic modeling, thereby improving test efficiency and result reliability.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for detecting gas generation after battery thermal runaway includes the following steps:

[0007] The battery under test was placed in a sealed explosion-proof container and thermal runaway was triggered by external heating.

[0008] Based on the dynamic pressure compensation model, pressure change data inside the sealed explosion-proof container during thermal runaway is collected in real time, and a pressure-temperature coupling matrix is ​​constructed by combining the temperature curves recorded by the temperature sensor.

[0009] The actual gas pressure distribution is obtained by correcting the pressure and temperature coupling matrix using nonlinear thermodynamic equations.

[0010] The gas composition inside the sealed explosion-proof container was analyzed in multiple dimensions using a Fourier transform infrared spectrometer, and the gas concentration distribution vector was extracted.

[0011] The total amount of gas produced during thermal runaway is obtained by fitting the gas concentration distribution vector with the actual gas pressure distribution using a multiple regression analysis algorithm.

[0012] Based on the second law of thermodynamics, the total amount of gas produced during thermal runaway is quantified using the entropy increase formula, resulting in the energy release characteristic matrix of the thermal runaway process.

[0013] The energy release feature matrix is ​​mapped to a high-dimensional safety assessment space to generate a battery thermal runaway risk index.

[0014] Based on the battery thermal runaway risk index, a support vector machine classifier is used to predict the thermal runaway probability distribution of the battery under test under different operating conditions.

[0015] Optionally, based on a dynamic pressure compensation model, pressure change data within the sealed explosion-proof container during thermal runaway is collected in real time, and a pressure-temperature coupling matrix is ​​constructed by combining the temperature curves recorded by temperature sensors, including:

[0016] The original pressure signal sequence inside the sealed explosion-proof container is acquired using a high-frequency pressure sensor at the first sampling rate.

[0017] Temperature signal sequences are synchronously acquired by a multi-point armored thermocouple array arranged inside a sealed explosion-proof container.

[0018] The Kalman filter algorithm is used to perform noise reduction and state estimation on the original pressure signal sequence and temperature signal sequence, respectively.

[0019] Based on the material thermal expansion coefficient and geometric model of the sealed explosion-proof tank, a thermally induced volume expansion error function of the tank is constructed.

[0020] The pressure signal sequence after noise reduction is compensated by the tank thermal volume expansion error function to eliminate the pressure measurement error caused by the expansion of the sealed explosion-proof tank itself.

[0021] The compensated pressure signal sequence and the denoised temperature signal sequence are aligned and interpolated on a shared timestamp reference to generate a pressure and temperature coupling matrix.

[0022] Optionally, the pressure and temperature coupling matrix is ​​corrected using nonlinear thermodynamic equations to obtain the actual gas pressure distribution, including:

[0023] The virial equation is selected as the nonlinear thermodynamic equation to characterize the behavior of high-temperature and high-pressure nonideal gases during thermal runaway.

[0024] Based on the data points in the pressure-temperature coupling matrix, an overdetermined system of equations is constructed.

[0025] The second and third virial coefficients of the virial equation were determined by Gibbs ensemble Monte Carlo simulation.

[0026] The singular value decomposition algorithm is used to solve the overdetermined system of equations, and the global optimal solution of the virial equation is obtained.

[0027] The global optimal solution is substituted into the virial equation, and the equivalent pressure of the ideal gas at different time points is obtained through iterative calculation, thus forming the actual gas pressure distribution.

[0028] Optionally, the gas composition inside the sealed explosion-proof container is analyzed in multiple dimensions using a Fourier transform infrared spectrometer, and the gas concentration distribution vector is extracted, including:

[0029] The mid-infrared detection beam is introduced into the sealed explosion-proof container through the optical window, and the transmission spectrum containing the absorption peaks of the gas-producing components is collected.

[0030] The net absorption spectrum data is obtained by removing the pre-collected inert gas background spectrum from the transmission spectrum.

[0031] Based on a standard gas database, a standard absorption spectral dictionary of potential gaseous products is constructed.

[0032] The net absorption spectral data is decomposed into a linear superposition of the components in the standard absorption spectral dictionary using a multivariate curve fitting algorithm.

[0033] The instantaneous molar concentration of each gas component is obtained by optimizing the superposition coefficient of the linear superposition using the least squares method.

[0034] The instantaneous molar concentrations of all measured gas components are arranged in a time series to form a gas concentration distribution vector.

