Lithium battery performance detection method, apparatus and device, and storage medium

By weighted calculation and multi-mode measurement of the multi-dimensional parameters of lithium batteries, combined with wavelet transformation and orthogonal frequency division multiplexing modulation technology, the performance detection of lithium batteries in the existing technology is solved by using multi-scale attention bidirectional long and short-term memory network. The multi-dimensional and nonlinear characteristics are difficult to capture, and high accuracy and reliability of battery health status detection is achieved.

CN118980944AInactive Publication Date: 2024-11-19SHENZHEN WEIXINGMING ENERGY TECH CO LTD
View PDF 0 Cites 24 Cited by

Patent Information

Application Number
CN202411339831.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-25
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing lithium battery performance detection methods are difficult to accurately capture dynamic characteristics and degradation processes under complex operating conditions, and usually only focus on single parameters or static characteristics, ignoring the multi-dimensional, nonlinear and time-varying characteristics of battery performance degradation, resulting in detection results. The accuracy and reliability are limited.

Method used

By weighted calculations of the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery, select the target measurement point, perform static and dynamic dual-mode measurements, combining wavelet transformation, orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operations , enter the multi-scale attention bidirectional long and short-term memory network for battery performance detection to obtain the battery health status value.

Benefits of technology

A comprehensive evaluation of the performance of lithium batteries has been achieved, the accuracy and reliability of detection has been improved, the anti-interference ability of the system has been enhanced, and the parallel detection and remote monitoring of large-scale battery packs have been realized, which has reduced human intervention and improved detection efficiency. and consistency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118980944A_ABST
    Figure CN118980944A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of lithium batteries, and discloses a lithium battery performance detection method, device and equipment and a storage medium, and the method comprises the steps: carrying out the weighted calculation of the residual capacity, internal resistance, temperature gradient and voltage deviation of a single lithium battery, obtaining a comprehensive score, and selecting a target measurement point; performing static and dynamic dual-mode measurement according to the target measurement point to obtain original battery performance data; performing wavelet transform denoising and feature extraction to obtain a static feature vector and a dynamic feature vector; performing digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; performing orthogonal frequency division multiplexing modulation and multi-mode charging and discharging operation on the calibration performance data to obtain multi-dimensional battery parameters; and inputting the multi-dimensional battery parameters into a multi-scale attention bidirectional long-short-term memory network for battery performance detection to obtain a battery health state value, thereby realizing parallel detection and remote monitoring of a large-scale battery pack.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of lithium batteries, and in particular to a performance detection method, device, equipment and storage medium for lithium batteries. Background Art

[0002] The performance of lithium batteries will gradually degrade with the use time and changes in environmental conditions, which not only affects the normal operation of the equipment, but also may cause safety hazards. Traditional battery performance detection methods often rely on simple voltage and current measurements or fixed models, which are difficult to accurately capture the dynamic characteristics and degradation process of batteries under complex working conditions.

[0003] In addition, existing detection technologies usually only focus on a single parameter or static characteristics, ignoring the multidimensional, nonlinear and time-varying characteristics of battery performance degradation, resulting in limited accuracy and reliability of detection results. Summary of the invention

[0004] The present application provides a lithium battery performance detection method, device, equipment and storage medium, thereby realizing parallel detection and remote monitoring of large-scale battery packs.

[0005] In a first aspect, the present application provides a performance detection method for a lithium battery, the performance detection method for a lithium battery comprising: Performing weighted calculation on the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery cell to obtain a comprehensive score, and selecting a target measurement point according to the comprehensive score; According to the target measurement point, static and dynamic dual-mode measurements are performed on the lithium battery cell to obtain original battery performance data; Performing wavelet transform denoising and feature extraction on the original battery performance data to obtain a static feature vector and a dynamic feature vector; Performing digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; Performing orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation on the calibration performance data to obtain multi-dimensional battery parameters; The multi-dimensional battery parameters are input into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection to obtain a battery health status value.

[0006] A second aspect of the present application provides a performance detection device for a lithium battery, the performance detection device for a lithium battery comprising: A selection module is used to perform weighted calculation on the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery cell to obtain a comprehensive score, and select a target measurement point according to the comprehensive score; A measurement module, used to perform static and dynamic dual-mode measurements on the lithium battery cell according to the target measurement point to obtain original battery performance data; An extraction module, used for performing wavelet transform denoising and feature extraction on the original battery performance data to obtain a static feature vector and a dynamic feature vector; A processing module, used for performing digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; An operation module, used to perform orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation on the calibration performance data to obtain multi-dimensional battery parameters; The detection module is used to input the multi-dimensional battery parameters into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection to obtain a battery health status value.

[0007] The third aspect of the present application provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the electronic device executes the above-mentioned lithium battery performance detection method.

[0008] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned lithium battery performance detection method.

[0009] Compared with the prior art, the present application has the following beneficial effects: by weighted calculation of the remaining capacity, internal resistance, temperature gradient and voltage deviation, a comprehensive score is obtained, a comprehensive evaluation of the battery performance is achieved, the one-sidedness that may be caused by a single parameter is avoided, and the accuracy and reliability of the detection are improved. Combined with static open circuit voltage measurement and dynamic charge and discharge test, the static characteristics and dynamic response of the battery are fully captured, which helps to more accurately reflect the actual use status and performance changes of the battery. The original data is denoised by wavelet transform technology, and the measurement noise is effectively eliminated through time-frequency analysis and statistical feature extraction, the signal-to-noise ratio of the feature is improved, and the feature vector is optimized through programmable gain adjustment, adaptive filtering and self-calibration technologies, the signal quality and data reliability are improved, and the anti-interference ability of the system is enhanced. The calibration data is modulated by OFDM technology, which improves the data transmission efficiency and anti-interference ability, and is conducive to the parallel detection and remote monitoring of large-scale battery packs. By executing multiple charge and discharge modes, various working conditions of the battery in actual use are fully simulated, more abundant and real performance data are obtained, and the representativeness and reliability of the test results are improved. Combining multi-scale convolution, bidirectional LSTM and self-attention mechanism, we can achieve multi-scale feature extraction and long-term and short-term dependency modeling of battery performance data, and improve the model's ability to capture the nonlinear dynamic behavior of the battery and the prediction accuracy. By building a complete process from raw data collection to the final health status output, we can achieve automation and intelligence of battery performance testing, reduce human intervention, and improve detection efficiency and consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0011] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0012] Figure 1 is a schematic flow chart of a performance testing method for a lithium battery provided by an embodiment of the present invention; Figure 2It is a schematic block diagram of the structure of a lithium battery performance detection device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0015] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0016] It should be further understood that the term "and / or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 , an embodiment of the performance detection method of the lithium battery in the embodiment of the present application includes: Step 100: Perform weighted calculation on the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery cell to obtain a comprehensive score, and select a target measurement point according to the comprehensive score; It is understandable that the execution subject of the present application may be a performance detection device for a lithium battery, or may be a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0017] Specifically, the coulomb counting method is used to measure the remaining capacity of the lithium battery cell, and the remaining capacity value is obtained by recording the current and time accumulation of the battery during the discharge process. By applying an AC pulse signal to the lithium battery cell, the voltage response of the battery is measured to obtain the internal resistance value, which reflects the changes in the electrical performance of the battery under different working conditions. The surface thermal imaging of the lithium battery cell is scanned by an infrared thermal imager to obtain the temperature distribution matrix of the battery surface. Based on the temperature distribution matrix, the temperature difference at different positions on the battery surface is calculated, and the maximum value of the temperature gradient is extracted to obtain the characteristic value of the temperature gradient, which reflects the uniformity of the temperature distribution of the battery during the charging and discharging process and its thermal performance characteristics. At the same time, by sampling the voltage of all cells in the lithium battery pack, the difference between the maximum voltage and the minimum voltage is calculated to obtain the voltage deviation value. The remaining capacity value, internal resistance value, temperature gradient characteristic value and voltage deviation value are weighted and assigned to obtain the weight coefficient of each parameter. Based on the weight coefficient of each parameter, the remaining capacity value, internal resistance value, temperature gradient characteristic value and voltage deviation value are weighted and summed to obtain a comprehensive score, which reflects the overall health status of the battery cell. Based on the comprehensive score, the lithium battery cells are arranged in descending order, and the top E% of lithium battery cells are selected as target measurement points, where E is a preset threshold. In this way, battery cells with poor performance can be effectively identified.

