A method, medium and system for quickly determining a damaged or defective lithium battery

Abstract

The present invention provides a method, medium and system for quickly determining damaged or defective lithium batteries, belonging to the technical field of product detection. By collecting data such as temperature, voltage, current, etc. of the battery under different charge and discharge states, the present invention constructs multi-dimensional features such as a temperature change matrix, a voltage matrix, a power matrix, a resistance matrix, a cross-correlation index matrix, a harmonic change matrix and a reverse charge and discharge power matrix to comprehensively characterize the battery performance. Thermal imaging is used to obtain the surface temperature distribution, and a DC internal resistance tester and an AC impedance spectroscopy analyzer are used to measure the internal resistance, and Fourier transform is performed to analyze the voltage harmonic characteristics. Based on a deep neural network model with a multi-head attention mechanism, the collected multi-dimensional matrix data is analyzed to output the battery health state score and the damage type. This method comprehensively considers the electrochemical performance and thermodynamic characteristics of the battery, and realizes the rapid and accurate evaluation of the health state of lithium batteries.

Description

Technical Field

[0001] The present invention belongs to the technical field of product detection, and specifically, relates to a method, medium and system for quickly determining damaged or defective lithium batteries. Background Art

[0002] As the core power source of modern electronic devices and new energy vehicles, the quality and performance of lithium batteries are directly related to the safety and service life of the devices. Traditional lithium battery quality detection mainly relies on single-dimensional parameter evaluation methods such as direct current internal resistance testing, capacity testing, and cycle life testing. These methods usually require a long testing cycle and are difficult to comprehensively reflect the health status of the battery in actual production and application environments. For example, capacity testing requires a complete charge-discharge cycle, which usually takes several hours to several days, while internal resistance testing can only provide static parameters and cannot dynamically reflect the performance changes of the battery under working conditions.

[0003] Most of the existing detection methods adopt isolated parameter evaluation, lacking a comprehensive analysis of the internal correlation between multi-dimensional parameters of the battery, resulting in insufficient identification of some potential defects. For example, it is difficult to detect internal micro-short circuit problems only through voltage measurement, and it is impossible to accurately judge the degradation of electrochemical performance simply through temperature monitoring. Especially in a large-scale production environment, the low efficiency of traditional detection methods cannot meet the requirements of rapid screening and timely feedback, increasing production costs and potential safety risks.

[0004] Therefore, the existing technology faces the technical problem of quickly and accurately identifying damaged or defective lithium batteries, and there is an urgent need for a new method that can comprehensively consider multi-dimensional parameters and quickly evaluate the health status of the battery, so as to improve the detection efficiency, reduce safety risks, and provide reliable quality assurance for battery production and application. Summary of the Invention

[0005] In view of this, the present invention provides a method, medium and system for quickly determining damaged or defective lithium batteries, which can solve the technical problem of quickly and accurately identifying damaged or defective lithium batteries in the existing technology.

[0006] The present invention is implemented as follows: A method for quickly determining damaged or defective lithium batteries provided by the first aspect of the present invention includes: collecting temperature data, voltage data, and current data of lithium batteries under different charge and discharge states; constructing a temperature change matrix, a voltage matrix, a power matrix, a resistance matrix, a cross-correlation index matrix, a harmonic change matrix, and a reverse charge and discharge power matrix; using a DC internal resistance tester and an AC impedance spectroscopy analyzer to measure the internal resistance data of lithium batteries, recording impedance values at multiple frequency points, and constructing a resistance matrix to characterize the internal electrochemical performance state of lithium batteries; calculating the correlation coefficient between the temperature change and voltage change of lithium batteries based on the data of the voltage matrix and the temperature change matrix, and constructing a cross-correlation index matrix; inputting the data of the resistance matrix, the cross-correlation index matrix, the harmonic change matrix, and the reverse charge and discharge power matrix into a pre-trained lithium battery health state evaluation model, and determining whether the lithium battery is a damaged or defective product according to the output result of the lithium battery health state evaluation model.

[0007] Among them, the temperature change matrix specifically refers to a two-dimensional data structure representing the change of the surface temperature of a lithium battery over time. The rows of the temperature change matrix represent the positions of different measurement points on the battery surface, the columns of the temperature change matrix represent different time points, and the element values of the temperature change matrix are the temperature values corresponding to the positions and times.

[0008] Among them, the voltage matrix specifically refers to a two-dimensional data structure representing the change of the terminal voltage of a lithium battery under different charge and discharge states. The rows of the voltage matrix represent different charge and discharge rate conditions, the columns of the voltage matrix represent different time points, and the element values of the voltage matrix are the voltage values corresponding to the conditions and times.

[0009] Among them, the power matrix specifically refers to a two-dimensional data structure representing the change of the output power of a lithium battery over time, which is calculated by multiplying the voltage data and the current data. The rows of the power matrix represent different load conditions, the columns of the power matrix represent different time points, and the element values of the power matrix are the power values corresponding to the conditions and times.

[0010] Among them, the resistance matrix specifically refers to a two-dimensional data structure representing the internal resistance characteristics of a lithium battery. The rows of the resistance matrix represent different measurement frequencies, the columns of the resistance matrix represent different charge states, and the element values of the resistance matrix are the internal resistance values measured under the corresponding frequencies and charge states.

[0011] Among them, the cross-correlation index matrix is specifically a two-dimensional data structure that quantifies the correlation between the voltage change and temperature change of a lithium battery. The rows of the cross-correlation index matrix represent voltage measurement points, the columns represent temperature measurement points, and the element values of the cross-correlation index matrix are the Pearson correlation coefficients of the corresponding voltage points and temperature points; the harmonic variation matrix is specifically a two-dimensional data structure that characterizes the spectral characteristics of the voltage waveform of a lithium battery. The rows of the harmonic variation matrix represent different charge and discharge states, the columns represent different frequency components, and the element values of the harmonic variation matrix are the amplitudes of the frequency components in the corresponding states; the reverse charge and discharge power matrix is specifically a two-dimensional data structure that characterizes the energy conversion characteristics of a lithium battery under reverse charge and discharge conditions, obtained by performing mathematical transformations on the power matrix, and is used to evaluate the energy conversion efficiency of the lithium battery.

[0012] Among them, the specific structure of the lithium battery health state evaluation model is a deep neural network structure based on the multi-head attention mechanism, including an input layer, multiple attention layers, a feature fusion layer, and an output layer; among them, the parameters of the multi-head attention mechanism are determined according to three key parameters: the internal resistance frequency response characteristics of the resistance matrix, the thermal field distribution characteristics of the temperature change matrix, and the spectral distribution characteristics of the harmonic variation matrix.

[0013] Among them, the steps for establishing the training data set in the training process of the lithium battery health state evaluation model include collecting lithium battery samples of different models and different aging degrees, performing a standardized test process for each sample to collect multi-matrix feature data, performing dimensionless processing on the collected data, and determining the actual health state of each sample as the training label.

[0014] The second aspect of the present invention provides a computer-readable storage medium, in which program instructions are stored. When the program instructions run on a computer, they are used to execute the above-mentioned method for quickly determining a damaged or defective lithium battery.

[0015] The third aspect of the present invention provides a system for quickly determining a damaged or defective lithium battery, including the above-mentioned computer-readable storage medium. The system can be any one of a computer, a server, and a single-chip microcomputer. The computer-readable storage medium is set inside the system, and a microprocessor for executing the program instructions stored in the computer-readable storage medium is set inside the system.

[0016] The present invention comprehensively characterizes the electrochemical and thermodynamic properties of lithium batteries by constructing multi-dimensional features such as a temperature change matrix, a voltage matrix, a power matrix, a resistance matrix, a cross-correlation index matrix, a harmonic change matrix, and a reverse charge-discharge power matrix, realizing a rapid and accurate assessment of the battery health state. This method overcomes the limitations of traditional single-parameter assessment, and through the deep neural network structure of the multi-head attention mechanism, effectively captures the complex correlations and non-linear characteristics among the various parameters of lithium batteries. In particular, the relationship between voltage fluctuations and temperature changes is quantified through the cross-correlation index matrix, and the dynamic response characteristics of the battery are analyzed using the harmonic change matrix and the reverse charge-discharge power matrix, significantly improving the ability to identify abnormal batteries and reducing the misjudgment rate.

[0017] By integrating multi-dimensional data and advanced algorithms, the present invention realizes a rapid and accurate assessment of the health state of lithium batteries, solves the problems of low identification efficiency and insufficient accuracy of damaged or defective lithium batteries in the prior art, provides an efficient and reliable quality assessment means for the fields of battery manufacturing, application, and recycling, and effectively improves the battery production efficiency and use safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the method of the present invention.

[0019] Figure 2 It is a comparison diagram of voltage curves of different sample batteries during 0.5C discharge in Example 2.

[0020] Figure 3 It is the Nyquist diagram of the electrochemical impedance spectrum and the equivalent circuit model diagram in Example 2.