[0035] Optionally, the gas concentration distribution vector is fitted to the actual gas pressure distribution using a multiple regression analysis algorithm to obtain the total amount of gas produced during thermal runaway, including:

[0036] A multiple linear regression model is constructed by taking the concentration of each component in the gas concentration distribution vector as the independent variable and the actual gas pressure distribution as the dependent variable.

[0037] The optimal number of principal components for a multiple linear regression model is determined using cross-validation.

[0038] Introduce a ridge regression regularization term into the multiple linear regression model;

[0039] A multiple linear regression model is trained using gradient descent, and a mapping relationship from independent variables to dependent variables is constructed.

[0040] Based on the time series, the total number of moles of gas evolved from the battery at each time point is calculated according to the mapping relationship, and the total amount of gas produced by thermal runaway is obtained by integration.

[0041] Optionally, based on the second law of thermodynamics, the total amount of gas produced during thermal runaway is quantified using the entropy increase formula to obtain the energy release characteristic matrix of the thermal runaway process, including:

[0042] The final mole fraction of each gas component at the end of thermal runaway is obtained from the gas concentration distribution vector;

[0043] Query the thermodynamic standard database to obtain the molar entropy of each gas component under standard conditions;

[0044] Temperature and pressure corrections are applied to the molar entropy of each gas component based on the pressure-temperature coupling matrix.

[0045] Based on the Gibbs formula for mixing entropy, the total entropy of the mixed gas is calculated according to the corrected molar entropy.

[0046] The difference between the total entropy value and the entropy value of the initial state is used to obtain the increase in the total entropy of the system.

[0047] The total entropy increase of the system, the peak gas production rate, and the total amount of gas produced during thermal runaway are combined into a three-dimensional vector, which serves as the energy release characteristic matrix of the thermal runaway process.

[0048] Optionally, the energy release feature matrix is ​​mapped to a high-dimensional safety assessment space to generate a battery thermal runaway risk index, including:

[0049] Construct a benchmark feature database containing safety level labels; the benchmark feature database is generated from historical battery test data;

[0050] Principal component analysis is used to reduce the dimensionality of the benchmark feature database, and an N-dimensional orthogonal security assessment space is constructed based on the N principal components with the highest contribution rates extracted as basis vectors.

[0051] Projecting the energy release feature matrix onto the orthogonal security assessment space yields the coordinate vector in the orthogonal security assessment space.

[0052] Calculate the Euclidean distance between the coordinate vector and the preset absolute safety origin;

[0053] By normalizing the Euclidean distance, a battery thermal runaway risk index is obtained.

[0054] Optionally, based on the battery thermal runaway risk index, a support vector machine classifier is used to predict the thermal runaway probability distribution of the battery under test under different operating conditions, including:

[0055] The battery thermal runaway risk index is used as the core input feature;

[0056] Battery design parameters, including cell capacity, positive and negative electrode material systems, and electrolyte composition, are used as auxiliary input features.

[0057] The core input features are concatenated with the auxiliary input features to obtain the enhanced feature vector;

[0058] The radial basis function is used as the kernel function of the support vector machine classifier, and the hyperparameters of the kernel function are optimized and trained using the Bayesian optimization algorithm.

[0059] The enhanced feature vectors are input into the trained support vector machine classifier to obtain the thermal runaway probability distribution.

[0060] A battery gas generation detection system after thermal runaway includes:

[0061] The thermal runaway triggering module is used to place the battery under test in a sealed explosion-proof container and trigger the thermal runaway process through external heating;

[0062] The pressure acquisition module is used to acquire pressure change data in the sealed explosion-proof container in real time during thermal runaway based on a dynamic pressure compensation model, and to construct a pressure-temperature coupling matrix by combining the temperature curves recorded by the temperature sensor.

[0063] The pressure correction module is used to correct the pressure and temperature coupling matrix through nonlinear thermodynamic equations to obtain the actual gas pressure distribution;

[0064] The gas analysis module is used to perform multidimensional analysis of the gas components in a sealed explosion-proof container using a Fourier transform infrared spectrometer, and to extract the gas concentration distribution vector.

[0065] The gas production calculation module is used to fit the gas concentration distribution vector with the actual gas pressure distribution through a multivariate regression analysis algorithm to obtain the total amount of gas produced during thermal runaway.

[0066] The energy assessment module is used to quantify the total amount of gas produced in thermal runaway based on the second law of thermodynamics and the entropy increase formula, so as to obtain the energy release characteristic matrix of the thermal runaway process.

[0067] The risk assessment module is used to map the energy release feature matrix to a high-dimensional safety assessment space to generate a battery thermal runaway risk index.