[0018] Step 200: Perform static and dynamic dual-mode measurements on the lithium battery cell according to the target measurement point to obtain original battery performance data; Specifically, the lithium battery cell is placed in a constant temperature environment and kept still to ensure that the influence of the ambient temperature on the battery performance is minimized during the test and the accuracy of the data is guaranteed. The open circuit voltage of the battery is measured using a high-precision voltmeter to obtain the static open circuit voltage value. The static open circuit voltage is the voltage of the battery under no-load condition, which reflects the initial state of the battery and provides a benchmark for subsequent measurements. A multi-frequency AC impedance spectrum test signal is applied to the lithium battery cell, and the impedance response of the battery is measured to obtain complex impedance data. Complex impedance data includes the real and imaginary parts of the impedance, which are key parameters for analyzing the electrochemical reaction process and transmission characteristics inside the battery. Based on the complex impedance data, the battery parameters are extracted by the equivalent circuit fitting method to construct a static equivalent circuit model. The equivalent circuit model can simulate the internal structure and electrochemical characteristics of the battery. After completing the static measurement, a pulse current excitation is applied to the lithium battery cell, and the voltage response curve of the battery is recorded. By analyzing the voltage response curve, the time constant of the battery is calculated using the exponential function fitting method, and the dynamic polarization characteristics of the battery are obtained, which reveals the response behavior of the battery under transient conditions and reflects the strength of the battery polarization phenomenon. At the same time, the lithium battery cell is connected to the programmable electronic load device to perform a multi-rate constant current discharge test. During the discharge process, the voltage, current and temperature data of the battery are recorded in real time to obtain the dynamic discharge characteristic curve of the battery. This curve can reflect the performance of the battery at different discharge rates. In order to analyze the dynamic discharge characteristic curve, a piecewise linear fitting is performed on it to calculate the slope and intercept of each section. Through these fitting parameters, the dynamic capacity decay index of the battery is obtained to effectively evaluate the capacity loss of the battery. Based on the constructed static equivalent circuit model and dynamic polarization characteristics, the state space equation is constructed. The internal state of the battery is estimated by the extended Kalman filter algorithm to obtain the initial values ​​of the state of charge and the state of health. The state of charge reflects the current remaining power of the battery, while the state of health is a key indicator for measuring the overall performance of the battery. The static open circuit voltage value, complex impedance data, dynamic polarization characteristics, dynamic discharge characteristic curve, dynamic capacity decay index, state of charge and initial values ​​of the state of health of the lithium battery are combined, and these measurement results are integrated to form the original battery performance data.

[0019] Step 300: Perform wavelet transform denoising and feature extraction on the original battery performance data to obtain static feature vectors and dynamic feature vectors; It should be noted that the original battery performance data is subjected to wavelet decomposition to decompose the data into approximate coefficients and detail coefficients. The approximate coefficients retain the low-frequency information of the data, and the detail coefficients contain the high-frequency part. In order to optimize the quality of the data, the signal-to-noise ratio of each layer is calculated based on the detail coefficients to obtain the corresponding denoising parameters. The signal-to-noise ratio is an important indicator for measuring signal quality, and its level determines the degree of influence of noise in the data. The detail coefficients are denoised using the soft threshold processing method. Soft threshold processing retains the key features of the signal by weakening the noise component instead of completely eliminating it. Combined with the previously extracted approximate coefficients, wavelet reconstruction is performed to obtain the denoised battery performance data. Time-frequency analysis is performed on the denoised data to extract the time-frequency feature matrix of the battery performance. Time-frequency analysis combines information in two dimensions, time and frequency, and can analyze the frequency response of the battery at different time points. Singular value decomposition is performed based on the time-frequency feature matrix, and the left singular vectors corresponding to the first k largest singular values ​​are selected to obtain the principal component features. Singular value decomposition can extract the most representative principal component in the data by selecting the largest singular value, thereby reducing redundant information. At the same time, the static parameters in the denoised battery performance data are statistically analyzed to calculate their statistical characteristics such as mean, variance, skewness and kurtosis. These statistical characteristics can reflect the basic distribution of battery performance and the characteristics of data fluctuations. For example, the mean reflects the central trend of the data, the variance measures the degree of dispersion of the data, and the skewness and kurtosis describe the symmetry and sharpness of the data. The principal component features and statistical features are fused. By comprehensively utilizing different types of features, the representativeness and effectiveness of the overall features are improved. During the fusion process, the Fisher discriminant criterion is used to calculate the weights of each feature to obtain a weighted feature vector. The Fisher discriminant criterion is a weight calculation method based on the degree of separation between classes, which can effectively distinguish feature data of different classes and ensure the rationality of weight distribution. Nonlinear dimensionality reduction is performed based on the weighted feature vector to separate static features from dynamic features. Nonlinear dimensionality reduction technology simplifies complex high-dimensional data while retaining the core information of the data, and finally obtains static feature vectors and dynamic feature vectors.

[0020] Step 400: Perform digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; Specifically, the static eigenvector and the dynamic eigenvector are input into a programmable gain amplifier. The programmable gain amplifier automatically adjusts the gain size according to the characteristics of the input signal, and accurately adjusts the gain through a digital potentiometer to amplify the input signal to obtain an initial amplified signal. The initial amplified signal is Fourier transformed to convert the signal from the time domain to the frequency domain to analyze its frequency components. The frequency distribution characteristics of the signal are obtained by calculating the power spectrum density of the signal. Based on the frequency distribution characteristics, the initial amplified signal is filtered to filter out unnecessary frequency components and noise to obtain a filtered signal. The filtered signal is input into a self-calibration circuit, and the gain error and offset error are calculated by comparing the reference voltage source. The reference voltage source provides a stable reference signal, and the self-calibration circuit calculates the gain error and offset error based on the reference signal. These errors are caused by nonlinearity in the circuit and device instability. The calibration parameters are obtained by calculation, and the filtered signal is linearly calibrated according to the calibration parameters to eliminate system errors. The calibrated signal is subjected to a Hilbert transform to calculate the instantaneous frequency and instantaneous amplitude of the signal, and the time-varying characteristics of the signal are extracted. The instantaneous frequency and instantaneous amplitude reflect the frequency change and amplitude change of the signal at different moments respectively. These time-varying features can reveal the dynamic behavior of the signal at different time points. Based on the extracted time-varying features, the power frequency interference in the signal is suppressed. Power frequency interference is a common external power supply noise, which usually interferes with the battery performance signal. By suppressing the signal and removing the power frequency interference, a de-interference signal is obtained. The de-interference signal is digitally sampled through an analog-to-digital converter, and the analog-to-digital converter converts the continuous analog signal into a discrete digital signal to obtain calibration performance data.

[0021] Step 500, performing orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation on the calibration performance data to obtain multi-dimensional battery parameters; Specifically, the calibration performance data is subjected to a fast Fourier transform, the original time domain signal is converted into a frequency domain signal, the frequency component of the signal is extracted, and it is divided into multiple subcarriers. Each subcarrier represents a different frequency range, and then 16QAM modulation is applied to each subcarrier. This modulation method can transmit more information on a single carrier, so that the data transmission efficiency can be improved, and the modulated subcarrier signal is obtained. The modulated subcarrier signal is subjected to an inverse fast Fourier transform to generate a time domain OFDM symbol. In order to avoid interference between symbols, a cyclic prefix is ​​added before the time domain OFDM symbol. By transmitting information in parallel on multiple subcarriers, the anti-interference ability and stability of the signal can be effectively improved, and the obtained orthogonal frequency division multiplexing modulated signal is more stable and efficient. The OFDM modulated signal is input into the programmable power supply system to perform constant current-constant voltage charging operation on the lithium battery pack. During the charging process, the voltage, current and temperature data of the battery are recorded in real time, and a charging characteristic curve is generated to show the characteristics of the battery at different charging stages. By analyzing the charging characteristic curve, the charging efficiency of the lithium battery and its response under different charging conditions can be understood. Perform multi-rate pulse discharge operations on lithium battery packs. Set different discharge rates and pulse widths to simulate the dynamic response characteristics of the battery under actual working conditions. During the discharge process, record the curves of voltage drop and voltage recovery to reflect the discharge capacity of the battery and its recovery performance under different load conditions. Based on the charging characteristic curve and dynamic response characteristics, the impedance characteristics of lithium batteries at different states of charge are measured using electrochemical impedance spectroscopy technology. The impedance characteristics can reflect the details of the electrochemical reactions inside the battery. These data are used to derive the impedance spectrum of the battery at different states of charge, revealing its electrochemical stability and performance. Feature extraction and parameter fitting of the charging characteristic curve, dynamic response characteristics, and impedance spectrum data are performed to obtain multi-dimensional battery parameters including internal resistance, polarization parameters, capacity, and power characteristics.

[0022] Step 600: Input the multi-dimensional battery parameters into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection to obtain a battery health status value.