[0021] Figure 4 It is a classification evaluation result diagram of healthy and defective battery samples in Example 2. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] As Figure 1 shown, it is a flowchart of a method for quickly determining damaged or defective lithium batteries provided in the first aspect of the present invention. This method includes the following steps:

[0024] S01. Collect the temperature data of the lithium battery operating in different charge-discharge states, obtain the multi-point temperature distribution on the surface of the lithium battery through a thermal imager, establish a multi-time-section temperature change curve, and construct a temperature change matrix to record the change characteristics of the surface temperature field of the lithium battery over time;

[0025] S02. Collect voltage data of the lithium battery under different charge and discharge states, record the voltage change amplitude, frequency, and stability under different load conditions, and construct a voltage matrix to characterize the variation law of the lithium battery voltage characteristics over time;

[0026] S03. Based on the collected voltage data and current data, calculate the output power of the lithium battery under different working states, and construct a power matrix to characterize the variation law of the lithium battery power characteristics over time;

[0027] S04. Use a DC internal resistance tester and an AC impedance spectroscopy analyzer to measure the internal resistance data of the lithium battery, record the impedance values at multiple frequency points, construct a resistance matrix to characterize the internal electrochemical performance state of the lithium battery, and analyze the internal resistance change mechanism through an electrochemical impedance equation to establish an electrochemical equivalent circuit model considering the electrode polarization effect;

[0028] S05. Based on the voltage matrix and the temperature change matrix data, calculate the correlation coefficient between the temperature change and the voltage change of the lithium battery, and construct a cross-correlation index matrix to quantify the influence degree of voltage fluctuation on temperature change;

[0029] S06. Perform Fourier transform on the voltage data, analyze the amplitude distribution of different frequency components, and construct a harmonic variation matrix to characterize the voltage harmonic characteristics of the lithium battery;

[0030] S07. Perform inverse transformation on the power matrix to construct an inverse charge and discharge power matrix to characterize the energy conversion efficiency of the lithium battery during the inverse charge and discharge process;

[0031] S08. Input the resistance matrix, the cross-correlation index matrix, the harmonic variation matrix, and the inverse charge and discharge power matrix data of the lithium battery to be tested into a pre-trained lithium battery state of health assessment model, and judge whether the lithium battery is damaged or defective according to the output result of the lithium battery state of health assessment model, and give the corresponding state of health level assessment result;

[0032] Among them, the temperature change matrix specifically refers to a two-dimensional data structure characterizing the change of the lithium battery surface temperature over time. The rows of the temperature change matrix represent the positions of different measurement points on the battery surface, the columns of the temperature change matrix represent different time points, and the element values of the temperature change matrix are the temperature values corresponding to the positions and times;

[0033] Among them, the voltage matrix specifically refers to a two-dimensional data structure characterizing the terminal voltage change of the lithium battery under different charge and discharge states. The rows of the voltage matrix represent different charge and discharge rate conditions, the columns of the voltage matrix represent different time points, and the element values of the voltage matrix are the voltage values corresponding to the conditions and times;

[0034] Among them, the power matrix specifically refers to a two-dimensional data structure representing the output power of the lithium battery changing with time, which is calculated by multiplying the voltage data and the current data. The rows of the power matrix represent different load conditions, the columns of the power matrix represent different time points, and the element values of the power matrix are the power values corresponding to the conditions and time.

[0035] Among them, the resistance matrix specifically refers to a two-dimensional data structure representing the internal resistance characteristics of the lithium battery. The rows of the resistance matrix represent different measurement frequencies, the columns of the resistance matrix represent different charging states, and the element values of the resistance matrix are the internal resistance values measured under the corresponding frequencies and charging states.

[0036] Among them, the cross-correlation index matrix specifically refers to a two-dimensional data structure quantifying the correlation between the voltage change and the temperature change of the lithium battery. The rows of the cross-correlation index matrix represent voltage measurement points, the columns of the cross-correlation index matrix represent temperature measurement points, and the element values of the cross-correlation index matrix are the Pearson correlation coefficients corresponding to the voltage points and temperature points.

[0037] Among them, the harmonic variation matrix specifically refers to a two-dimensional data structure representing the spectral characteristics of the voltage waveform of the lithium battery. The rows of the harmonic variation matrix represent different charge and discharge states, the columns of the harmonic variation matrix represent different frequency components, and the element values of the harmonic variation matrix are the amplitudes of the frequency components under the corresponding states.

[0038] Among them, the reverse charge and discharge power matrix specifically refers to a two-dimensional data structure representing the energy conversion characteristics of the lithium battery under reverse charge and discharge conditions, which is obtained by performing a mathematical transformation on the power matrix and is used to evaluate the energy conversion efficiency of the lithium battery.

[0039] The electrochemical impedance equation is used to describe the internal electrode polarization process and charge transfer impedance characteristics of the lithium battery, analyze the variation laws of the internal Ohmic impedance, charge transfer impedance and diffusion impedance of the battery with frequency. The inputs include the different measurement frequencies, the electrode potential, the electrolyte concentration, the temperature data and the electrode active area, and the outputs are the impedance values in complex form and the corresponding equivalent circuit parameters.

[0040] The specific structure of the lithium battery health state evaluation model is a deep neural network structure based on the multi-head attention mechanism, including the input layer, the multiple attention layers, the feature fusion layer, and the output layer. The input layer receives the feature vectors of the resistance matrix, the cross-correlation index matrix, the harmonic variation matrix, and the reverse charge-discharge power matrix. The multi-head attention layer focuses on the information of different feature dimensions and assigns weights. The feature fusion layer integrates the outputs of the multi-head attention layer to form a comprehensive feature representation of the battery health state. The output layer maps the comprehensive feature to the health state score and the damage type classification result through the activation function. The parameters of the multi-head attention mechanism in the lithium battery health state evaluation model are determined according to three key parameters: the internal resistance frequency response characteristics of the resistance matrix, the thermal field distribution characteristics of the temperature variation matrix, and the spectral distribution characteristics of the harmonic variation matrix;

[0041] The steps for establishing the training dataset during the training process of the lithium battery health state evaluation model specifically include collecting a large number of lithium battery samples of different models and different aging degrees, performing a standardized test process for each sample to collect data of the resistance matrix, the temperature variation matrix, the voltage matrix, the power matrix, the cross-correlation index matrix, the harmonic variation matrix, and the reverse charge-discharge power matrix, performing dimensionless processing on the collected data to eliminate the dimensional differences between different physical quantities, determining the actual health state of each sample as the training label through a test device, forming the training data pairs by combining the processed multi-matrix feature data with the corresponding health state labels, and finally dividing the dataset into the training set and the validation set according to the ratio of 8:2;

[0042] The steps for training the lithium battery health state evaluation model specifically include initializing the model parameters, training the lithium battery health state evaluation model using the mini-batch stochastic gradient descent method, dynamically adjusting the learning rate during the training process to avoid local optimal solutions, calculating the error between the prediction result of the lithium battery health state evaluation model and the true label through the cross-entropy loss function, optimizing the model parameters using the backpropagation algorithm, evaluating the performance of the lithium battery health state evaluation model on the validation set and recording the key metrics, terminating the training process using the early stopping strategy when the performance metrics on the validation set do not show significant improvement for multiple consecutive cycles, and finally saving the trained model parameters as the final model for subsequent actual lithium battery health state evaluation tasks;

[0043] The electrode potential refers to the electrochemical potential of the lithium battery electrode material in the electrolyte environment, which is obtained by measuring with a reference electrode; the electrolyte concentration refers to the concentration of the lithium salt in the internal electrolyte of the lithium battery, with the unit of mol / L; the electrode active area refers to the effective surface area of the lithium battery electrode material participating in the electrochemical reaction, with the unit of ; the current data refers to the current values of the lithium battery under different working conditions, which are collected by a current sensor; the multi-matrix feature data refers to the feature set extracted based on the resistance matrix, temperature change matrix, voltage matrix, power matrix, cross-correlation index matrix, harmonic change matrix and reverse charge-discharge power matrix; the state of health score refers to the numerical value that quantitatively represents the health degree of the lithium battery, with the range of 0 to 100; the damage type classification result refers to the classification output of the lithium battery fault or degradation type.

[0044] Among them, the multi-dimensional feature optimization function is used to comprehensively consider the contribution weights of various characteristic parameters of the lithium battery to the battery state of health, and adjusts the model parameters through an iterative optimization algorithm to improve the discrimination accuracy. The input includes the resistance matrix eigenvalue, the cross-correlation index matrix eigenvalue, the harmonic change matrix frequency spectrum distribution feature and the reverse charge-discharge power matrix efficiency parameter, and the output is the optimized model parameter set and the respective feature weight coefficients; the respective feature weight coefficients refer to the contribution degree coefficients of different features to the judgment of the lithium battery state of health; the internal resistance frequency response characteristic refers to the curve characteristic of the lithium battery internal resistance changing with frequency; the thermal field distribution characteristic refers to the spatial characteristic of the lithium battery surface temperature distribution; the frequency spectrum distribution characteristic refers to the energy distribution characteristic of the lithium battery voltage waveform in the frequency domain.

[0045] The specific implementation manners of the above steps are described in detail below.

[0046] The specific implementation manner of step S01 is to perform high-precision multi-point monitoring on the lithium battery surface temperature distribution. First, use an infrared thermal imager with at least 0.05°C temperature resolution, and evenly arrange no less than 9 temperature acquisition points on the lithium battery surface, including 3 monitoring points in the positive electrode area, the negative electrode area and the middle area respectively. During the acquisition process, the lithium battery is placed in a constant temperature environment (25±2°C), and a complete charge-discharge cycle test is carried out according to three charging rates of 0.5C, 1C, 2C and three discharging rates of 0.5C, 1C, 2C. Record the temperature data of all monitoring points every 2 seconds, and continuously record at least 3 complete charge-discharge cycles. Eliminate the random noise of the collected original temperature data through the moving average filtering algorithm, and then use the bicubic spline interpolation method to generate a continuous temperature change curve. Construct a temperature change matrix based on the processed temperature data , where the rows of the matrix represent different monitoring points (points 1 to 9), and the columns represent the sampling time points ( to ), and the matrix element value represents the The temperature value of a monitoring point at moment. The function of this step is to capture possible local abnormal heating points inside the lithium battery through the spatio-temporal distribution characteristics of the temperature field. The temperature distribution of a healthy lithium battery should be relatively uniform, the abnormal temperature rise should not exceed 5°C, and the temperature difference between monitoring points is usually less than 3°C.