[0068] The probability prediction module is used to predict the probability distribution of thermal runaway of the battery under test under different operating conditions based on the battery thermal runaway risk index and a support vector machine classifier.

[0069] According to specific embodiments provided by the present invention, the following technical effects are disclosed: The method for detecting gas production state after battery thermal runaway provided by the present invention includes: placing the battery under test in a sealed explosion-proof container and triggering the thermal runaway process by external heating; based on a dynamic pressure compensation model, collecting pressure change data in the sealed explosion-proof container in real time during the thermal runaway process, and constructing a pressure-temperature coupling matrix by combining the temperature curve recorded by a temperature sensor; correcting the pressure-temperature coupling matrix by a nonlinear thermodynamic equation to obtain the actual gas pressure distribution; performing multidimensional analysis of the gas components in the sealed explosion-proof container by a Fourier transform infrared spectrometer and extracting the gas concentration distribution vector; fitting the gas concentration distribution vector with the actual gas pressure distribution by a multivariate regression analysis algorithm to obtain the total amount of gas produced during thermal runaway; quantifying the total amount of gas produced during thermal runaway by the entropy increase formula based on the second law of thermodynamics to obtain the energy release feature matrix of the thermal runaway process; mapping the energy release feature matrix to a high-dimensional safety assessment space to generate a battery thermal runaway risk index; and predicting the thermal runaway probability distribution of the battery under test under different operating conditions by a support vector machine classifier based on the battery thermal runaway risk index. This method achieves high-precision measurement and analysis of gas production during battery thermal runaway through dynamic pressure compensation, multidimensional gas component analysis, and nonlinear thermodynamic modeling, thereby improving battery testing efficiency and early warning capabilities. Attached Figure Description

[0070] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in 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.

[0071] Figure 1 This is a flowchart of the battery gas generation state detection method after thermal runaway according to the present invention;

[0072] Figure 2 This is a flowchart of the energy release feature matrix acquisition process according to an embodiment of the present invention. Detailed Implementation

[0073] 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 of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0075] like Figure 1 As shown, the present invention provides a method for detecting gas generation after battery thermal runaway, comprising the following steps:

[0076] Step 100: Place the battery under test in a sealed explosion-proof container and trigger the thermal runaway process by external heating;

[0077] In some embodiments, the heating device employs resistance wire heating, enabling precise control of the heating power to simulate thermal runaway conditions under different operating conditions. The sealed explosion-proof container is made of high-temperature and high-pressure resistant stainless steel, and its geometry is optimized to ensure that it can withstand extreme pressure during experiments while also meeting the calculation requirements of the volume expansion error function. Furthermore, the sealed explosion-proof container is equipped with an optical window for guiding the detection beam of the Fourier transform infrared spectrometer.

[0078] Step 200: Based on the dynamic pressure compensation model, real-time pressure change data inside the sealed explosion-proof container during thermal runaway is collected, and a pressure-temperature coupling matrix is ​​constructed by combining the temperature curves recorded by the temperature sensor.

[0079] Specifically, a high-frequency pressure sensor uses a sampling rate of 1000Hz to acquire the original pressure signal sequence inside the tank to capture the transient characteristics of pressure changes during thermal runaway. A multi-point armored thermocouple array, composed of K-type armored thermocouples arranged at different spatial locations inside the sealed explosion-proof tank, synchronously acquires the temperature signal sequence inside the tank to monitor the spatial distribution differences of the temperature field within the tank. The thermocouples are arranged in a cross-shaped configuration inside the sealed explosion-proof tank. Then, the Kalman filter algorithm in the dynamic pressure compensation model is used to denoise and estimate the state of the original pressure and temperature signal sequences to eliminate pressure measurement errors caused by external environmental interference. Specifically, let the initial pressure inside the tank be... The initial temperature is During thermal runaway, the expressions for the changes in pressure and temperature inside the vessel over time are: ,in The coefficient representing the sensitivity of temperature changes to pressure. The change in temperature relative to the initial value is represented by this model, which can correct for pressure measurement deviations caused by temperature fluctuations, thereby improving data reliability. Next, based on the material thermal expansion coefficient and geometric model of the sealed explosion-proof tank, a thermally induced volumetric expansion error function for the tank is constructed and used to compensate for the noise-reduced pressure signal sequence, eliminating pressure measurement errors caused by the expansion of the sealed explosion-proof tank itself. The expression is: ,in The coefficient of thermal expansion is This represents the initial volume of the tank. The pressure data is the real-time temperature. Finally, the compensated pressure data and temperature data are interpolated and aligned under a unified time reference to generate a pressure-temperature coupling matrix.