[0023] Specifically, the multi-dimensional battery parameters are time series segmented. Through a sliding window with a length of T and a step size of S, N input sequences are generated, each sequence contains M feature dimensions, so as to obtain a time series feature matrix with a shape of (N, T, M). The segmentation operation converts the original multi-dimensional parameters into a standard input format processed by a neural network, which is helpful for the subsequent network time series modeling. The time series feature matrix is ​​input into a multi-scale convolution layer. The multi-scale convolution layer consists of three parallel one-dimensional convolution sublayers, which use convolution operations with convolution kernel sizes of 3, 5 and 7 respectively. The step size of each sublayer is set to 1, and ReLU is used as the activation function. Each sublayer outputs 32 feature maps. Through multi-scale convolution operations, the time series patterns at different scales are captured, and three multi-scale time feature tensors with shapes of (N, T, 32) are output. The maximum pooling operation is performed on these three multi-scale time feature tensors, with a pooling kernel size of 2 and a step size of 1 to reduce the complexity of the feature dimension while retaining the main time series information. After pooling, the shape of each feature tensor is still (N, T, 32), but the information is simplified and refined. The three downsampled feature tensors are concatenated in the feature dimension to generate a fused feature tensor with a shape of (N, T, 96). The fused feature tensor is input into the bidirectional long short-term memory network (BiLSTM) layer. The BiLSTM network contains 128 hidden units, 64 forward units and 64 backward units, which are processed by the tanh activation function to generate a bidirectional sequence representation tensor with a shape of (N, T, 128). Bidirectional LSTM can capture the forward dependency of the data and model the reverse time dependency, so that the network has a more comprehensive understanding of the overall performance of the battery parameters in time. After the bidirectional sequence representation tensor is generated, the self-attention mechanism is applied. The bidirectional sequence representation tensor is converted into the query matrix Q, the key matrix K and the value matrix V respectively through linear transformation. The attention score matrix is ​​obtained by calculating the dot product of the query matrix Q and the key matrix K and dividing it by the scaling factor 8. The attention score matrix is ​​then converted into a weight matrix through the softmax function, and multiplied with the value matrix V to generate a weighted feature representation tensor with a shape of (N, T, 128). The self-attention mechanism can give higher weights to important information in time series data and improve the network's sensitivity to the battery status at critical moments. The weighted feature representation tensor is converted into a two-dimensional tensor with a shape of (NT, 128) and input into the fully connected layer. The fully connected layer contains 64 neurons and uses the ReLU activation function to extract more advanced abstract features and generate a high-level abstract feature tensor with a shape of (NT, 64). After the fully connected layer, a dropout operation is performed on the high-level abstract feature tensor with a dropout rate of 0.5. By randomly discarding some neuron connections, the network is prevented from overfitting, and a regularized feature tensor with a shape of (N*T, 64) is obtained.The regularized feature tensor is input to the output layer, which contains a single neuron and is processed using the Sigmoid activation function. The Sigmoid function compresses the output value between 0 and 1, and finally obtains a numerical value representing the health status of the battery. This health status value can provide accurate quantitative results for the performance evaluation of lithium batteries, helping to determine the current condition of the battery and its future service life.

[0024] In the embodiment of the present application, by weighted calculation of the remaining capacity, internal resistance, temperature gradient and voltage deviation, a comprehensive score is obtained, a comprehensive evaluation of the battery performance is achieved, the one-sidedness that may be caused by a single parameter is avoided, and the accuracy and reliability of the detection are improved. Combined with static open circuit voltage measurement and dynamic charge and discharge test, the static characteristics and dynamic response of the battery are fully captured, which helps to more accurately reflect the actual use status and performance changes of the battery. The original data is denoised by wavelet transform technology, and the measurement noise is effectively eliminated by time-frequency analysis and statistical feature extraction, the signal-to-noise ratio of the feature is improved, and the feature vector is optimized by programmable gain adjustment, adaptive filtering and self-calibration and other technologies, the signal quality and data reliability are improved, and the anti-interference ability of the system is enhanced. The calibration data is modulated by OFDM technology, which improves the data transmission efficiency and anti-interference ability, and is conducive to the parallel detection and remote monitoring of large-scale battery packs. By executing multiple charge and discharge modes, the various working conditions of the battery in actual use are fully simulated, and more abundant and real performance data are obtained, which improves the representativeness and reliability of the test results. Combining multi-scale convolution, bidirectional LSTM and self-attention mechanism, we can achieve multi-scale feature extraction and long-term and short-term dependency modeling of battery performance data, and improve the model's ability to capture the nonlinear dynamic behavior of the battery and the prediction accuracy. By building a complete process from raw data collection to the final health status output, we can achieve automation and intelligence of battery performance testing, reduce human intervention, and improve detection efficiency and consistency.

[0025] In a specific embodiment, the process of executing step 100 may specifically include the following steps: The remaining capacity of the lithium battery cell is measured by the coulomb counting method to obtain the remaining capacity value, and an AC pulse signal is applied to the lithium battery cell to measure the voltage response to obtain the internal resistance value; The surface of the lithium battery cell is scanned by infrared thermal imaging to obtain a temperature distribution matrix, and the maximum value of the temperature gradient is calculated based on the temperature distribution matrix to obtain a temperature gradient characteristic value; Sampling the voltage of all cells in the lithium battery pack, calculating the difference between the maximum voltage and the minimum voltage, obtaining the voltage deviation value, and weighting the remaining capacity value, internal resistance value, temperature gradient characteristic value and voltage deviation value to obtain the weight coefficient of each parameter; The remaining capacity value, internal resistance value, temperature gradient characteristic value and voltage deviation value are weighted and summed according to the weight coefficient of each parameter to obtain a comprehensive score; Based on the comprehensive scores, the lithium battery cells are arranged in descending order, and the top E% of lithium battery cells are selected as target measurement points, where E is a preset threshold.

[0026] Specifically, the remaining capacity of a lithium battery cell is measured by the coulomb counting method. The coulomb counting method calculates the remaining capacity of the battery by real-time monitoring of the battery's discharge current and time accumulation. Assume that the battery's discharge current at a certain moment is , after a period of time , the remaining capacity is expressed as: ; Among them, among them, is the remaining capacity of the battery, is the initial capacity of the battery, is the instantaneous value of the current, is the duration of the discharge process. Integrate to get the remaining capacity of the battery at different discharge stages. Apply an AC pulse signal to the lithium battery cell and measure the voltage response to get the internal resistance of the battery. The AC pulse signal is a small-amplitude alternating current that causes a change in voltage when it acts on the battery. According to Ohm's law, the internal resistance of the battery is determined by measuring the ratio of voltage to current. For a low-frequency AC pulse signal, the internal resistance It is expressed as: ; in, is the change in battery voltage under the action of the pulse signal, is the corresponding current change value. By measuring the voltage response at different frequencies multiple times, a more accurate internal resistance value can be obtained. At the same time, in order to evaluate the thermal characteristics of lithium batteries during use, an infrared thermal imager is used to perform thermal imaging scanning on the battery surface to obtain a temperature distribution matrix. The thermal imager captures the temperature changes of different areas on the battery surface in real time and generates a matrix ,in and They represent the two-dimensional coordinates of the battery surface, and each element of the matrix represents the temperature of that point. By analyzing the temperature distribution matrix, the maximum temperature gradient is extracted to reflect the uniformity of the thermal distribution of the battery during operation. Calculated by the following formula: ; Maximum temperature gradient It can indicate the area where the battery is overheated or has poor heat dissipation. By sampling the voltage of all cells in the lithium battery pack, the difference between the maximum voltage and the minimum voltage is calculated to obtain the voltage deviation value. Voltage deviation is an important parameter to measure the consistency of single cells in a battery pack. For a battery pack, the greater the voltage difference between cells, the worse the overall performance balance. Voltage deviation It is expressed as: ; in, is the voltage of the largest single cell in the battery pack, is the voltage of the smallest single cell. By monitoring the voltage deviation, the overall consistency of the battery pack and whether there are battery cells with serious performance degradation can be determined. Weights are assigned to the remaining capacity value, internal resistance value, temperature gradient characteristic value and voltage deviation value to obtain their respective weight coefficients. Assume that the weight coefficient of the remaining capacity is , the weight coefficient of internal resistance is , the weight coefficient of the temperature gradient is , the weight coefficient of voltage deviation is , then the comprehensive score Calculated by weighted summation: ; Among them, each weight coefficient Represents the relative importance of different parameters in battery performance evaluation. The weight coefficient is adjusted based on experimental data or engineering experience to ensure that the comprehensive score can accurately reflect the overall health status of the battery. , sort the lithium battery cells in descending order, and select the top E% of the batteries as the target measurement points. Preset threshold It is set according to actual needs and is usually used to screen out batteries with poor performance for more detailed inspection or maintenance. For example, if the comprehensive score in a battery pack is lower than a certain threshold, the battery has serious degradation or potential failure and needs to be handled in time.