[0047] The specific implementation of step S02 is to monitor the terminal voltage of the lithium battery under different charge and discharge conditions using a high-precision voltage acquisition system (with an accuracy of not less than 1 mV). First, in the constant current charging mode, the lithium battery is charged at three charging rates of 0.2C, 0.5C, and 1C respectively, and the voltage value is recorded every 0.5 seconds until the battery voltage reaches the cut-off voltage; then, in the constant current discharge mode, the lithium battery is discharged at four discharge rates of 0.2C, 0.5C, 1C, and 2C respectively, and the voltage value is also recorded every 0.5 seconds until the battery voltage drops to the discharge cut-off voltage. The recorded voltage data is filtered using a Butterworth low-pass filter, and the cut-off frequency is set to 10 Hz to remove high-frequency interference components. According to the processed voltage data, the voltage change rate ( ), and the voltage stability index are calculated. The voltage stability index is characterized by calculating the standard deviation of the voltage data, and a stability index lower than 10 mV is regarded as a stable state. Finally, a voltage matrix V is constructed, where the rows of the matrix represent different charge and discharge rates (a total of 7 conditions from 0.2C charging to 2C discharging), and the columns represent the sampling time points ( to ), and the matrix element value represents the voltage value at the th rate condition at moment. The purpose of this step is to reflect the electrochemical performance state of the lithium battery through voltage characteristics. The voltage curve of a healthy lithium battery should have good repeatability and stability under the same conditions, and the voltage fluctuation amplitude does not exceed 2% of the charge and discharge platform voltage.

[0048] The specific implementation of step S03 is to calculate the real-time output power of the lithium battery under different working states based on the aforementioned collected voltage data and synchronously recorded current data. The current data is collected through a Hall effect current sensor (with an accuracy better than 0.5%), and the sampling frequency is the same as that of the voltage data collection (2 Hz). For each sampling time point, the instantaneous power value is calculated by multiplying the voltage and current: , where is the voltage value at the th condition at moment, is the current value at the corresponding moment. At the same time, calculate the energy conversion efficiency under each working condition. The charging energy efficiency is defined as the ratio of the actual stored energy to the input energy, and the discharging energy efficiency is defined as the ratio of the output energy to the theoretical stored energy. Construct a power matrix according to the calculation results , where the rows of the matrix represent different load conditions (consistent with the voltage matrix), and the columns represent the sampling time points ( to ), and the matrix element value represents the power value at the th load condition moment. This step aims to evaluate the energy conversion performance of the lithium battery through the power characteristics. The energy conversion efficiency of a healthy lithium battery should generally be higher than 85%, and the power curve should maintain a relatively stable morphological characteristic under different load conditions.

[0049] The specific implementation of step S04 is to comprehensively measure the internal resistance characteristics of the lithium battery using a high-precision DC internal resistance tester and an electrochemical workstation. First, use the DC internal resistance tester to measure the DC internal resistance values of the lithium battery under different state of charge (SOC = 0%, 20%, 40%, 60%, 80%, 100%), and control the measurement current within the range of 0.1C to 0.2C. The measurement interval is not less than 5 minutes each time, and each SOC point is measured 3 times and the average value is taken. Then use the electrochemical workstation to conduct electrochemical impedance spectroscopy (EIS) tests. The frequency scanning range is from 0.01 Hz to 100 kHz. Select no less than 30 frequency points at equal intervals in the logarithmic coordinate, and control the amplitude of the excitation signal within 10 mV to ensure the linear response of the system. Establish an electrochemical equivalent circuit model based on the measurement data. This model includes parameters such as ohmic internal resistance Rs, charge transfer resistance , double-layer capacitance and Warburg impedance , etc. Describe the internal impedance characteristics of the battery through the electrochemical impedance equation , where is the angular frequency. Construct a resistance matrix R according to the measured data. The rows of the matrix represent different measurement frequency points ( to ), and the columns represent different SOC states (0% to 100%). The matrix element value represents the complex impedance value measured at the frequency state. The purpose of this step is to reveal the internal structure and electrochemical reaction state of the lithium battery through the internal resistance characteristics. The internal resistance value of a healthy lithium battery is low and stable. The typical value does not exceed 50 mΩ in a 0.5C capacity lithium battery, and the frequency characteristic curve should present a typical semicircle plus oblique line shape.

[0050] The specific implementation of step S05 is to analyze the correlation between the voltage change and temperature change of the lithium battery based on the previously constructed voltage matrix and temperature change matrix. First, the voltage data and temperature data are aligned in time and normalized, and all data are converted into the interval [-1, 1] to eliminate the dimension difference. Then, the data in the key time windows (such as the rapid charge and discharge stage, the constant voltage charging conversion point, etc.) are selected, and the Pearson correlation coefficient between each voltage measurement point and each temperature measurement point is calculated: ; where is the value of the th voltage measurement point at moment, is the value of the th temperature measurement point at moment, and are the average values of the corresponding measurement points respectively. Based on the calculation results, a cross-correlation index matrix is constructed. The rows of the matrix represent the voltage measurement points, the columns represent the temperature measurement points, and the matrix element value represents the correlation coefficient between the th voltage point and the th temperature point. At the same time, the overall correlation index is calculated to characterize the global voltage-temperature coupling characteristics of the battery. The function of this step is to quantify the intensity of the voltage-temperature coupling effect. The voltage-temperature correlation coefficient of a healthy lithium battery should usually be kept above 0.7, and the correlation characteristics in different regions of the same battery should be consistent, and the difference in the correlation coefficient should not exceed 0.15.

[0051] The specific implementation of step S06 is to perform frequency-domain analysis on the collected voltage data to evaluate the harmonic characteristics of the voltage waveform. First, the voltage time-series data in each charge and discharge state are segmented, with 4096 data points taken for each segment, and the overlap rate between adjacent segments is 50%. The Hanning window function is applied to each segment of data to reduce spectral leakage, and then the fast Fourier transform (FFT) is performed to convert the time-domain signal into a frequency-domain representation: ; where is the time-domain voltage data, is the corresponding frequency-domain representation, and N is the number of data points. The power spectral density (PSD) is calculated, and the amplitudes of different frequency components are extracted. The low-frequency harmonic components in the frequency range of 0.1 Hz to 10 Hz are focused on. This frequency band is usually related to the internal electrochemical reaction and material aging characteristics of the battery. According to the analysis results, a harmonic variation matrix is constructed. The rows of the matrix represent different charge and discharge states (consistent with the voltage matrix), the columns represent different frequency components ( to ), and the matrix element value represents the value at the th charge and discharge state The amplitude of the frequency component. This step aims to reveal the characteristics of the internal dynamic process of the lithium battery through spectral features. The voltage spectrum of a healthy lithium battery should exhibit the characteristic that the amplitude decays smoothly as the frequency increases, and the amplitude of the harmonic components other than the main frequency component should not exceed 5% of the main frequency component.

[0052] The specific implementation of step S07 is to perform an inverse transformation on the aforementioned power matrix to evaluate the energy conversion efficiency of the lithium battery under reverse charge and discharge conditions. The specific implementation method is to first separate the charge power data and discharge power data in the power matrix P, and then reorder the charge power data in reverse time sequence to generate a reverse charge power curve; the discharge power data is also reordered in reverse time sequence to generate a reverse discharge power curve. Calculate the cumulative energy conversion efficiency during the reverse charge and discharge process: ; where and are the discharge power and charge power functions respectively. Based on the calculation results, construct a reverse charge and discharge power matrix R. The rows of the matrix represent different load conditions (the same as the power matrix), and the columns represent the sampling time points ( to ), and the matrix element value represents the power value at the th moment during the reverse charge and discharge process under the th load condition. The purpose of this step is to evaluate the energy storage reversibility of the lithium battery through the reverse energy conversion characteristics. The reverse energy conversion efficiency of a healthy lithium battery should be no less than 80%, and the efficiency fluctuation under different load conditions should not exceed 5%.

[0053] The specific implementation of step S08 is to input the above matrix data into a pre-trained lithium battery health state evaluation model to achieve a comprehensive evaluation of the lithium battery state. First, feature extraction and dimensionality reduction are performed on the resistance matrix, cross-correlation index matrix, harmonic matrix, and reverse charge-discharge power matrix. The principal component analysis (PCA) algorithm is used to retain the main feature dimensions with an explained variance ratio of not less than 95%. Then, the dimensionality-reduced feature vectors are normalized and input into a pre-trained deep neural network model, which can adaptively focus on the important information of different feature dimensions based on the multi-head attention mechanism. The forward calculation process of the model includes feature embedding, multi-head self-attention calculation, feature fusion, and classification / regression output. The final model output includes a health state score (0-100 points) and a damage type classification result. The scoring thresholds are set as follows: 90-100 points represent a healthy state, 75-90 points represent a mild aging state, 60-75 points represent a moderate aging state, 40-60 points represent a severe aging state, and below 40 points represent a damaged state. The output damage types also include various categories such as capacity attenuation type, internal resistance increase type, poor contact type, and reduced electrochemical activity type. The purpose of this step is to achieve an accurate quantitative evaluation of the lithium battery health state, combine multi-dimensional features to achieve high-precision discrimination of damaged or defective products, and the judgment accuracy should reach more than 95% on the test dataset.