[0080] It should be noted that by combining high-frequency sampling with a multi-point temperature array, high-resolution spatiotemporal acquisition of pressure and temperature signals was achieved; and by establishing a physical model of tank expansion, the container deformation error term was included for the first time. Dynamic compensation calculations were incorporated, and timestamp synchronization technology was adopted to ensure accurate matching of pressure and temperature data on a millisecond-level time scale. This solved the problem of pressure measurement distortion caused by thermal expansion of the sealed container, and enabled accurate synchronous acquisition of pressure and temperature data, thereby significantly improving the accuracy of thermal runaway gas generation detection.

[0081] Step 300: Correct the pressure and temperature coupling matrix using nonlinear thermodynamic equations to obtain the actual gas pressure distribution;

[0082] Specifically, the virial equation is chosen as the nonlinear thermodynamic equation. This equation characterizes the deviation of the corrected ideal gas equation through the statistical mechanics of intermolecular forces, and its expression is: Where P is pressure, n is the number of moles of gas, R is the ideal gas constant, T is temperature, V is volume, and B(T) and C(T) are the second and third virial coefficients, respectively. An overdetermined system of equations is then constructed based on the data points in the pressure-temperature coupling matrix. In this embodiment, a matrix form is used. The second and third virial coefficients are determined through Gibbs ensemble Monte Carlo simulation, a molecular simulation method based on statistical mechanics principles used to calculate the virial coefficients corresponding to the intermolecular forces of gas under high temperature and pressure. Then, the singular value decomposition algorithm is used to solve the overdetermined system of equations to obtain the global optimal solution of the virial equations. This solution is then substituted into the virial equations, and the equivalent pressure of the ideal gas at different time points is obtained through iterative calculation, forming the actual gas pressure distribution.

[0083] Step 400: Perform multidimensional analysis of the gas components inside the sealed explosion-proof container using a Fourier transform infrared spectrometer, and extract the gas concentration distribution vector;

[0084] Specifically, a high-temperature resistant optical window with a transmission band of 2.5~15μm is used to transmit light through a Fourier transform infrared spectrometer (FTIR) with wavelengths ranging from 4000 to 400cm. -1 A mid-infrared detection beam is guided into a sealed explosion-proof container. The beam passes through the container via a coaxial reflection path, and a mercury-cadmium-tellurium detector collects the transmission spectrum containing the absorption peaks of the gas-producing components in real time. The expression for the intensity of the transmission spectrum is:

[0085] ;

[0086] Where v is the wave number and t is time. For the incident light intensity, Let i be the absorption cross section of the i-th gas component. Let L be the instantaneous molecular concentration of the i-th gas, L be the optical path length, and M be the number of gas species to be measured. The pre-collected background spectrum of inert gases is removed from the transmission spectrum to obtain the net absorption spectrum data. Then, based on a standard gas database, this embodiment uses the HITRAN database to extract the absorption lines of potential gaseous products (H2, CO, CO2, CH4, C2H4, etc.) under standard conditions, and performs temperature and pressure broadening corrections to form a standard absorption spectrum dictionary for potential gaseous products. The expression for the broadening correction process is:

[0087] ;

[0088] Where T ref The reference temperature is 296K in this embodiment, P ref The reference pressure is 1 atm in this embodiment. Next, a multivariate curve fitting algorithm is used to decompose the net absorption spectrum data into a linear superposition of the components in the standard absorption spectral dictionary, expressed as:

[0089] ;

[0090] Where, d i Let c be the standard absorption spectral line of the i-th gas. i Let be the superposition coefficients for each component. Then, the least squares method is used to optimize the superposition coefficients of the linear superposition, yielding the instantaneous molar concentration of each gas component, expressed as: ;where N A is Avogadro's constant. Finally, the instantaneous molar concentrations of all measured gas components are arranged in time series to form a gas concentration distribution vector.

[0091] It should be noted that by real-time online detection of gas components during thermal runaway, the testing delay caused by waiting for thermal equilibrium in traditional methods is eliminated. This also solves the analytical error problem caused by overlapping absorption peaks in mixed gases and improves the detection accuracy of low-concentration harmful gases. Furthermore, the establishment of the dynamic concentration distribution vector provides time-series data support for subsequent calculations of total gas production, avoiding the applicability limitations of the ideal gas law in transient processes.