[0027] In a specific embodiment, the process of executing step 200 may specifically include the following steps: The lithium battery cell is placed in a constant temperature environment and left to stand, and the open circuit voltage is measured using a high-precision voltmeter to obtain a static open circuit voltage value, and a multi-frequency AC impedance spectrum test signal is applied to the lithium battery cell to measure the impedance response and obtain complex impedance data; Based on the complex impedance data, the battery parameters are extracted by the equivalent circuit fitting method to obtain the static equivalent circuit model. The pulse current excitation is applied to the lithium battery cell, the voltage response curve is recorded, and the time constant is calculated by the exponential function fitting method to obtain the dynamic polarization characteristics. Connect the lithium battery cell to a programmable electronic load, perform a multi-rate constant current discharge test, record the voltage, current and temperature data, obtain a dynamic discharge characteristic curve, and perform piecewise linear fitting on the dynamic discharge characteristic curve, calculate the slope and intercept of each segment, and obtain a dynamic capacity attenuation index; Based on the static equivalent circuit model and dynamic polarization characteristics, the state space equation is constructed, and the internal state of the battery is estimated through the extended Kalman filter algorithm to obtain the initial values ​​of the state of charge and health state; The static open circuit voltage value, complex impedance data, dynamic polarization characteristics, dynamic discharge characteristic curve, dynamic capacity decay index, state of charge and initial value of health state are combined to obtain the original battery performance data.

[0028] Specifically, the lithium battery cell is placed in a constant temperature environment to ensure the stability of the test environment and the repeatability of the data. During the static state of the battery, the internal chemical reaction tends to equilibrium, which can better reflect the actual open circuit voltage state of the battery. The open circuit voltage of the battery is measured using a high-precision voltmeter to obtain the static open circuit voltage value. The open circuit voltage is the potential difference of the battery under no-load conditions. It is a basic parameter in battery performance testing and can reflect the state of charge of the battery. A multi-frequency AC impedance spectrum test signal is applied to the lithium battery cell, and its impedance response is measured to obtain complex impedance data. The AC impedance spectrum test studies the internal electrochemical process by applying a small amplitude AC signal at different frequencies and analyzing the voltage response of the battery. The complex impedance data contains real and imaginary parts, which represent the resistive component and capacitive or inductive component of the battery, respectively, and can effectively reveal the electrochemical reaction and charge transfer mechanism of the battery. Based on the complex impedance data, the key parameters of the battery are extracted by the equivalent circuit fitting method to obtain a static equivalent circuit model. The equivalent circuit model is a simplified representation of the complex electrochemical process inside the battery, usually containing multiple components such as resistance, capacitance and inductance. By fitting the model with the actual impedance data, each parameter in the equivalent circuit, such as the internal resistance, polarization impedance and capacitance of the battery, is determined. After establishing a static model, a pulse current excitation is applied to the lithium battery cell, and its voltage response curve is recorded. Pulse current excitation is a method of instantaneous large current stimulation that can reveal the dynamic polarization characteristics of the battery. By analyzing the voltage response curve, the time constant of the battery is calculated by the exponential function fitting method. The time constant reflects the speed at which the battery recovers from the polarization state to the stable state, and is an important indicator of the dynamic performance of the battery. Assume that the change in voltage response is expressed as: ; in, It's in time The voltage at the moment, is the voltage in steady state, is the amplitude of the voltage change, is the time constant. By fitting the voltage response data, the time constant is calculated , thereby characterizing the dynamic polarization characteristics of the battery. Connect the lithium battery cell to a programmable electronic load and perform a multi-rate constant current discharge test. By setting different discharge rates, recording the voltage, current and temperature data of the battery under different discharge conditions, a dynamic discharge characteristic curve is generated to reflect the performance of the battery under different load conditions. In order to analyze the attenuation characteristics of the battery, a piecewise linear fit is performed on the dynamic discharge characteristic curve, and the slope and intercept of each section are calculated to obtain the dynamic capacity attenuation index. Assume that a certain section of the discharge characteristic is approximately a linear relationship: ; in, It's time The voltage at the moment, is the slope, which indicates how fast the voltage decreases over time during the discharge process. is the intercept. By fitting the different stages of the discharge curve, the attenuation rate at different rates is obtained, which can accurately evaluate the loss of battery capacity. After completing the extraction of static and dynamic parameters, the state space equation is constructed based on the static equivalent circuit model and dynamic polarization characteristics. The state space equation is a mathematical model that describes the internal state changes of the battery. It can estimate the initial values ​​of the battery's state of charge (SOC) and state of health (SOH) based on the battery's voltage, current, temperature and other information. In order to improve the accuracy of the estimation, the extended Kalman filter algorithm is used to solve the state space equation. The extended Kalman filter is a commonly used nonlinear state estimation method that can accurately estimate the battery state in the presence of large noise and uncertainty. The general form of the state space equation is expressed as: ; ; in, The battery is The state vector at the moment, is the input (such as current), and are the state transfer equation and the observation equation, and It is the system noise and measurement noise. Through the extended Kalman filter algorithm, the initial values ​​of SOC and SOH of the battery are iteratively estimated under the continuously updated voltage and current data. The static open circuit voltage value, complex impedance data, dynamic polarization characteristics, dynamic discharge characteristic curve, dynamic capacity decay index state of charge and initial value of health status are combined to form complete original battery performance data.

[0029] The lithium battery cell is connected to the programmable electronic load through a high-precision connecting line to ensure a stable and reliable connection; the working mode of the programmable electronic load is set to a constant current discharge mode, and a plurality of different discharge rates are preset, including but not limited to 0.2C, 0.5C, 1C and 2C; at each preset discharge rate, a constant current discharge test is performed until the battery voltage drops to a cut-off voltage; during the discharge process, a high-precision data acquisition system is used to synchronously record the voltage, current and surface temperature data of the battery at a sampling frequency of not less than 1Hz; the recorded voltage, current and time data are sorted out, and a curve of voltage variation over time is drawn to obtain a dynamic discharge characteristic curve; the dynamic discharge characteristic curve is segmented and the curve is divided into a plurality of characteristic segments, including but Not limited to the initial segment, the stable segment and the fast decline segment; apply the least squares method to each characteristic segment for linear fitting to obtain the linear equation of each segment; calculate the slope and intercept of each linear equation, wherein the slope represents the voltage change rate, and the intercept represents the initial voltage of the segment; based on the slope of each segment, calculate the changing trend of the voltage change rate to obtain the voltage decay rate; use the integral method to calculate the actual discharge capacity at each discharge rate; compare the actual discharge capacity at different discharge rates, calculate the rate of change of capacity with the discharge rate, and obtain the rate performance index; combine the voltage decay rate and the rate performance index to construct a dynamic capacity decay index, which is used to characterize the performance decay characteristics of the battery under different discharge conditions; output the dynamic capacity decay index as a basis for evaluating the dynamic performance of the battery. Based on the static equivalent circuit model, a battery voltage equation is established, wherein the voltage equation includes open circuit voltage, ohmic internal resistance voltage drop and polarization voltage terms; based on the dynamic polarization characteristics, a dynamic equation of polarization voltage is established, wherein the dynamic equation describes the relationship between the polarization voltage and time and current; the battery capacity is introduced into the equation group as a state variable, and a capacity decay model is established to describe the change of capacity with the number of cycles; the voltage equation, polarization voltage dynamic equation and capacity decay model are combined to construct a nonlinear state space equation, wherein the state variables include state of charge, polarization voltage and battery capacity; the nonlinear state space equation is discretized and linearized to obtain a linearized discrete state space equation; Based on the linearized discrete state space equation, an extended Kalman filter is designed, including a state prediction equation and a measurement update equation; the state vector and covariance matrix of the extended Kalman filter are initialized, wherein the initial state vector includes an estimated value of an initial state of charge, an initial polarization voltage, and an initial battery capacity; the current and voltage measurement data of the battery are input, and the internal state of the battery is iteratively estimated through a prediction-update loop of the extended Kalman filter; the state of charge estimate is extracted from the state estimation result of the extended Kalman filter; the initial state of health value is calculated based on the estimated value of the battery capacity and the rated capacity; and the state of charge estimate and the initial state of health value are output as an initial evaluation result of the battery state.

[0030] In a specific embodiment, the process of executing step 300 may specifically include the following steps: Perform wavelet decomposition on the original battery performance data to obtain approximate coefficients and detail coefficients, and calculate the signal-to-noise ratio of each layer based on the detail coefficients to obtain denoising parameters; Perform soft threshold processing on detail coefficients according to denoising parameters, and perform wavelet reconstruction in combination with approximate coefficients to obtain denoised battery performance data; Perform time-frequency analysis on the denoised battery performance data to obtain a time-frequency feature matrix, and perform singular value decomposition based on the time-frequency feature matrix to select the left singular vectors corresponding to the first k largest singular values ​​to obtain the principal component features; Perform statistical analysis on the static parameters in the denoised battery performance data, calculate the mean, variance, skewness and kurtosis, and obtain statistical characteristics; The principal component features and statistical features are fused, and the feature weights are calculated using the Fisher discriminant criterion to obtain the weighted feature vector; Nonlinear dimensionality reduction is performed based on the weighted feature vector to separate the static features from the dynamic features, and obtain the static feature vector and the dynamic feature vector.