[0054] The detailed structure of the lithium battery health state evaluation model specifically includes 4 main parts: an input layer, a multi-head attention encoding layer, a feature fusion layer, and an output layer. The input layer receives the feature vectors of the resistance matrix, cross-correlation index matrix, harmonic matrix, and reverse charge-discharge power matrix after preprocessing. The input dimension is determined according to the dimension after feature extraction, and a typical value is 256 dimensions. The multi-head attention encoding layer contains 3 parallel attention heads. Each attention head independently calculates the query vector (Query), key vector (Key), and value vector (Value), and then calculates the attention weights and weighted aggregates the feature information through the scaled dot-product attention mechanism. The dimension of each attention head is 64, and the total calculation formula is: ; where is the dimension of the attention head. The feature fusion layer adopts a two-layer fully connected network structure. The first layer contains 128 neurons and uses the ReLU activation function; the second layer contains 64 neurons and also uses the ReLU activation function. The output layer is divided into two branches: the health state scoring branch uses a fully connected layer and the Sigmoid activation function to output a score value in the range of 0-100; the damage type classification branch uses a fully connected layer and the Softmax activation function to output the probability distribution of various damage types. During the model training process, the cross-entropy loss function is used to measure the classification accuracy, and the mean square error loss function is used to measure the scoring accuracy. The two losses are combined in a ratio of 6:4 as the total loss function to guide the model optimization.

[0055] The detailed steps for establishing the training dataset of the lithium battery health state assessment model are as follows. First, collect samples. Obtain at least 500 lithium battery samples of different models and different aging degrees from different manufacturers, including brand-new batteries, naturally aged batteries, accelerated-aged batteries, and known faulty batteries, ensuring a balanced distribution of sample categories. Perform a standardized test process on each sample. Collect data of the resistance matrix, temperature change matrix, voltage matrix, power matrix, cross-correlation index matrix, harmonic change matrix, and reverse charge-discharge power matrix at a constant temperature environment (25 ± 2 °C) according to the aforementioned method. Use professional lithium battery detection equipment (such as the Neware battery comprehensive test system) to determine the actual health state of each sample, including key indicators such as capacity retention rate, internal resistance growth rate, cycle life prediction, etc., as training labels. The data preprocessing link includes outlier detection and processing (identifying outliers using the 3σ criterion), data normalization (scaling all features to the interval [0, 1] using the Min-Max method), and data augmentation (expanding training samples by adding Gaussian noise, random rotation, etc.). Finally, randomly divide the dataset into a training set and a validation set according to a ratio of 8:2, ensuring that the proportions of various samples in the two sets are consistent. The model training uses the mini-batch stochastic gradient descent method, with the batch size set to 32 and the initial learning rate set to 0.001. The cosine annealing strategy is used to dynamically adjust the learning rate. The early stopping strategy is used during the training process. When the performance indicators on the validation set do not improve significantly for 10 consecutive epochs, the training is terminated. The accuracy of the final model on the independent test set reaches 96.5%, and the F1 score reaches 0.94, meeting the requirements for accurately determining damaged or defective lithium batteries.

[0056] In the second aspect of the present invention, a computer-readable storage medium is provided. Program instructions are stored in the computer-readable storage medium, and when the program instructions run on a computer, they are used to execute the above-mentioned method for quickly determining damaged or defective lithium batteries.

[0057] In the third aspect of the present invention, a system for quickly determining damaged or defective lithium batteries is provided, including the above-mentioned computer-readable storage medium. The system can be any one of a computer, a server, and a single-chip microcomputer. The computer-readable storage medium is set inside the system, and a microprocessor for executing the program instructions stored in the computer-readable storage medium is set inside the system.

[0058] The following describes in detail the mathematical models or calculation processes involved in the present invention.

[0059] In step S01, the construction of the temperature change matrix involves collecting and processing the surface temperature data of the lithium battery. The specific representation of the temperature change matrix is as follows:

[0060] ;

[0061] In the formula, is a temperature change matrix; represents the th measurement point at time, with the unit of °C; is the number of temperature measurement points, usually not less than 9; is the number of time sampling points.

[0062] The formula of the moving average filtering algorithm used in the temperature data processing is as follows:

[0063] ;

[0064] In the formula, is the temperature value of the th measurement point at time after filtering; is the half-width of the filtering window, and the typical value is 2; is the original temperature data.

[0065] This filtering algorithm is based on the principle of local data averaging. By calculating the arithmetic mean value of the data within the window, the influence of random noise is reduced. The window size determines the smoothing degree. A larger window provides a stronger smoothing effect but may cause loss of useful information.

[0066] In step S02, the construction of the voltage matrix involves recording voltage data under different charge and discharge conditions. The specific representation of the voltage matrix is as follows:

[0067] ;

[0068] In the formula, is the voltage matrix; represents the th charge and discharge rate condition at time, with the unit of V; is the number of charge and discharge rate conditions, usually 7; is the number of time sampling points.

[0069] The calculation formula of the voltage stability index is as follows:

[0070] ;

[0071] In the formula, is the voltage stability index under the th charge and discharge rate condition, with the unit of V; is the average voltage value under the th charge and discharge rate condition, .

[0072] This index quantifies the voltage fluctuation degree by calculating the standard deviation of voltage data, reflecting the voltage stability of the lithium battery. For a healthy battery, the value is usually less than 10 mV, and a larger value implies that there may be abnormalities inside the battery.

[0073] In step S03, the power matrix is calculated by multiplying the voltage data and the current data. The specific representation of the power matrix is as follows:

[0074] ;

[0075] In the formula, is the power matrix; represents the power value at the th load condition at the moment, with the unit of W; is the corresponding voltage value; is the corresponding current value; is the number of load conditions; is the number of time sampling points.

[0076] The calculation formula of the energy conversion efficiency is as follows:

[0077] ;

[0078] ;

[0079] In the formula, is the charging energy efficiency; is the discharging energy efficiency; is the stored energy; is the input energy; is the output energy; and are the voltage and current functions during the charging process respectively; and are the voltage and current functions during the discharging process respectively; and are the total times of charging and discharging respectively.

[0080] These efficiency calculations are based on the principle of energy conservation, and the input-output ratio during the energy conversion process is calculated through integration, reflecting the energy conversion performance of the lithium battery. For a healthy battery, and should usually be greater than 85%.

[0081] In step S04, an electrochemical equivalent circuit model is established and a resistance matrix is constructed. The specific representation of the electrochemical impedance equation is as follows:

[0082] ;

[0083] In the formula, is the impedance value in complex form, with the unit of Ω; is the imaginary unit; is the angular frequency, with the unit of rad / s; is the frequency, with the unit of Hz; is the internal resistance of Ohm, representing the electronic conduction impedance of the electrolyte and electrode materials, with the unit of Ω; is the charge transfer resistance, representing the impedance of the charge transfer process at the electrode - electrolyte interface, with the unit of Ω; is the double - layer capacitance, representing the charge accumulation ability at the electrode - electrolyte interface, with the unit of F; is the Warburg impedance coefficient, characterizing the diffusion process of ions in the electrode material, with the unit of .

[0084] This equation is based on the theory of electrochemical impedance spectroscopy, comprehensively considering three main processes: Ohmic conduction, charge transfer, and ion diffusion inside the battery. The first term represents the pure resistance characteristic; the second term represents the parallel RC - circuit characteristic of the interface charge transfer process; the third term represents the Warburg impedance characteristic of the finite diffusion process. For a healthy battery, the and values are lower, the value is higher, and the value is moderate.

[0085] The specific representation of the resistance matrix is as follows:

[0086] ;

[0087] In the formula, is the resistance matrix; represents the complex impedance value measured at the SOC state of frequency, which is in complex form ; ; and are the real and imaginary parts of the impedance respectively; is the number of frequency points, not less than 30; is the number of SOC states, usually 6 (0%, 20%, 40%, 60%, 80%, 100%).

[0088] In step S05, calculate the correlation between the lithium - battery voltage change and the temperature change, and construct a cross - correlation index matrix. The calculation formula of the Pearson correlation coefficient is as follows:

[0089] ;

[0090] In the formula, is the Pearson correlation coefficient between the th voltage measurement point and the th temperature measurement point; is the normalized voltage value of the th voltage measurement point at moment; is the normalized temperature value of the th temperature measurement point at moment; and are the average values of the corresponding measurement points respectively; is the number of time sampling points.

[0091] This coefficient is based on the statistical correlation theory and quantifies the linear correlation degree between variables by calculating the ratio of the product of the covariance and standard deviation of two standardized variables. The value range is [-1, 1]. The closer the value is to ±1, the stronger the correlation. A positive value indicates a positive correlation, and a negative value indicates a negative correlation. The correlation coefficient between the voltage and temperature of a healthy battery is usually greater than 0.7, indicating a strong synchronization between the voltage and temperature changes during the charge and discharge process.

[0092] The specific representation of the cross-correlation index matrix is as follows:

[0093] ;

[0094] In the formula, is the cross-correlation index matrix; is the Pearson correlation coefficient between the th voltage measurement point and the th temperature measurement point; is the number of voltage measurement points; is the number of temperature measurement points, usually not less than 9.