[0092] Step 500: Fit the gas concentration distribution vector to the actual gas pressure distribution using a multivariate regression analysis algorithm to obtain the total amount of gas produced during thermal runaway;

[0093] Specifically, a multiple linear regression model is constructed by taking the concentration of each component in the gas concentration distribution vector as the independent variable and the actual gas pressure distribution as the dependent variable. ,in For the intercept term, These are the regression coefficients. K-fold cross-validation is then used to determine the optimal number of principal components for the multiple linear regression model. A ridge regression regularization term is then introduced into the model. To avoid overfitting, a multiple linear regression model is then trained using gradient descent, with the prediction pressure calculated in the k-th iteration during training. = ,in For bias terms, Let be the regression coefficient of the i-th gas component; then calculate the mean squared error loss function. Where N is the number of time points, =0.01; After updating the model weights using regression coefficients, a mapping relationship from independent variables to dependent variables is generated. Finally, based on the time series and the mapping relationship, the total number of moles of gas evolved from the battery at each time point is calculated, and an integration operation is performed. The total amount of gas produced by thermal runaway was obtained. .

[0094] It should be noted that during thermal runaway, gas concentration and pressure changes exhibit a dynamic coupling relationship. By establishing a multiple linear regression model, using real-time concentration data of each gas component as input variables and pressure change data as output variables, the synergistic effect of multiple gas components on pressure can be captured. Cross-validation was used to verify the model performance with different numbers of principal components by splitting the dataset, ensuring that the selected key variables accurately reflect the pressure change pattern. After introducing a ridge regression regularization term, the coefficient oscillations caused by the correlation between gas components under high temperature and high pressure conditions were suppressed by adding weight decay constraints. Gradient descent was used to train the model, which can adaptively adjust parameter weights and accurately fit the nonlinear relationship between concentration and pressure. The influence of the settling time boundary determination error on the calculation results was eliminated, and the problem of deviation in the total gas production calculation caused by the insufficient applicability of the ideal gas equation was solved, improving the real-time performance and accuracy of gas production detection.

[0095] Step 600: Based on the second law of thermodynamics, the total amount of gas produced during thermal runaway is quantified using the entropy increase formula to obtain the energy release characteristic matrix of the thermal runaway process; specific steps are as follows. Figure 2 As shown, it includes:

[0096] Step 601: Obtain the final mole fraction of each gas component at the end of thermal runaway based on the gas concentration distribution vector.

[0097] Step 602: Query the thermodynamic standard database to obtain the molar entropy of each gas component under standard conditions.

[0098] Step 603: Perform temperature and pressure corrections on the molar entropy of each gas component based on the pressure-temperature coupling matrix, expressed as:

[0099] ;

[0100] Where Cp is the constant pressure heat capacity. This represents the molar entropy under standard conditions.

[0101] Step 604: Based on the Gibbs formula for mixing entropy, calculate the total entropy of the mixed gas according to the corrected molar entropy. The calculation formula is as follows: .

[0102] Step 605: Calculate the difference between the total entropy value and the entropy value of the initial state to obtain the total entropy increase of the system;

[0103] Step 606: Combine the total entropy increase of the system, the peak gas production rate, and the total amount of gas produced during thermal runaway into a three-dimensional vector, which serves as the energy release characteristic matrix of the thermal runaway process.

[0104] It should be noted that after obtaining the mole fraction of each component through the gas concentration distribution vector, benchmark data is established by combining it with the standard entropy database. Then, the standard entropy value is dynamically corrected by using the real-time environmental parameters in the pressure and temperature coupling matrix. This accurately quantifies the energy release characteristics of the thermal runaway process and solves the problem of the total gas production calculation deviation caused by ignoring the entropy change of gas mixing and the fluctuation of environmental parameters in the existing technology.

[0105] Step 700: Map the energy release feature matrix to a high-dimensional safety assessment space to generate a battery thermal runaway risk index;

[0106] Specifically, firstly, a benchmark feature database containing safety level labels is constructed, which includes historical battery test data and their corresponding safety level labels. Then, covariance matrix factorization is used to extract the principal components with the largest variance in the dataset to eliminate redundant information between features and retain core safety influencing factors, thereby achieving data dimensionality reduction. The N principal components with the highest contribution rates are then used as basis vectors to construct an N-dimensional orthogonal safety assessment space. Next, matrix multiplication is used to transform the energy release feature matrix into the orthogonal safety assessment space, obtaining the coordinate vector of the matrix in the orthogonal safety assessment space. Then, the Euclidean distance between the coordinate vector and the preset absolute safety origin is calculated based on the square root of the sum of the squares of each component, quantifying the deviation of the current state from the ideal safety state. Finally, the maximum-minimum method is used to map the Euclidean distance to the 0-1 interval, obtaining the battery thermal runaway risk index.