[0031] Specifically, the original battery performance data is subjected to wavelet decomposition. Wavelet decomposition is a multi-scale analysis tool that can decompose the signal into components in different frequency ranges and effectively capture the local characteristics of the signal in the time domain and frequency domain. Wavelet decomposition decomposes the signal into two parts: approximate coefficients and detail coefficients. The approximate coefficients retain the low-frequency components of the signal and reflect the overall trend of the signal, while the detail coefficients capture the high-frequency components of the signal and reflect short-term changes or noise in the signal. After obtaining the detail coefficients, the signal-to-noise ratio (SNR) is calculated for the detail coefficients of each layer to evaluate the intensity of the noise in the signal and determine the denoising parameters. The signal-to-noise ratio is defined as the ratio of the effective component of the signal to the noise component. Assume that the signal power is The noise power is , then the signal-to-noise ratio (SNR) is expressed as: ; By calculating the SNR of the detail coefficients of each layer, the noise level of different frequency bands in the signal is evaluated, and the corresponding denoising parameters are determined. The detail coefficients are soft-thresholded according to the denoising parameters. Soft thresholding achieves the denoising effect by suppressing detail coefficients with smaller amplitudes (i.e., noise components) while retaining detail coefficients with larger amplitudes. The processed detail coefficients are combined with the unmodified approximate coefficients for wavelet reconstruction to obtain the denoised battery performance data. The denoised battery performance data is subjected to time-frequency analysis. Time-frequency analysis can simultaneously consider the time and frequency characteristics of the signal and reveal the frequency changes of the signal in different time periods. Through time-frequency analysis, a time-frequency feature matrix is ​​constructed, which not only contains the frequency distribution of the signal, but also captures the dynamic characteristics of the signal changing over time. Based on the time-frequency feature matrix, singular value decomposition is performed. Singular value decomposition is a matrix decomposition technique that can decompose the original matrix into three parts: left singular matrix, singular value matrix, and right singular matrix. The singular values ​​in the singular value matrix reflect the importance of the feature, and larger singular values ​​represent features that have a greater impact on the original data. By selecting the left singular vectors corresponding to the first k largest singular values, the principal component features of the signal can be extracted to achieve dimensionality reduction and retain the main features of the signal. At the same time, the static parameters in the denoised battery performance data are statistically analyzed. By calculating statistical features such as mean, variance, skewness and kurtosis, the distribution and change characteristics of the signal are revealed. The mean reflects the central trend of the signal, the variance measures the fluctuation range of the signal, the skewness indicates the symmetry of the signal distribution, and the kurtosis measures the sharpness of the signal distribution. By calculating these statistical features, static feature information is provided for subsequent feature fusion. The principal component features and statistical features are fused. By combining time-frequency features with statistical features, the important information in the battery performance data can be more comprehensively reflected. In order to ensure that the weights of the influence of different features on the final result are reasonable, the Fisher discriminant criterion is used to calculate the weights of each feature. The Fisher discriminant criterion ensures that the extracted features can effectively distinguish data of different categories by maximizing inter-class differences and minimizing intra-class differences. The calculation formula of feature weight is expressed as: ; in, and is the mean of the two types of data, and is the variance of the two types of data. By maximizing , get the optimal feature weight allocation scheme and form a weighted feature vector. Based on the weighted feature vector, perform nonlinear dimensionality reduction to separate static features from dynamic features. Nonlinear dimensionality reduction is a technology that can reduce redundant information while retaining the main features of the data. Map high-dimensional data to low-dimensional space while retaining the main change patterns in the data. Through the dimensionality reduction process, extract the static feature vector and dynamic feature vector in the battery performance data respectively.

[0032] In a specific embodiment, the process of executing step 400 may specifically include the following steps: The static eigenvector and the dynamic eigenvector are input into a programmable gain amplifier, and the gain is adjusted by a digital potentiometer to obtain an initial amplified signal; Performing Fourier transform on the initial amplified signal, calculating the power spectrum density of the signal, obtaining the frequency distribution characteristics, and filtering the initial amplified signal based on the frequency distribution characteristics to obtain a filtered signal; Inputting the filtered signal into the self-calibration circuit, calculating the gain error and the offset error by comparing with the reference voltage source, obtaining the calibration parameters, and performing linear calibration on the filtered signal according to the calibration parameters to obtain the calibrated signal; Performing Hilbert transform on the calibrated signal, calculating the instantaneous frequency and instantaneous amplitude of the signal, obtaining time-varying characteristics, and suppressing the power frequency interference in the calibrated signal based on the time-varying characteristics to obtain a de-interference signal; The interference-removed signal is digitally sampled through an analog-to-digital converter to obtain calibration performance data.

[0033] Specifically, the static eigenvector and the dynamic eigenvector are input into a programmable gain amplifier. The programmable gain amplifier adjusts the gain through a digital potentiometer according to the different strengths of the signal, so that the amplitude of the signal is amplified within an appropriate range. The digital potentiometer accurately adjusts the resistance value in the circuit, thereby changing the gain of the amplifier. Assume that the gain is , the input signal is , then the output signal The relationship is expressed as: ; By adjusting the gain , control the amplitude of the output signal to make it suitable for subsequent signal processing. Through amplification, the initial signal is enhanced, thereby enhancing the signal detection accuracy and the ability to analyze signal characteristics in subsequent processing steps. The initial amplified signal is Fourier transformed to convert the time domain signal into a frequency domain signal so that the frequency components of the signal can be extracted and analyzed. Through Fourier transformation, the power spectral density (PSD) of the signal is obtained, which represents the power distribution of each frequency component in the signal. Power spectral density Defined as the signal at each frequency The power at the location can help analyze the main frequency components of the signal. The power spectrum density formula is: ; By analyzing the power spectral density, the frequency distribution characteristics of the signal are obtained, from which the main frequency components and the frequency range of the noise are identified. Based on the frequency distribution characteristics, the initial amplified signal is filtered. The filtering operation is intended to retain the effective frequency components in the signal while filtering out unnecessary noise. Using a bandpass filter, the signal components within the frequency range of interest are selected, and high-frequency noise or low-frequency drift beyond this range is suppressed to obtain a filtered signal. The filtered signal is input into the self-calibration circuit. The self-calibration circuit calculates the gain error and offset error by comparing the signal with the reference voltage source. The gain error is usually caused by device drift or instability in the circuit, and the offset error comes from noise or external interference in the system. The reference voltage source provides a stable reference voltage for the system, and the error in the system is identified by comparison with the signal. Assume that the actual gain of the signal is , the target gain is , then the gain error It is expressed as: ; Similarly, the bias error It is expressed as the difference between the input signal and the reference voltage source. By calculating these errors, the corresponding calibration parameters are obtained, and the filtered signal is linearly calibrated according to the calibration parameters to ensure that the amplitude and offset of the signal are within the ideal range and obtain a more accurate calibration signal. After completing the linear calibration, the Hilbert transform is performed on the calibrated signal. The Hilbert transform is a time-frequency analysis tool that calculates the instantaneous frequency and instantaneous amplitude by constructing the analytical signal of the signal. The instantaneous frequency reflects the frequency change of the signal at different time points, while the instantaneous amplitude represents the change of the signal amplitude over time. Assume that the signal is , whose Hilbert transform is , then the analytical signal It is expressed as: ; in, The instantaneous amplitude and instantaneous frequency Respectively expressed as: ; By calculating the instantaneous frequency and instantaneous amplitude, the time-varying characteristics of the signal are extracted to reflect the dynamic changes of the battery performance data. Based on the time-varying characteristics, the power frequency interference in the signal is suppressed. Power frequency interference is a common external power supply noise, usually 50Hz or 60Hz. By analyzing the instantaneous frequency characteristics, the frequency component of the power frequency interference is identified, and the frequency is suppressed by designing a narrowband filter to effectively remove the influence of the power supply noise on the signal, and a purer interference-free signal is obtained. The interference-free signal is digitally sampled through an analog-to-digital converter. The analog-to-digital converter can convert a continuous analog signal into a discrete digital signal so that it can be processed and stored by a digital system. The core of analog-to-digital conversion is to sample the signal at a predetermined sampling frequency and quantize each sampling point into a corresponding digital value. The sampling frequency should comply with the Nyquist sampling theorem, that is, the sampling frequency should be at least twice the highest frequency of the signal to avoid signal aliasing. Through analog-to-digital conversion, digital calibration performance data is obtained.