[0095] In step S06, Fourier transform is performed on the voltage data to construct a harmonic matrix. The calculation formula of the fast Fourier transform is as follows:

[0096] , ;

[0097] In the formula, is the voltage data in the frequency domain; is the voltage data in the time domain; is the number of data points, taking 4096; is the imaginary unit; is the frequency index; is the time index.

[0098] This transformation is based on the theory of Fourier analysis, which decomposes a time-domain signal into a superposition of sine and cosine waves of different frequencies, reflecting the spectral characteristics of the signal. The FFT algorithm reduces the computational complexity through a divide-and-conquer strategy and is suitable for large-scale data processing.

[0099] The calculation formula for the power spectral density is as follows:

[0100] ;

[0101] In the formula, is the power spectral density at the th frequency point; and are the real and imaginary parts of respectively; is the number of data points.

[0102] The specific representation of the harmonic variation matrix is as follows:

[0103] ;

[0104] In the formula, is the harmonic variation matrix; represents the amplitude of the th frequency component in the th charge-discharge state; is the number of charge-discharge states; is the number of frequency components.

[0105] In step S07, calculate the cumulative energy conversion efficiency during the reverse charge-discharge process and construct the reverse charge-discharge power matrix. The calculation formula for the cumulative energy conversion efficiency is as follows:

[0106] ;

[0107] In the formula, is the cumulative energy conversion efficiency during the reverse charge-discharge process, expressed as a percentage; is the discharge power function; is the charge power function; is the total charge-discharge time.

[0108] This efficiency calculation evaluates the reversibility of battery energy storage by comparing the ratio of the energy released during discharge to the energy input during charge. The of a healthy battery is usually not less than 80%, and a lower value indicates irreversible reactions or damage inside the battery.

[0109] The specific representation of the reverse charge-discharge power matrix is as follows:

[0110] ;

[0111] In the formula, is the reverse charge-discharge power matrix; represents the th load condition during the reverse charge-discharge process at time , with the unit of W; is the number of load conditions; is the number of time sampling points.

[0112] The methods for obtaining the parameters used in the lithium battery state of health assessment model are as follows:

[0113] , , and parameters are obtained through electrochemical impedance spectroscopy experiments. The specific steps include: (1) Using an electrochemical workstation to perform EIS tests on the lithium battery, with a frequency range of 0.01 Hz to 100 kHz; (2) Using an equivalent circuit model to perform complex nonlinear least squares fitting on the measured impedance data to obtain the parameter values of each equivalent circuit element. The typical value range: is 5 - 50 mΩ, is 10 - 100 mΩ, is 0.1 - 10 F, is 0.05 - 0.5 .

[0114] parameters are obtained through an infrared thermal imager experiment. The specific steps include: (1) Using an infrared thermal imager with a resolution of 0.05 °C to image the surface of the lithium battery to obtain a temperature distribution map; (2) Uniformly selecting no less than 9 measurement points on the battery surface; (3) Recording the temperature values of each point every 2 seconds during the charge-discharge process. The typical value range: 25 - 45 °C.

[0115] parameters are obtained through a voltage acquisition system experiment. The specific steps include: (1) Connecting the lithium battery with a voltage acquisition system with an accuracy of not less than 1 mV; (2) Recording the voltage value every 0.5 seconds under different charge-discharge rate conditions. The typical value range: 3.0 - 4.2 V (applicable to lithium-ion batteries).

[0116] parameters are obtained through a current sensor experiment. The specific steps include: (1) Connecting the charge-discharge circuit of the lithium battery with a Hall effect current sensor with an accuracy better than 0.5%; (2) Recording the current value every 0.5 seconds under different charge-discharge rate conditions. The typical value range: 0.1 - 10 A (depending on the battery capacity and charge-discharge rate).

[0117] The health status score threshold is determined based on the statistics of a large number of lithium battery samples: a score of 90 - 100 indicates a healthy state, 75 - 90 indicates a mildly aged state, 60 - 75 indicates a moderately aged state, 40 - 60 indicates a severely aged state, and below 40 indicates a damaged state. This score reflects the degree to which the performance of the lithium battery is maintained relative to a new battery and is a comprehensive quantitative indicator of the health status.

[0118] The above equations and calculation processes together constitute a method system for comprehensively evaluating the health status of lithium batteries. By combining multi-dimensional analysis of electrochemical characteristics, thermal characteristics, and electrical characteristics, rapid and accurate determination of damaged or defective lithium batteries can be achieved.

[0119] Optionally, for the electrochemical equivalent circuit model mentioned in step S04, the relationship between the model parameters and the battery characteristics is detailed as follows:

[0120] ;

[0121] In the formula, is the ohmic internal resistance, with the unit of Ω; is the temperature, with the unit of °C; is the electrolyte concentration, with the unit of mol / L; is the electrode contact area, with the unit of ; is the distance between electrodes, with the unit of m; is the electrolyte conductivity, which is related to temperature and concentration, with the unit of S / m; is the correction factor; is the error term.

[0122] ;

[0123] In the formula, is the charge transfer resistance, with the unit of Ω; is the gas constant, taking 8.314 J / (mol·K); is the absolute temperature, with the unit of K; is the number of electrons participating in the reaction; is the Faraday constant, taking 96485 C / mol; is the exchange current density, with the unit of A / ; is the electrode active area, with the unit of ; is the error term.

[0124] ;

[0125] In the formula, is the Warburg impedance coefficient, with the unit of ; is the diffusion coefficient of ions in the electrode material, with the unit of / s; is the reactant concentration, with the unit of mol / ; is the active area of the electrode, with the unit of ; is the error term.

[0126] Optionally, for the multi-dimensional feature optimization function mentioned in step S08, it is specifically expressed as follows:

[0127] ;

[0128] In the formula, is the multi-dimensional feature optimization function; is the set of model parameters; is the number of training samples; is the cross-entropy loss function; is the mean squared error loss function; and are respectively the true class label and the predicted class of the th sample; and are respectively the true health score and the predicted score of the th sample; and are the loss weight coefficients, usually ; is the regularization coefficient; is the number of model parameters.

[0129] Among them, the specific expression of the cross-entropy loss function is:

[0130] ;

[0131] In the formula, is the cross-entropy loss function; is the one-hot encoding of the true label; is the probability of the class predicted by the model; is the total number of classes.

[0132] Among them, the specific expression of the mean squared error loss function is:

[0133] ;

[0134] In the formula, is the mean squared error loss function; is the true health score; is the predicted health score.

[0135] Specifically, the principle of the present invention is as follows: The technical principle of the present invention is based on a comprehensive analysis of multi-dimensional parameters of lithium batteries and a health state evaluation model driven by deep learning. First, from the perspective of electrochemical principles, the health state of a lithium battery is closely related to the internal electrochemical reaction kinetics, charge transfer mechanism, and mass transfer process. When the battery is damaged or has quality problems, these processes will undergo abnormal changes, which are then reflected in multiple dimensions such as the voltage characteristics, internal resistance parameters, temperature distribution, and power output of the battery.

[0136] The present invention innovatively constructs a series of feature matrices, each matrix targeting different aspects of the battery's performance. The temperature change matrix captures the spatio-temporal distribution characteristics of the battery surface temperature field, which can reflect internal hot spots and non-uniform heat release phenomena; the voltage matrix and the harmonic change matrix characterize the voltage stability and harmonic components of the battery, and can effectively identify internal short circuits and material aging problems; the resistance matrix comprehensively characterizes the internal electrochemical state of the battery through multi-frequency impedance measurements; and the cross-correlation index matrix quantifies the correlation between voltage and temperature changes, providing key cross-validation information.

[0137] The core of the present invention lies in adopting a deep neural network structure with a multi-head attention mechanism, which can adaptively focus on important information in different feature dimensions and assign corresponding weights. Through training with the mini-batch stochastic gradient descent method, the model can learn the complex non-linear mapping relationship between the multi-dimensional parameters of lithium batteries and their health states. In particular, the parameters of the multi-head attention mechanism are determined according to the internal resistance frequency response characteristics, thermal field distribution characteristics, and spectral distribution characteristics, making the model have strong generalization ability and interpretability.

[0138] In addition, the present invention optimizes the model performance by constructing a diverse training data set covering lithium battery samples of different models and different aging degrees, and adopting a cross-entropy loss function and an early stopping strategy, ensuring the reliability and stability of the evaluation results. This method based on multi-dimensional data fusion and deep learning can extract key features under complex noise backgrounds, achieve accurate evaluation of the health state of lithium batteries, and provide a scientific basis for industrial production and applications.

[0139] A specific embodiment 1 of the present invention is provided below, and the specific implementation methods of each step in this embodiment 1 are described in detail as follows.