[0107] It should be noted that by constructing the evaluation space, matrix mapping, and Euclidean distance normalization, a quantitative assessment of battery thermal runaway risk was achieved, eliminating assessment bias caused by subjective judgment. By using mathematical modeling to transform multidimensional energy characteristics into a single risk index, the efficiency of safety assessment was significantly improved.

[0108] Step 800: Based on the battery thermal runaway risk index, predict the thermal runaway probability distribution of the battery under test under different operating conditions using a support vector machine classifier.

[0109] Specifically, the battery thermal runaway risk index is used as the core input feature, and battery design parameters, including cell capacity, positive and negative electrode material systems, and electrolyte composition, are used as auxiliary input features. These two features are then concatenated into an enhanced feature vector. The radial basis function is then used as the kernel function of the support vector machine classifier, with the expression: The hyperparameters of the kernel function are optimized using a Bayesian optimization algorithm. Optimization training is performed. Finally, the enhanced feature vectors are input into the trained support vector machine classifier, and the final output is the thermal runaway probability distribution.

[0110] Furthermore, the support vector machine classifier first calculates the kernel space distance between the augmented feature vector and the support vector, then calculates the original classification score through the built-in decision function, and maps the original classification score of the decision function to the [0,1] interval through the Sigmoid function. At the same time, it uses the real labels of historical test data to perform logistic regression fitting and calibration. Finally, for various preset operating conditions, such as different ambient temperatures, charge and discharge rates and aging states, it outputs the corresponding thermal runaway probability values ​​in parallel, forming discrete operating condition-probability key-value pairs, thereby constructing a thermal runaway probability distribution map.

[0111] The present invention also provides a battery gas production state detection system after thermal runaway, comprising:

[0112] The thermal runaway triggering module is used to place the battery under test in a sealed explosion-proof container and trigger the thermal runaway process through external heating;

[0113] The pressure acquisition module is used to acquire pressure change data in the sealed explosion-proof container in real time during thermal runaway based on a dynamic pressure compensation model, and to construct a pressure-temperature coupling matrix by combining the temperature curves recorded by the temperature sensor.

[0114] The pressure correction module is used to correct the pressure and temperature coupling matrix through nonlinear thermodynamic equations to obtain the actual gas pressure distribution;

[0115] The gas analysis module is used to perform multidimensional analysis of the gas components in a sealed explosion-proof container using a Fourier transform infrared spectrometer, and to extract the gas concentration distribution vector.

[0116] The gas production calculation module is used to fit the gas concentration distribution vector with the actual gas pressure distribution through a multivariate regression analysis algorithm to obtain the total amount of gas produced during thermal runaway.

[0117] The energy assessment module is used to quantify the total amount of gas produced in thermal runaway based on the second law of thermodynamics and the entropy increase formula, so as to obtain the energy release characteristic matrix of the thermal runaway process.

[0118] The risk assessment module is used to map the energy release feature matrix to a high-dimensional safety assessment space to generate a battery thermal runaway risk index.

[0119] The probability prediction module is used to predict the probability distribution of thermal runaway of the battery under test under different operating conditions based on the battery thermal runaway risk index and a support vector machine classifier.

[0120] The beneficial effects of this invention are as follows:

[0121] 1) By using Kalman filtering and tank thermal expansion error correction, the pressure measurement distortion caused by the thermal expansion of the sealed explosion-proof tank itself is eliminated, and the error problem caused by container deformation in traditional methods is solved.

[0122] 2) The virial equation was used to correct the behavior of non-ideal gases, overcoming the limitations of the ideal gas law under high temperature and high pressure, and significantly improving the accuracy and precision of detecting gas pressure distribution.

[0123] 3) By using Fourier transform infrared spectroscopy (FTIR) to analyze the components of mixed gas online, and combining multivariate curve fitting and background spectrum subtraction techniques, the gas concentration distribution vector can be obtained, avoiding the test delay caused by static waiting in traditional methods.

[0124] 4) A dynamic mapping relationship between gas concentration and pressure change was established through multiple regression analysis, which solved the modeling problem of the synergistic effect of mixed gas components on pressure;

[0125] 5) Combined with integral calculation, the total gas production is directly calculated without relying on the stable state after settling, eliminating the measurement deviation caused by the boundary judgment error of settling time and gas leakage in the traditional method;

[0126] 6) Based on the second law of thermodynamics, the energy dissipation of the gas production process was quantified by the entropy increase formula, and an energy characteristic matrix containing entropy increase, peak gas production and total amount was generated, providing a physical basis for risk assessment.