[0034] The filtered signal is input into a self-calibration circuit, wherein the self-calibration circuit includes a programmable gain amplifier, an analog-to-digital converter and a digital signal processor; the filtered signal is preliminarily amplified by the programmable gain amplifier to obtain an amplified signal; the amplified signal and a preset reference voltage signal are simultaneously input into the analog-to-digital converter for digital sampling to obtain a digitized amplified signal and a reference signal; in the digital signal processor, the digitized amplified signal and the reference signal are compared and analyzed to calculate the difference between the two; based on the difference, the gain error and the offset error of the system are calculated, wherein the gain error is expressed as the ratio of the actual gain to the ideal gain, and the offset error is expressed as the DC offset of the output signal; the gain error and the offset error are combined to obtain calibration parameters; according to the calibration parameters, a linear calibration equation is constructed, wherein the calibration equation is in the form of: y=ax+b, wherein a is the gain correction factor and b is the offset correction amount; the filtered signal is substituted into the linear calibration equation, and a point-to-point calibration calculation is performed to obtain a calibrated signal; the calibrated signal is output to the next processing stage or storage unit. Performing a discrete Hilbert transform on the calibrated signal to obtain an analytical signal, wherein the analytical signal includes a real part and an imaginary part; calculating the instantaneous phase of the signal based on the real and imaginary parts of the analytical signal, specifically: the instantaneous phase is equal to the inverse tangent of the imaginary part divided by the real part; performing a time difference operation on the instantaneous phase and dividing it by the sampling time interval to obtain the instantaneous frequency; calculating the modulus of the analytical signal to obtain the instantaneous amplitude; combining the instantaneous frequency and the instantaneous amplitude to obtain a time-varying feature; constructing an adaptive notch filter based on the time-varying feature, wherein the center frequency of the filter is determined according to the statistical distribution of the instantaneous frequency, and the bandwidth of the filter is determined according to the rate of change of the instantaneous amplitude; passing the calibrated signal through the adaptive notch filter to suppress the power frequency interference component and obtain a de-interferenced signal.

[0035] In a specific embodiment, the process of executing step 500 may specifically include the following steps: Performing fast Fourier transform on the calibration performance data to obtain a frequency domain signal, dividing the frequency domain signal into a plurality of subcarriers, applying 16QAM modulation to each subcarrier to obtain a modulated subcarrier signal; Performing an inverse fast Fourier transform on the modulated subcarrier signal to obtain a time-domain OFDM symbol, and adding a cyclic prefix before the time-domain OFDM symbol to obtain an orthogonal frequency division multiplexing modulated signal; The orthogonal frequency division multiplexing modulation signal is input to the programmable power supply, and the constant current-constant voltage charging operation is performed on the lithium battery pack, and the voltage, current and temperature data during the charging process are recorded to obtain the charging characteristic curve; Perform multi-rate pulse discharge operation on the lithium battery pack, set different discharge rates and pulse widths, record the voltage drop and recovery curve during the discharge process, and obtain the dynamic response characteristics; Based on the charging characteristic curve and dynamic response characteristics, the impedance characteristics of lithium batteries at different charge states are measured by electrochemical impedance spectroscopy technology to obtain impedance spectrum data; The charging characteristic curve, dynamic response characteristics and impedance spectrum data are subjected to feature extraction and parameter fitting to obtain multi-dimensional battery parameters including internal resistance, polarization parameters, capacity and power characteristics.

[0036] Specifically, the calibration performance data of the lithium battery is subjected to a fast Fourier transform to convert the time domain signal into a frequency domain signal. Decomposing the complex time domain signal into different frequency components makes the frequency domain analysis easier. Assume that the calibration performance data is , after Fourier transform, we get the frequency domain signal

[0037] ; The transformation result Represents the amplitude and phase information of the signal at different frequencies. By analyzing the frequency domain signal, the response characteristics of the lithium battery at different frequencies are revealed. The frequency domain signal is divided into multiple subcarriers, each of which represents a signal within a specific frequency range. In order to improve data transmission efficiency, 16QAM (16-order quadrature amplitude modulation) is applied to each subcarrier. 16QAM is an efficient digital modulation method that maps information to 16 different symbol points by simultaneously adjusting the amplitude and phase of the signal. This modulation technology can not only transmit more information, but also increase the data rate within a limited bandwidth. Assume that the subcarrier signal is , then the signal after 16QAM modulation is expressed as: ; in, and Representing amplitude and phase respectively, the subcarrier signal after 16QAM modulation contains more information. The modulated subcarrier signal is subjected to inverse fast Fourier transform, and the frequency domain signal is converted back to the time domain to obtain an orthogonal frequency division multiplexing (OFDM) symbol. OFDM technology effectively improves the signal's anti-interference ability by transmitting multiple subcarrier signals in parallel. In order to avoid inter-symbol interference (ISI), a cyclic prefix is ​​added before the time domain OFDM symbol to ensure that the OFDM symbol has stronger anti-interference and stability during transmission, and an orthogonal frequency division multiplexing modulated signal is obtained. The orthogonal frequency division multiplexing modulated signal is input into the programmable power supply system to perform constant current-constant voltage charging operation on the lithium battery pack. During the charging process, the voltage, current and temperature data of the battery are recorded in real time to generate a charging characteristic curve. In the constant current charging stage, the current remains unchanged and the voltage gradually increases; in the constant voltage charging stage, the voltage remains constant and the current gradually decreases. This process helps to analyze the characteristics of lithium batteries under different charging conditions. The charging characteristic curve can reveal the battery's charging efficiency, internal impedance changes, and thermal performance. Perform multi-rate pulse discharge operations on the lithium battery pack. By setting different discharge rates and pulse widths, the load response behavior of the battery under actual working conditions is simulated. The voltage drop and recovery curve of the battery during the discharge process are recorded in real time. The voltage drop reflects the response speed of the battery under load, and the recovery curve can show the time it takes for the battery to recover from the discharge state to the stable state. These data together constitute the dynamic response characteristics of the battery. In order to analyze the performance of the battery, these curves are feature extracted, including the calculation of dynamic indicators such as internal resistance, polarization parameters and recovery time at each discharge rate. Based on the charge and discharge tests, the impedance characteristics of lithium batteries at different states of charge (SOC) are measured by electrochemical impedance spectroscopy. The electrochemical impedance spectroscopy can reveal the impedance changes of the battery at different frequencies, and help analyze the charge transfer, electrochemical reactions and polarization effects inside the battery. Assume that the impedance is , then decomposed into the real part and the imaginary part

[0038] ; By measuring impedance at different frequencies, the impedance spectrum data of lithium batteries is generated. The charging characteristic curve, dynamic response characteristics and impedance spectrum data are feature extracted and parameter fitted to generate multi-dimensional battery parameters, including internal resistance, polarization parameters, capacity and power characteristics. By fitting the impedance spectrum data, the equivalent circuit model of the lithium battery is obtained, including key parameters such as electrolyte resistance, double layer capacitance, diffusion impedance, etc. For example, the internal resistance reflects the internal conductivity of the battery, and the polarization parameter describes the voltage loss caused by the battery polarization effect.

[0039] In a specific embodiment, the process of executing step 600 may specifically include the following steps: The multi-dimensional battery parameters are segmented into time series, and a sliding window with a length of T and a step size of S is used to generate N input sequences, each of which contains M feature dimensions, to obtain a time series feature matrix with a shape of (N, T, M); The temporal feature matrix is ​​input into the multi-scale convolution layer. The multi-scale convolution layer contains three parallel one-dimensional convolution sub-layers. The convolution kernel size of the first sub-layer is 3, the convolution kernel size of the second sub-layer is 5, and the convolution kernel size of the third sub-layer is 7. The step size of each sub-layer is 1, the activation function is ReLU, and each sub-layer outputs 32 feature maps, obtaining three multi-scale temporal feature tensors with shapes of (N, T, 32); Perform the maximum pooling operation on the three multi-scale time feature tensors respectively, with the pooling kernel size of 2 and the step size of 1, to obtain three downsampled feature tensors with the shape of (N, T, 32), and then concatenate these three downsampled feature tensors in the feature dimension to obtain a fused feature tensor with the shape of (N, T, 96); The fused feature tensor is input into the bidirectional long short-term memory network layer. The bidirectional long short-term memory network layer contains 128 hidden units, including 64 forward units and 64 backward units. The tanh activation function is used to obtain a bidirectional sequence representation tensor with a shape of (N, T, 128). Apply the self-attention mechanism to the bidirectional sequence representation tensor and calculate the attention weight matrix. The specific steps are as follows: the bidirectional sequence representation tensor is linearly transformed to obtain the query matrix Q, key matrix K and value matrix V respectively, then the dot product of Q and K is calculated and divided by 8, and then the attention weight is obtained by the softmax function. Finally, the attention weight is multiplied by V to obtain a weighted feature representation tensor with a shape of (N, T, 128); The weighted feature representation tensor is converted into a two-dimensional tensor of shape (NT, 128), and then input into a fully connected layer, which contains 64 neurons and uses the ReLU activation function to obtain a high-level abstract feature tensor of shape (NT, 64); Perform a dropout operation on the high-level abstract feature tensor with a dropout rate of 0.5 to obtain a regularized feature tensor with a shape of (N*T, 64); The regularized feature tensor is input into the output layer, which is a single neuron. The Sigmoid activation function is used to obtain the battery health status value.