[0140] The specific implementation of step S01 is to conduct high-precision multi-point monitoring on the surface temperature distribution of a lithium battery. First, an infrared thermal imager with a temperature resolution of at least 0.05°C is used to evenly arrange no less than 9 temperature acquisition points on the surface of the lithium battery, including 3 monitoring points in the positive electrode area, 3 monitoring points in the negative electrode area, and 3 monitoring points in the middle area. During the acquisition process, the lithium battery is placed in a constant temperature environment (25 ± 2°C), and a complete charge-discharge cycle test is carried out at three charging rates of 0.5C, 1C, and 2C and three discharging rates of 0.5C, 1C, and 2C. The temperature data of all monitoring points are recorded every 2 seconds for at least 3 complete charge-discharge cycles. The original temperature data collected are processed by a moving average filtering algorithm to eliminate random noise. The specific calculation formula of the filtering algorithm is:

[0141] ;

[0142] In the formula, is the temperature value of the th measurement point at moment after filtering; is the half-width of the filtering window, and the typical value is 2; is the original temperature data. This algorithm reduces the influence of random noise by calculating the arithmetic mean of the data within the window. Then, bicubic spline interpolation is used to generate a continuous temperature change curve. Based on the processed temperature data, a temperature change matrix T is constructed, and its specific representation is:

[0143] ;

[0144] In the formula, is the temperature change matrix; represents the temperature value of the th measurement point at moment, and the unit is °C; is the number of temperature measurement points, usually no less than 9; is the number of time sampling points. The function of this step is to capture possible local abnormal heating points inside the lithium battery through the spatio-temporal distribution characteristics of the temperature field. The temperature distribution of a healthy lithium battery should be relatively uniform, the abnormal temperature rise should not exceed 5°C, and the temperature difference between monitoring points is usually less than 3°C.

[0145] The specific implementation of step S02 is to use a high-precision voltage acquisition system (with an accuracy of not less than 1 mV) to monitor the terminal voltage of the lithium battery under different charge and discharge conditions. First, in the constant current charging mode, the lithium battery is charged at three charging rates of 0.2C, 0.5C, and 1C respectively, and the voltage value is recorded every 0.5 seconds until the battery voltage reaches the cut-off voltage. Then, in the constant current discharge mode, the lithium battery is discharged at four discharge rates of 0.2C, 0.5C, 1C, and 2C respectively, and the voltage value is also recorded every 0.5 seconds until the battery voltage drops to the discharge cut-off voltage. The recorded voltage data is filtered by a Butterworth low-pass filter with a cut-off frequency set to 10 Hz to remove high-frequency interference components. According to the processed voltage data, the voltage change rate (dV / dt) and voltage stability index under different load conditions are calculated. The voltage stability index is characterized by calculating the standard deviation of the voltage data, and the specific calculation formula is:

[0146] ;

[0147] In the formula, is the voltage stability index under the th charge and discharge rate condition, with the unit of V; is the average voltage value under the th charge and discharge rate condition, ; is the voltage value at the th moment under the th charge and discharge rate condition; is the number of time sampling points. A stability index lower than 10 mV is regarded as a stable state. Finally, a voltage matrix V is constructed, and its specific representation is:

[0148] ;

[0149] In the formula, is the voltage matrix; represents the voltage value at the th moment under the th charge and discharge rate condition, with the unit of V; is the number of charge and discharge rate conditions, usually 7; is the number of time sampling points. The purpose of this step is to reflect the electrochemical performance state of the lithium battery through voltage characteristics. For a healthy lithium battery, the voltage curve should have good repeatability and stability under the same conditions, and the voltage fluctuation amplitude does not exceed 2% of the charge and discharge platform voltage.

[0150] The specific implementation of step S03 is to calculate the real-time output power of the lithium battery under different working states based on the aforementioned collected voltage data and simultaneously recorded current data. The current data is collected by a Hall effect current sensor (with an accuracy better than 0.5%), and the sampling frequency is consistent with that of the voltage data collection (2 Hz). For each sampling time point, the instantaneous power value is calculated by multiplying the voltage and the current: , where is the voltage value at time tj under the ith working condition, is the current value at the corresponding time. At the same time, the energy conversion efficiency under each working condition is calculated. The charging energy efficiency is defined as the ratio of the actual stored energy to the input energy, and the discharging energy efficiency is defined as the ratio of the output energy to the theoretical stored energy. The specific calculation formulas are as follows:

[0151] ;

[0152] ;

[0153] In the formula, is the charging energy efficiency; is the discharging energy efficiency; is the stored energy; is the input energy; is the output energy; and are the voltage and current functions during the charging process respectively; and are the voltage and current functions during the discharging process respectively; and are the total times of charging and discharging respectively. According to the calculation results, a power matrix P is constructed, and its specific representation is:

[0154] ;

[0155] In the formula, is the power matrix; represents the power value at time under the th load condition, with the unit of W; is the corresponding voltage value; is the number of load conditions; is the number of time sampling points. This step aims to evaluate the energy conversion performance of the lithium battery through power characteristics. The energy conversion efficiency of a healthy lithium battery should generally be higher than 85%, and the power curve should maintain a relatively stable morphological characteristic under different load conditions.

[0156] The specific implementation of step S04 is to comprehensively measure the internal resistance characteristics of a lithium battery using a high-precision DC internal resistance tester and an electrochemical workstation. First, use the DC internal resistance tester to measure the DC internal resistance values of the lithium battery at different states of charge (SOC = 0%, 20%, 40%, 60%, 80%, 100%). The measurement current is controlled within the range of 0.1C to 0.2C, and the interval between each measurement is not less than 5 minutes. Each SOC point is measured 3 times and the average value is taken. Then, use the electrochemical workstation to conduct electrochemical impedance spectroscopy (EIS) tests. The frequency scanning range is from 0.01 Hz to 100 kHz. At least 30 frequency points are equally spaced and selected in the logarithmic coordinate. The amplitude of the excitation signal is controlled within 10 mV to ensure the linear response of the system. Based on the measurement data, an electrochemical equivalent circuit model is established. This model includes parameters such as ohmic internal resistance Rs, charge transfer resistance Rct, double-layer capacitance Cdl, and Warburg impedance W. The electrochemical impedance equation is specifically expressed as:

[0157] ;

[0158] In the formula, is the impedance value in complex form, with the unit of Ω; is the imaginary unit; is the angular frequency, with the unit of rad / s; is the frequency, with the unit of Hz; is the ohmic internal resistance, representing the electronic conduction impedance of the electrolyte and electrode materials, with the unit of Ω; is the charge transfer resistance, representing the impedance of the charge transfer process at the electrode - electrolyte interface, with the unit of Ω; is the double-layer capacitance, representing the charge accumulation ability at the electrode - electrolyte interface, with the unit of F; is the Warburg impedance coefficient, characterizing the diffusion process of ions in the electrode material, with the unit of .

[0159] The relationship between the parameters in the model and the physical characteristics of the battery is expressed as:

[0160] ;

[0161] In the formula, is the ohmic internal resistance, with the unit of Ω; is the temperature, with the unit of °C; is the electrolyte concentration, with the unit of mol / L; is the electrode contact area, with the unit of ; is the distance between electrodes, with the unit of m; is the electrolyte conductivity, related to temperature and concentration, with the unit of S / m; is the correction factor; is the error term.

[0162] ;

[0163] In the formula, is the charge transfer resistance, with the unit of Ω; is the gas constant, taking 8.314 J / (mol·K); is the absolute temperature, with the unit of K; is the number of electrons participating in the reaction; is the Faraday constant, taking 96485 C / mol; is the exchange current density, with the unit of A / ; is the electrode active area, with the unit of ; is the error term.

[0164] ;

[0165] In the formula, is the Warburg impedance coefficient, with the unit of ; is the diffusion coefficient of ions in the electrode material, with the unit of / s; is the reactant concentration, with the unit of mol / ; is the electrode active area, with the unit of ; is the error term.

[0166] Construct a resistance matrix R according to the measured data. The matrix is specifically expressed as:

[0167] ;

[0168] In the formula, is the resistance matrix; represents the complex impedance value measured at the SOC state at frequency, which is in complex form ; ; and are the real part and the imaginary part of the impedance respectively; is the number of frequency points, not less than 30; is the number of SOC states, usually 6 (0%, 20%, 40%, 60%, 80%, 100%). The purpose of this step is to reveal the internal structure and electrochemical reaction state of the lithium battery through the internal resistance characteristics. The internal resistance value of a healthy lithium battery is low and stable. The typical value does not exceed 50 mΩ in a 0.5C capacity lithium battery, and the frequency characteristic curve should present a typical semi-circle plus oblique line shape.

[0169] The specific implementation of step S05 is to analyze the correlation between the voltage change and temperature change of the lithium battery based on the previously constructed voltage matrix and temperature change matrix. First, perform time alignment and normalization on the voltage data and temperature data, convert all data into the range of [-1, 1] to eliminate the dimension difference. Then select the data at key time windows (such as rapid charge and discharge stages, constant voltage charging conversion points, etc.), and calculate the Pearson correlation coefficient between each voltage measurement point and each temperature measurement point. The specific calculation formula is:

[0170] ;

[0171] In the formula, is the Pearson correlation coefficient between the th voltage measurement point and the th temperature measurement point; is the normalized voltage value of the th voltage measurement point at time ; is the normalized temperature value of the th temperature measurement point at time ; and are the average values of the corresponding measurement points respectively; is the number of time sampling points. Based on the calculation results, construct the cross-correlation index matrix C, and the matrix is specifically expressed as:

[0172] ;

[0173] In the formula, is the cross-correlation index matrix; is the Pearson correlation coefficient between the th voltage measurement point and the th temperature measurement point; is the number of voltage measurement points; is the number of temperature measurement points, usually not less than 9. At the same time, calculate the overall correlation index to characterize the global voltage-temperature coupling characteristics of the battery. The function of this step is to quantify the intensity of the voltage-temperature coupling effect. The voltage-temperature correlation coefficient of a healthy lithium battery should usually be kept above 0.7, and the correlation characteristics in different regions of the same battery should be consistent, and the difference in the correlation coefficient does not exceed 0.15.