[0127] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0128] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for detecting gas generation after battery thermal runaway, characterized in that, Includes the following steps: The battery under test was placed in a sealed explosion-proof container and thermal runaway was triggered by external heating. Based on the dynamic pressure compensation model, pressure change data inside the sealed explosion-proof container during thermal runaway is collected in real time, and a pressure-temperature coupling matrix is ​​constructed by combining the temperature curves recorded by the temperature sensor. The actual gas pressure distribution is obtained by correcting the pressure and temperature coupling matrix using nonlinear thermodynamic equations. The gas composition inside the sealed explosion-proof container was analyzed in multiple dimensions using a Fourier transform infrared spectrometer, and the gas concentration distribution vector was extracted. The total amount of gas produced by thermal runaway is obtained by fitting the gas concentration distribution vector with the actual gas pressure distribution using a multiple regression analysis algorithm. Based on the second law of thermodynamics, the total amount of gas produced during thermal runaway is quantified using the entropy increase formula, resulting in an energy release characteristic matrix of the thermal runaway process. The energy release feature matrix is ​​mapped to a high-dimensional safety assessment space to generate a battery thermal runaway risk index. Based on the battery thermal runaway risk index, the thermal runaway probability distribution of the battery under test under different operating conditions is predicted by a support vector machine classifier.

2. The method for detecting gas generation after battery thermal runaway according to claim 1, characterized in that, Based on a dynamic pressure compensation model, pressure change data within the sealed explosion-proof container during thermal runaway is collected in real time. A pressure-temperature coupling matrix is ​​constructed by combining this data with temperature curves recorded by temperature sensors, including: The original pressure signal sequence inside the sealed explosion-proof container is acquired using a high-frequency pressure sensor at a first sampling rate. Temperature signal sequences are synchronously acquired by a multi-point armored thermocouple array arranged inside the sealed explosion-proof container. The original pressure signal sequence and the temperature signal sequence are denoised and state estimated using the Kalman filter algorithm, respectively. Based on the material thermal expansion coefficient and geometric model of the sealed explosion-proof can, a thermally induced volume expansion error function of the can is constructed. The pressure signal sequence after noise reduction is compensated by the thermal volume expansion error function of the tank body to eliminate the pressure measurement error caused by the expansion of the sealed explosion-proof tank itself. The compensated pressure signal sequence and the denoised temperature signal sequence are aligned and interpolated on a shared timestamp reference to generate the pressure and temperature coupling matrix.

3. The method for detecting gas generation state after battery thermal runaway according to claim 1, characterized in that, The actual gas pressure distribution is obtained by correcting the pressure-temperature coupling matrix using nonlinear thermodynamic equations, including: The virial equation is selected as the nonlinear thermodynamic equation to characterize the high-temperature and high-pressure nonideal gas behavior during thermal runaway gas production. Based on the data points in the pressure and temperature coupling matrix, an overdetermined system of equations is constructed. The second and third virial coefficients of the virial equation were determined by Gibbs ensemble Monte Carlo simulation. The overdetermined system of equations is solved using the singular value decomposition algorithm to obtain the global optimal solution of the virial equations; The global optimal solution is substituted into the virial equation, and the equivalent pressure of the ideal gas at different time points is obtained through iterative calculation to form the actual gas pressure distribution.

4. The method for detecting gas generation after battery thermal runaway according to claim 1, characterized in that, The gas composition inside the sealed explosion-proof container was analyzed in multiple dimensions using a Fourier transform infrared spectrometer, and the gas concentration distribution vector was extracted, including: A mid-infrared detection beam is introduced into the sealed explosion-proof container through an optical window, and a transmission spectrum containing the absorption peaks of the gas-producing components is collected. The pre-collected inert gas background spectrum is removed from the transmission spectrum to obtain the net absorption spectrum data; Based on a standard gas database, a standard absorption spectral dictionary of potential gaseous products is constructed. The net absorption spectral data is decomposed into a linear superposition of the components in the standard absorption spectral dictionary using a multivariate curve fitting algorithm. The instantaneous molar concentration of each gas component is obtained by optimizing the superposition coefficient of the linear superposition using the least squares method. The instantaneous molar concentrations of all measured gas components are arranged in a time series to form the gas concentration distribution vector.