[0040] Specifically, the multi-dimensional battery parameters are segmented into time series. and the step length is The sliding window method generates input sequences, each of which contains feature dimensions. The data is organized into a shape of The time series feature matrix. The continuous time series data is divided into a format suitable for input into the deep learning network to ensure that the model can simultaneously learn the relationship between the time dimension and the feature dimension. The time series feature matrix is ​​input into the multi-scale convolution layer. The multi-scale convolution layer contains three parallel convolution sub-layers with convolution kernel sizes of 3, 5, and 7 respectively. The convolution kernel of each sub-layer is used to capture temporal patterns and change characteristics at different scales. The sub-layer with a convolution kernel size of 3 captures short-term superpotentials, the sub-layer with a size of 5 captures medium-term changes, and the sub-layer with a convolution kernel size of 7 identifies longer-term change patterns. The step size of each convolution sub-layer is set to 1, and ReLU is used as the activation function. ReLU can effectively introduce nonlinearity and improve the model's ability to learn complex patterns. Each convolution sub-layer outputs 32 feature maps, resulting in three shapes. The multi-scale time feature tensor represents the battery performance characteristics at different time scales. The maximum pooling operation is performed on each multi-scale time feature tensor, with a pooling kernel size of 2 and a step size of 1. The purpose of maximum pooling is to reduce the dimension of the feature while retaining the main feature information. The pooling operation retains the most significant features by extracting the maximum value within the local window, thereby achieving downsampling. After pooling, the three tensors obtained are , but the data volume is compressed, improving the computational efficiency of the model. The three downsampled feature tensors are concatenated in the feature dimension to form a shape of The fused feature tensor of . Feature information of different scales is fused together. The fused feature tensor is input into the bidirectional long short-term memory network (BiLSTM) layer. Bidirectional LSTM is a recurrent neural network structure that can process time series data and can simultaneously consider the forward and backward dependencies in the time series. The bidirectional LSTM layer contains 128 hidden units, of which 64 are forward and 64 are backward. The activation function uses tanh, which can help the model learn the nonlinear mapping between input and output and better capture the complex timing patterns in battery performance data. The shape of the bidirectional sequence representation tensor output by the bidirectional LSTM layer is , which contains the time series information captured by the bidirectional LSTM. The self-attention mechanism is applied to the bidirectional sequence representation tensor to calculate the importance of each time step in the time series, focusing on those time points that contribute most to the final prediction. The bidirectional sequence representation tensor is transformed linearly to obtain the query matrix , key matrix Sum Matrix The query matrix is ​​used to calculate the attention weights with the key matrix. The dot product of the query matrix and the key matrix is ​​calculated, then divided by the scaling factor 8, and the result is normalized to the attention weight matrix through the softmax function. The attention weight matrix reflects the contribution of each time step to the overall sequence. The attention weight matrix is ​​combined with the value matrix Multiply them together to get the weighted feature representation tensor, the shape is The weighted feature representation tensor retains the information of important time steps in the sequence, thereby improving the predictive power of the model. Convert the weighted feature representation tensor to a shape of The fully connected layer contains 64 neurons and uses the ReLU activation function, generating a 2D tensor of shape The function of the fully connected layer is to further map the high-dimensional features extracted previously and extract higher-level abstract features that can better characterize the battery health status. In order to prevent the model from overfitting, a dropout operation is performed on the high-level abstract feature tensor after the fully connected layer, and the dropout rate is set to 0.5. Dropout makes the model more robust by randomly discarding some neurons, reduces dependence on training data, and improves the model's generalization ability on test data. After dropout processing, the shape is The regularized feature tensor is input into the output layer, which contains a single neuron and uses the Sigmoid activation function. The Sigmoid function compresses the output to a range between 0 and 1, which is suitable as the predicted output of the battery health state value. The output of Sigmoid is interpreted as the probability value of the battery health state. A value close to 1 indicates a good battery health state, and a value close to 0 indicates a poor battery health state.

[0041] The above describes the performance detection method of the lithium battery in the embodiment of the present application. The following describes the performance detection device 10 of the lithium battery in the embodiment of the present application. Figure 2 In one embodiment of the present application, a performance detection device 10 for a lithium battery includes: The selection module 11 is used to perform weighted calculation on the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery cell to obtain a comprehensive score, and select a target measurement point according to the comprehensive score; The measuring module 12 is used to perform static and dynamic dual-mode measurements on the lithium battery cells according to the target measuring points to obtain the original battery performance data; An extraction module 13 is used to perform wavelet transform denoising and feature extraction on the original battery performance data to obtain a static feature vector and a dynamic feature vector; A processing module 14 is used to perform digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; An operation module 15 is used to perform orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation on the calibration performance data to obtain multi-dimensional battery parameters; The detection module 16 is used to input multi-dimensional battery parameters into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection and obtain a battery health status value.

[0042] Through the synergy of the above components, the remaining capacity, internal resistance, temperature gradient and voltage deviation are weighted to obtain a comprehensive score, which realizes a comprehensive evaluation of battery performance, avoids the one-sidedness that may be caused by a single parameter, and improves the accuracy and reliability of detection. Combined with static open circuit voltage measurement and dynamic charge and discharge test, the static characteristics and dynamic response of the battery are fully captured, which helps to more accurately reflect the actual use status and performance changes of the battery. Wavelet transform technology is used to reduce the noise of the original data, and time-frequency analysis and statistical feature extraction are used to effectively eliminate measurement noise and improve the signal-to-noise ratio of the feature. Through programmable gain adjustment, adaptive filtering and self-calibration and other technologies, the feature vector is optimized to improve the signal quality and data reliability, and enhance the anti-interference ability of the system. The calibration data is modulated using OFDM technology to improve the data transmission efficiency and anti-interference ability, which is conducive to the parallel detection and remote monitoring of large-scale battery packs. By executing multiple charge and discharge modes, various working conditions of the battery in actual use are fully simulated to obtain more abundant and real performance data, and the representativeness and reliability of the test results are improved. Combining multi-scale convolution, bidirectional LSTM and self-attention mechanism, we can achieve multi-scale feature extraction and long-term and short-term dependency modeling of battery performance data, and improve the model's ability to capture the nonlinear dynamic behavior of the battery and the prediction accuracy. By building a complete process from raw data collection to the final health status output, we can achieve automation and intelligence of battery performance testing, reduce human intervention, and improve detection efficiency and consistency.

[0043] The present application also provides an electronic device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the lithium battery performance detection method in the above-mentioned embodiments.

[0044] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the lithium battery performance detection method.

[0045] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0046] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0047] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A performance testing method for a lithium battery, characterized in that: The performance detection method of the lithium battery comprises: Performing weighted calculation on the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery cell to obtain a comprehensive score, and selecting a target measurement point according to the comprehensive score; According to the target measurement point, static and dynamic dual-mode measurements are performed on the lithium battery monomer to obtain original battery performance data; Performing wavelet transform denoising and feature extraction on the original battery performance data to obtain a static feature vector and a dynamic feature vector; Performing digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; Performing orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation on the calibration performance data to obtain multi-dimensional battery parameters; The multi-dimensional battery parameters are input into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection to obtain a battery health status value.

2. The performance testing method of a lithium battery according to claim 1, characterized in that: The weighted calculation of the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery monomer is performed to obtain a comprehensive score, and the target measurement point is selected according to the comprehensive score, including: The remaining capacity of the lithium battery cell is measured by the coulomb counting method to obtain the remaining capacity value, and an AC pulse signal is applied to the lithium battery cell to measure the voltage response to obtain the internal resistance value; Performing surface thermal imaging scanning on the lithium battery cell using an infrared thermal imager to obtain a temperature distribution matrix, and calculating a maximum temperature gradient based on the temperature distribution matrix to obtain a temperature gradient characteristic value; Sampling the voltages of all cells in the lithium battery pack, calculating the difference between the maximum voltage and the minimum voltage to obtain a voltage deviation value, and weighting the remaining capacity value, the internal resistance value, the temperature gradient characteristic value, and the voltage deviation value to obtain a weight coefficient of each parameter; According to the weight coefficients of the parameters, the remaining capacity value, the internal resistance value, the temperature gradient characteristic value and the voltage deviation value are weighted and summed to obtain a comprehensive score; The lithium battery cells are arranged in descending order based on the comprehensive score, and the lithium battery cells ranked in the top E% are selected as target measurement points, where E is a preset threshold.