[0174] The specific implementation of step S06 is to perform frequency-domain analysis on the collected voltage data to evaluate the harmonic characteristics of the voltage waveform. First, the voltage time-series data under each charge-discharge state is segmented. Each segment takes 4096 data points, and the overlap rate between adjacent segments is 50%. Apply the Hann window function to each segment of data to reduce spectral leakage, and then perform the Fast Fourier Transform (FFT) to convert the time-domain signal into a frequency-domain representation. The specific calculation formula is:

[0175] , ;

[0176] In the formula, is the frequency-domain voltage data; is the time-domain voltage data; is the number of data points, taking 4096; is the imaginary unit; is the frequency index; is the time index. Calculate the power spectral density (PSD). The specific calculation formula is:

[0177] ;

[0178] In the formula, is the power spectral density at the th frequency point; and are the real part and the imaginary part of respectively; is the number of data points. Extract the amplitudes of different frequency components, and focus on the low-frequency harmonic components in the frequency range of 0.1 Hz to 10 Hz. This frequency band is usually related to the internal electrochemical reactions and material aging characteristics of the battery. Construct a harmonic variation matrix H according to the analysis results. The specific representation of the matrix is:

[0179] ;

[0180] In the formula, is the harmonic variation matrix; represents the amplitude of the th charge-discharge state at the frequency component; is the number of charge-discharge states; is the number of frequency components. The purpose of this step is to reveal the characteristics of the internal dynamic process of the lithium battery through spectral features. The voltage spectrum of a healthy lithium battery should exhibit the characteristic that the amplitude decays smoothly with the increase of frequency, and the amplitude of the harmonic components other than the main frequency component should not exceed 5% of the main frequency component.

[0181] The specific implementation of step S07 is to perform an inverse transformation on the aforementioned power matrix to evaluate the energy conversion efficiency of the lithium battery under reverse charge and discharge conditions. The specific implementation method is to first separate the charging power data and the discharging power data in the power matrix P, and then re-arrange the charging power data in reverse chronological order to generate a reverse charging power curve; the discharging power data is also re-arranged in reverse chronological order to generate a reverse discharging power curve. Calculate the cumulative energy conversion efficiency during the reverse charge and discharge process. The specific calculation formula is:

[0182] ;

[0183] In the formula, is the cumulative energy conversion efficiency of the reverse charge and discharge process, expressed as a percentage; is the discharging power function; is the charging power function; is the total charge and discharge time. Based on the calculation results, construct a reverse charge and discharge power matrix R. The specific representation of the matrix is:

[0184] ;

[0185] In the formula, is the reverse charge and discharge power matrix; represents the power value at the th moment during the reverse charge and discharge process under the th load condition, with the unit of W; is the number of load conditions; is the number of time sampling points. The purpose of this step is to evaluate the energy storage reversibility of the lithium battery through the reverse energy conversion characteristics. The reverse energy conversion efficiency of a healthy lithium battery should be no less than 80%, and the efficiency fluctuation under different load conditions should not exceed 5%.

[0186] The specific implementation of step S08 is to input the aforementioned matrix data into a pre-trained lithium battery health state evaluation model to achieve a comprehensive evaluation of the lithium battery state. First, perform feature extraction and dimensionality reduction on the resistance matrix, cross-correlation index matrix, harmonic matrix, and reverse charge and discharge power matrix. Use the principal component analysis (PCA) algorithm to retain the main feature dimensions with an explained variance ratio of no less than 95%. Then, standardize the dimensionality-reduced feature vectors and input them into a pre-trained deep neural network model. This model can adaptively focus on the important information of different feature dimensions based on the multi-head attention mechanism. The forward calculation process of the model includes feature embedding, multi-head self-attention calculation, feature fusion, and classification / regression output, etc. The multi-dimensional feature optimization function is specifically represented as:

[0187] ;

[0188] In the formula, Function optimized for multi-dimensional features; Set of model parameters; Number of training samples; Cross-entropy loss function; Mean squared error loss function; and are respectively the true class label and predicted class of the th sample; and are respectively the true health score and predicted score of the th sample; and are loss weight coefficients, usually ; Regularization coefficient; Number of model parameters.

[0189] The specific representation of the cross-entropy loss function is:

[0190] ;

[0191] In the formula, is the cross-entropy loss function; is the one-hot encoding of the true label; is the probability of the class predicted by the model; is the total number of classes.

[0192] The specific representation of the mean squared error loss function is:

[0193] ;

[0194] In the formula, is the mean squared error loss function; is the true health score; is the predicted health score.

[0195] The final model output includes the health status score (0 - 100 points) and the damage type classification result. The score threshold is set as follows: 90 - 100 points represents a healthy state, 75 - 90 points represents a mild aging state, 60 - 75 points represents a moderate aging state, 40 - 60 points represents a severe aging state, and below 40 points represents a damaged state. The damage types output at the same time include multiple categories such as capacity attenuation type, internal resistance increase type, poor contact type, and electrochemical activity reduction type. The purpose of this step is to achieve an accurate quantitative assessment of the health status of lithium batteries, combine multi-dimensional features to achieve high-precision discrimination of damaged or defective products, and the judgment accuracy rate should reach over 95% on the test data set.

[0196] To better understand and implement the present invention, the following provides Example 2 of a specific application scenario of the present invention: Based on the lithium battery rapid determination method of the present invention, researchers conducted a group of typical 18650-type lithium-ion battery defect detection experiments in a certain power battery manufacturing and testing laboratory. This experiment selected 10 groups of 18650-type lithium battery samples in different states, including 5 groups of healthy batteries and 5 groups of batteries that may have defects. All samples are NCM523 cathode material batteries with a nominal capacity of 3200 mAh and a nominal voltage of 3.7 V. The test environment temperature is controlled at 25 ± 1 °C, and the relative humidity is controlled at 45 ± 5%.

[0197] First, execute step S01. Use a FLIR T1050sc infrared thermal imager (temperature resolution 0.02 °C) to evenly arrange 9 temperature monitoring points on the surface of each battery, at the coordinate positions shown in Table 1:

[0198] Table 1 Layout coordinates of temperature monitoring points on the battery surface (mm)

[0199]

[0200] Conduct a 1C rate charging test on sample 1, and collect the temperature of all monitoring points every 2 seconds to obtain partial original temperature data as shown in Table 2:

[0201] Table 2 Partial temperature data of sample 1 during 1C charging (°C)

[0202]

[0203] Perform moving average filtering processing on the temperature data (window half-width k = 2) to obtain a filtered temperature change matrix. The highest temperature of each monitoring point of the healthy battery sample 1 is 34.92 °C, and the maximum temperature difference is 0.73 °C; while the highest temperature of the suspected defective sample 6 reaches 37.83 °C under the same conditions, and the maximum temperature difference reaches 3.26 °C, indicating that there may be local abnormal hot spots.

[0204] In step S02, use a high-precision voltage acquisition system with a precision of 0.5 mV to record the battery voltage data at different charge and discharge rates. Taking sample 1 and sample 6 as examples, conduct a 0.5C discharge test, and the voltage data compared at similar discharge capacity values are shown in Table 3:

[0205] Table 3 Partial voltage comparison data of sample 1 and sample 6 during 0.5C discharge (V)

[0206]

[0207] Calculate the voltage stability index of sample 1 is 8.7 mV, while that of Sample 6 is 18.6 mV, more than twice that of a healthy battery, indicating a significant reduction in the voltage stability of Sample 6. Figure 2 shows the comparison of the voltage curves of a healthy battery (Sample 1) and a suspected defective battery (Sample 6) during discharge at a rate of 0.5C. The horizontal axis of the graph represents the percentage of discharge capacity (from 0% to 100%), and the vertical axis represents the battery voltage (unit: V). The healthy battery is represented by circular marker points and a solid line, and the suspected defective battery is represented by square marker points and a solid line. The voltage differences (unit: mV) at 30%, 70%, and 90% depth of discharge are specifically marked in the figure, and the voltage performance differences between the two batteries during discharge are visually shown through the marked arrows. It can be clearly seen from the graph that as the depth of discharge increases, the voltage of the suspected defective battery drops significantly faster than that of the healthy battery, especially in the later stage of discharge (70% - 100%), where the voltage difference reaches 77 - 117 mV, indicating a significant reduction in the voltage stability of the suspected defective battery.

[0208] In step S03, the power characteristics are calculated based on the voltage data and current data, and a power matrix is constructed. At the same time, the energy efficiency of different samples is calculated. The charging energy efficiency of the healthy sample 1 is 92.8%, and the discharging energy efficiency is 94.3%; the charging energy efficiency of the suspected defective sample 6 is 87.2%, and the discharging energy efficiency is 89.1%, with the overall efficiency being approximately 5% lower.

[0209] In step S04, an electrochemical workstation is used to measure the internal resistance of all samples, and an electrochemical impedance spectroscopy test is carried out in the frequency range of 0.1 Hz to 10 kHz. Some of the measurement results are shown in Table 4:

[0210] Table 4 Partial impedance data (mΩ) of Sample 1 and Sample 6 at SOC = 50%

[0211]

[0212] Based on the fitting of the electrochemical impedance equation, the equivalent circuit parameters of Sample 1 and Sample 6 are shown in Table 5:

[0213] Table 5 Comparison of electrochemical equivalent circuit model parameters

[0214]

[0215] The ohmic internal resistance of Sample 6 and the charge transfer resistance are both significantly higher than those of the healthy sample 1, indicating the deterioration of its internal electrochemical performance. Figure 3 shows the Nyquist plot of the electrochemical impedance spectra of a healthy battery (Sample 1) and a suspected defective battery (Sample 6). The horizontal axis represents the real part of the impedance ( , unit: mΩ), and the vertical axis represents the negative imaginary part of the impedance ( (unit: mΩ). The impedance data of the two batteries are represented by circular and square marker points and connecting lines respectively. The key frequency points (10000 Hz, 100 Hz, 1 Hz, and 0.1 Hz) are marked in the figure, as well as the equivalent circuit model parameters of the two batteries, including the ohmic internal resistance Rs, the charge transfer resistance and the double-layer capacitance . It can be clearly seen from this chart by comparison that the impedance semicircle of the suspected defective battery (Sample 6) is significantly larger than that of the healthy battery. Its ohmic internal resistance and charge transfer resistance are 34.97 mΩ and 82.36 mΩ respectively, which are much higher than 21.86 mΩ and 43.75 mΩ of the healthy battery, indicating that the internal electrochemical performance of the suspected defective battery has deteriorated significantly.