5. The method for detecting gas generation state after battery thermal runaway according to claim 1, characterized in that, The gas concentration distribution vector is fitted to the actual gas pressure distribution using a multiple regression analysis algorithm to obtain the total amount of gas produced during thermal runaway, including: A multiple linear regression model is constructed by taking the concentration of each component in the gas concentration distribution vector as the independent variable and the actual gas pressure distribution as the dependent variable. The optimal number of principal components for the multiple linear regression model was determined using cross-validation. A ridge regression regularization term is introduced into the aforementioned multiple linear regression model; The multiple linear regression model is trained using gradient descent, and a mapping relationship from the independent variable to the dependent variable is constructed. Based on the time series, the total number of moles of gas evolved from the battery at each time point is calculated according to the mapping relationship, and an integration operation is performed to obtain the total amount of gas generated by thermal runaway.

6. The method for detecting gas generation state after battery thermal runaway according to claim 1, characterized in that, Based on the second law of thermodynamics, the total amount of gas produced during thermal runaway is quantified using the entropy increase formula, resulting in an energy release characteristic matrix of the thermal runaway process, including: The final mole fraction of each gas component at the end of thermal runaway is obtained based on the gas concentration distribution vector. Query the thermodynamic standard database to obtain the molar entropy of each gas component under standard conditions; The molar entropy of each gas component is corrected for temperature and pressure based on the pressure and temperature coupling matrix. Based on the Gibbs formula for mixing entropy, the total entropy of the mixed gas is calculated according to the corrected molar entropy. The total entropy value is subtracted from the entropy value of the initial state to obtain the increase in the total entropy of the system. The total entropy increase of the system, the peak gas production rate, and the total amount of gas produced during thermal runaway are combined into a three-dimensional vector, which serves as the energy release characteristic matrix of the thermal runaway process.

7. The method for detecting gas generation after battery thermal runaway according to claim 1, characterized in that, The energy release feature matrix is ​​mapped to a high-dimensional safety assessment space to generate a battery thermal runaway risk index, including: A benchmark feature database containing safety level labels is constructed; the benchmark feature database is generated from historical battery test data. Principal component analysis is used to reduce the dimensionality of the benchmark feature database, and an N-dimensional orthogonal security assessment space is constructed based on the N principal components with the highest contribution rates extracted as basis vectors. Projecting the energy release feature matrix onto the orthogonal security assessment space yields a coordinate vector in the orthogonal security assessment space. Calculate the Euclidean distance between the coordinate vector and the preset absolute safety origin; The Euclidean distance is normalized to obtain the battery thermal runaway risk index.

8. The method for detecting gas generation state after battery thermal runaway according to claim 1, characterized in that, Based on the battery thermal runaway risk index, a support vector machine classifier is used to predict the thermal runaway probability distribution of the battery under test under different operating conditions, including: The battery thermal runaway risk index is used as the core input feature; Battery design parameters, including cell capacity, positive and negative electrode material systems, and electrolyte composition, are used as auxiliary input features. The core input features are concatenated with the auxiliary input features to obtain the enhanced feature vector; The radial basis function is used as the kernel function of the support vector machine classifier, and the hyperparameters of the kernel function are optimized and trained using the Bayesian optimization algorithm. The enhanced feature vector is input into the trained support vector machine classifier to obtain the thermal runaway probability distribution.

9. A system for detecting gas generation after battery thermal runaway, characterized in that, include: The thermal runaway triggering module is used to place the battery under test in a sealed explosion-proof container and trigger the thermal runaway process through external heating; The pressure acquisition module is used to acquire pressure change data inside the sealed explosion-proof container in real time during thermal runaway based on a dynamic pressure compensation model, and to construct a pressure-temperature coupling matrix by combining the temperature curves recorded by the temperature sensor. The pressure correction module is used to correct the pressure and temperature coupling matrix using a nonlinear thermodynamic equation to obtain the actual gas pressure distribution. The gas analysis module is used to perform multidimensional analysis of the gas components in the sealed explosion-proof container using a Fourier transform infrared spectrometer, and to extract the gas concentration distribution vector. The gas production calculation module is used to fit the gas concentration distribution vector with the actual gas pressure distribution through a multivariate regression analysis algorithm to obtain the total amount of gas produced during thermal runaway. The energy assessment module is used to quantify the total amount of gas produced by the thermal runaway based on the second law of thermodynamics and the entropy increase formula, so as to obtain the energy release characteristic matrix of the thermal runaway process. The risk assessment module is used to map the energy release feature matrix to a high-dimensional safety assessment space to generate a battery thermal runaway risk index. The probability prediction module is used to predict the probability distribution of thermal runaway of the battery under test under different operating conditions based on the battery thermal runaway risk index and using a support vector machine classifier.

Images