[0216] In step S05, the voltage-temperature cross-correlation index matrix is calculated. The average correlation coefficient of the healthy Sample 1 is 0.913, and the maximum correlation coefficient difference is 0.092; while the average correlation coefficient of Sample 6 is 0.762, and the maximum correlation coefficient difference is 0.287, indicating that its voltage-temperature coupling characteristics are non-uniform and there may be local defects.

[0217] In step S06, the voltage data is subjected to spectral analysis. The harmonic components of Sample 6 in the 0.1 - 10 Hz frequency band are abnormally enhanced, especially in the 0.5 - 2 Hz frequency band, and the harmonic amplitude reaches 7.8% of the main frequency component, exceeding the 5% threshold of the healthy battery.

[0218] In step S07, the reverse charge-discharge power matrix is calculated and the reversibility of energy conversion is evaluated. The average reverse energy conversion efficiency of the healthy sample group is 86.3% with a standard deviation of 1.7%; while the average efficiency of the suspected defective sample group is 75.4% with a standard deviation of 3.9%, which is significantly lower than the healthy standard.

[0219] Finally, in step S08, all the characteristic data are input into the pre-trained lithium battery health state evaluation model, and the evaluation results are shown in Table 6:

[0220] Table 6 Evaluation Results of Lithium Battery Health State

[0221]

[0222] Figure 4Comprehensively presents the health assessment results of 10 battery samples. The main coordinate axis (on the left) represents the health score (0 - 100 points), and is represented by bar charts of different colors: green indicates a healthy state, orange indicates moderate aging, red indicates severe aging, and purple indicates a damaged state. Orange and red dashed lines in the figure respectively mark the moderate aging threshold (70 points) and the severe aging threshold (50 points). The secondary coordinate axis (on the right) uses a blue line graph to represent the determination credibility (unit: %). For battery samples with defects, the specific defect types are marked above the bar chart, such as reduced electrochemical activity type, increased internal resistance type, capacity attenuation type, and poor contact type. It can be directly seen from the chart that samples 1 - 5 are healthy batteries with a health score above 90 points; the health scores of samples 6 and 7 are relatively low, 51.3 points and 36.8 points respectively, belonging to the severe aging and damaged states; the health scores of samples 8 and 10 are between 60 - 70 points, belonging to the moderate aging state; the health score of sample 9 is only 32.4 points, belonging to the damaged state. The determination credibility of all samples is above 94.5%, indicating that the evaluation results have high reliability. Traditional lithium - battery defect detection mainly relies on single - parameter evaluation, such as capacity testing, DC internal resistance testing, or simple charge - discharge curve analysis. These methods often require a complete charge - discharge cycle, which takes 8 - 12 hours and is difficult to detect certain specific types of defects. For example, traditional capacity testing may not be able to detect potential safety hazards caused by local poor contact, and DC internal resistance testing is difficult to distinguish the increase in internal resistance caused by different mechanisms. In addition, traditional methods usually have a serial detection process and are difficult to achieve online real - time evaluation.

[0223] In contrast, the rapid determination method of lithium - battery in the present invention is based on multi - dimensional matrix feature analysis and deep - learning models, achieving a comprehensive three - dimensional evaluation of the battery. This method comprehensively considers data in multiple dimensions such as voltage, current, temperature, internal resistance, and voltage - temperature coupling relationship, and deeply reveals the root causes of defects through electrochemical mechanism analysis. Experimental results show that the detection accuracy of this method reaches more than 95%, and it can effectively distinguish different types of defect mechanisms. In terms of detection efficiency, this method only needs to perform partial charge - discharge tests, and the entire detection process can be completed within 1 hour, which is about 80% more efficient than traditional methods. In addition, the multi - dimensional feature analysis of this method has higher sensitivity and can discover potential safety hazards that are easily overlooked in traditional methods, such as local hot spots, micro - short circuits, etc. These innovation points make the present invention have significant application value and technical advantages in the fields of quality inspection of lithium - battery production lines and safety management of energy - storage systems.

[0224] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 7 and 8 below.

[0225] Table 7 Variable Explanation Table (First Part)

[0226]

[0227] Table 8 Explanation Table of Variables (Second Part)

[0228]

[0229] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A method for quickly determining a damaged or defective lithium battery, characterized in that, Including: Collecting temperature data, voltage data, and current data of the lithium battery under different charge and discharge states; Constructing a temperature change matrix, a voltage matrix, a power matrix, a resistance matrix, a mutual correlation index matrix, a harmonic change matrix, and a reverse charge and discharge power matrix; Measuring the internal resistance data of the lithium battery using a DC internal resistance tester and an AC impedance spectroscopy analyzer, recording impedance values at multiple frequency points, and constructing a resistance matrix to characterize the internal electrochemical performance state of the lithium battery; Based on the data of the voltage matrix and the temperature change matrix, calculating the correlation coefficient between the temperature change and the voltage change of the lithium battery, and constructing a mutual correlation index matrix; Inputting the data of the resistance matrix, the mutual correlation index matrix, the harmonic change matrix, and the reverse charge and discharge power matrix into a pre-trained lithium battery health state evaluation model, and judging whether the lithium battery is damaged or defective according to the output result of the lithium battery health state evaluation model; The temperature change matrix specifically refers to a two-dimensional data structure that characterizes the surface temperature change of the lithium battery over time. The rows of the temperature change matrix represent different measurement point positions on the battery surface, the columns represent different time points, and the element values of the temperature change matrix are the temperature values corresponding to the positions and times; The voltage matrix specifically refers to a two-dimensional data structure that characterizes the terminal voltage change of the lithium battery under different charge and discharge states. The rows of the voltage matrix represent different charge and discharge rate conditions, the columns represent different time points, and the element values of the voltage matrix are the voltage values corresponding to the conditions and times; The power matrix specifically refers to a two-dimensional data structure that characterizes the output power change of the lithium battery over time, calculated by multiplying the voltage data and the current data. The rows of the power matrix represent different load conditions, the columns represent different time points, and the element values of the power matrix are the power values corresponding to the conditions and times; The resistance matrix specifically refers to a two-dimensional data structure that characterizes the internal resistance characteristics of the lithium battery. The rows of the resistance matrix represent different measurement frequencies, the columns represent different charge states, and the element values of the resistance matrix are the internal resistance values measured at the corresponding frequencies and charge states; The mutual correlation index matrix specifically refers to a two-dimensional data structure that quantifies the correlation between the voltage change and the temperature change of the lithium battery. The rows of the mutual correlation index matrix represent the voltage measurement points, the columns represent the temperature measurement points, and the element values of the mutual correlation index matrix are the Pearson correlation coefficients corresponding to the voltage points and the temperature points; The harmonic change matrix specifically refers to a two-dimensional data structure that characterizes the spectral characteristics of the lithium battery voltage waveform. The rows of the harmonic change matrix represent different charge and discharge states, the columns represent different frequency components, and the element values of the harmonic change matrix are the amplitudes of the frequency components in the corresponding states; The reverse charge and discharge power matrix specifically refers to a two-dimensional data structure that characterizes the energy conversion characteristics of the lithium battery under reverse charge and discharge conditions, obtained by performing a mathematical transformation on the power matrix, and is used to evaluate the energy conversion efficiency of the lithium battery.

2. The method for quickly determining damaged or defective lithium batteries according to claim 1, wherein The specific structure of the lithium battery health state evaluation model is a deep neural network structure based on a multi-head attention mechanism, including an input layer, multiple attention layers, a feature fusion layer, and an output layer; Among them, the multi-head attention mechanism parameters are determined according to three key parameters: the internal resistance frequency response characteristics of the resistance matrix, the thermal field distribution characteristics of the temperature change matrix, and the spectral distribution characteristics of the harmonic change matrix.

3. The method for quickly determining a damaged or defective lithium battery according to claim 2, wherein The steps for establishing a training data set during the training process of the lithium battery health state evaluation model include collecting lithium battery samples of different models and different aging degrees, performing a standardized test process for each sample to collect multi-matrix feature data, performing dimensionless processing on the collected data, and determining the actual health state of each sample as a training label.

4. A computer-readable storage medium, characterized in that, Program instructions are stored in the computer-readable storage medium, and when the program instructions run on a computer, they are used to execute the method for quickly determining a damaged or defective lithium battery according to any one of claims 1-3.

5. A rapid determination system for damaged or defective lithium batteries, characterized in that, The system includes the computer-readable storage medium according to claim 4, the system is any one of a computer, a server, and a single-chip microcomputer, the computer-readable storage medium is disposed within the system, and a microprocessor for executing the program instructions stored in the computer-readable storage medium is disposed within the system.

Images