3. The performance testing method of a lithium battery according to claim 1, characterized in that: The step of performing static and dynamic dual-mode measurements on the lithium battery cell according to the target measurement point to obtain original battery performance data includes: Placing the lithium battery cell in a constant temperature environment and letting it stand, using a high-precision voltmeter to measure the open circuit voltage to obtain a static open circuit voltage value, and applying a multi-frequency AC impedance spectrum test signal to the lithium battery cell to measure the impedance response to obtain complex impedance data; Based on the complex impedance data, battery parameters are extracted by an equivalent circuit fitting method to obtain a static equivalent circuit model, and pulse current excitation is applied to the lithium battery cell, a voltage response curve is recorded, and a time constant is calculated by an exponential function fitting method to obtain dynamic polarization characteristics; Connecting the lithium battery cell to a programmable electronic load, performing a multi-rate constant current discharge test, recording voltage, current and temperature data, obtaining a dynamic discharge characteristic curve, and performing piecewise linear fitting on the dynamic discharge characteristic curve, calculating the slope and intercept of each segment, and obtaining a dynamic capacity attenuation index; Based on the static equivalent circuit model and dynamic polarization characteristics, a state space equation is constructed, and the internal state of the battery is estimated by an extended Kalman filter algorithm to obtain initial values ​​of the state of charge and the state of health; The static open circuit voltage value, the complex impedance data, the dynamic polarization characteristics, the dynamic discharge characteristic curve, the dynamic capacity decay index, the state of charge and the initial value of the health state are combined to obtain the original battery performance data.

4. The performance testing method of a lithium battery according to claim 1, characterized in that: The performing wavelet transform denoising and feature extraction on the original battery performance data to obtain a static feature vector and a dynamic feature vector includes: Performing wavelet decomposition on the original battery performance data to obtain approximate coefficients and detail coefficients, and calculating the signal-to-noise ratio of each layer based on the detail coefficients to obtain denoising parameters; Performing soft threshold processing on the detail coefficient according to the denoising parameter, and performing wavelet reconstruction in combination with the approximate coefficient to obtain denoised battery performance data; Performing time-frequency analysis on the denoised battery performance data to obtain a time-frequency feature matrix, and performing singular value decomposition based on the time-frequency feature matrix to select left singular vectors corresponding to the first k largest singular values ​​to obtain principal component features; Performing statistical analysis on static parameters in the denoised battery performance data, calculating mean, variance, skewness and kurtosis, and obtaining statistical features; The principal component features are integrated with the statistical features, and the feature weights are calculated by using the Fisher discriminant criterion to obtain a weighted feature vector; Nonlinear dimensionality reduction is performed based on the weighted feature vector to separate static features from dynamic features, thereby obtaining a static feature vector and a dynamic feature vector.

5. The performance testing method of a lithium battery according to claim 1, characterized in that: The step of performing digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data includes: Inputting the static characteristic vector and the dynamic characteristic vector into a programmable gain amplifier, adjusting the gain through a digital potentiometer, and obtaining an initial amplified signal; Performing Fourier transform on the initial amplified signal, calculating the power spectrum density of the signal to obtain frequency distribution characteristics, and filtering the initial amplified signal based on the frequency distribution characteristics to obtain a filtered signal; Inputting the filtered signal into a self-calibration circuit, calculating a gain error and an offset error by comparing with a reference voltage source to obtain calibration parameters, and performing linear calibration on the filtered signal according to the calibration parameters to obtain a calibrated signal; Performing Hilbert transform on the calibrated signal, calculating the instantaneous frequency and instantaneous amplitude of the signal, obtaining time-varying characteristics, and suppressing power frequency interference in the calibrated signal based on the time-varying characteristics to obtain a de-interferenced signal; The interference-removed signal is digitally sampled through an analog-to-digital converter to obtain calibration performance data.

6. The performance testing method of a lithium battery according to claim 1, characterized in that: The calibration performance data is subjected to orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation to obtain multi-dimensional battery parameters, including: Performing a fast Fourier transform on the calibration performance data to obtain a frequency domain signal, dividing the frequency domain signal into a plurality of subcarriers, applying 16QAM modulation to each subcarrier to obtain a modulated subcarrier signal; Performing an inverse fast Fourier transform on the modulated subcarrier signal to obtain a time-domain OFDM symbol, and adding a cyclic prefix before the time-domain OFDM symbol to obtain an orthogonal frequency division multiplexing modulated signal; Input the orthogonal frequency division multiplexing modulation signal into the programmable power supply, perform constant current-constant voltage charging operation on the lithium battery pack, record the voltage, current and temperature data during the charging process, and obtain a charging characteristic curve; Performing a multi-rate pulse discharge operation on the lithium battery pack, setting different discharge rates and pulse widths, recording the voltage drop and recovery curve during the discharge process, and obtaining dynamic response characteristics; Based on the charging characteristic curve and the dynamic response characteristic, the impedance characteristics of the lithium battery under different charge states are measured by electrochemical impedance spectroscopy technology to obtain impedance spectrum data; The charging characteristic curve, the dynamic response characteristic and the impedance spectrum data are subjected to feature extraction and parameter fitting to obtain multi-dimensional battery parameters including internal resistance, polarization parameters, capacity and power characteristics.

7. The performance testing method of a lithium battery according to claim 1, characterized in that: The multi-dimensional battery parameters are input into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection to obtain a battery health status value, including: The multi-dimensional battery parameters are segmented into time series, and a sliding window with a length of T and a step size of S is used to generate N input sequences, each sequence containing M feature dimensions, to obtain a time series feature matrix with a shape of (N, T, M); Input the temporal feature matrix into a multi-scale convolutional layer, wherein the multi-scale convolutional layer comprises three parallel one-dimensional convolutional sub-layers, the convolution kernel size of the first sub-layer is 3, the convolution kernel size of the second sub-layer is 5, the convolution kernel size of the third sub-layer is 7, the step size of each sub-layer is 1, the activation function is ReLU, and each sub-layer outputs 32 feature maps, thereby obtaining three multi-scale temporal feature tensors with shapes of (N, T, 32); Performing maximum pooling operations on the three multi-scale time feature tensors respectively, with a pooling kernel size of 2 and a step size of 1, to obtain three downsampled feature tensors with a shape of (N, T, 32), and concatenating the three downsampled feature tensors in the feature dimension to obtain a fused feature tensor with a shape of (N, T, 96); Input the fused feature tensor into a bidirectional long short-term memory network layer, wherein the bidirectional long short-term memory network layer comprises 128 hidden units, of which 64 are forward and 64 are backward, and a tanh activation function is used to obtain a bidirectional sequence representation tensor with a shape of (N, T, 128); Applying a self-attention mechanism to the bidirectional sequence representation tensor and calculating an attention weight matrix, the specific steps are: performing a linear transformation on the bidirectional sequence representation tensor to obtain a query matrix Q, a key matrix K, and a value matrix V, respectively, and then calculating the dot product of Q and K and dividing it by 8, and then performing a softmax function to obtain an attention weight, and finally multiplying the attention weight by V to obtain a weighted feature representation tensor with a shape of (N, T, 128); The weighted feature representation tensor is converted into a two-dimensional tensor of shape (NT, 128), and then input into a fully connected layer, wherein the fully connected layer includes 64 neurons and uses a ReLU activation function to obtain a high-level abstract feature tensor of shape (NT, 64); Perform a dropout operation on the high-level abstract feature tensor, with a dropout rate of 0.5, to obtain a regularized feature tensor with a shape of (N*T, 64); The regularized feature tensor is input into the output layer, the output layer is a single neuron, and a Sigmoid activation function is used to obtain the battery health status value.

8. A performance testing device for a lithium battery, characterized in that: The performance detection device of the lithium battery comprises: A selection module is used to perform weighted calculation on the remaining capacity, internal resistance, temperature gradient and voltage deviation of the lithium battery cell to obtain a comprehensive score, and select a target measurement point according to the comprehensive score; A measurement module, used to perform static and dynamic dual-mode measurements on the lithium battery cell according to the target measurement point to obtain original battery performance data; An extraction module, used for performing wavelet transform denoising and feature extraction on the original battery performance data to obtain a static feature vector and a dynamic feature vector; A processing module, used for performing digital signal processing on the static feature vector and the dynamic feature vector to obtain calibration performance data; An operation module, used to perform orthogonal frequency division multiplexing modulation and multi-mode charge and discharge operation on the calibration performance data to obtain multi-dimensional battery parameters; The detection module is used to input the multi-dimensional battery parameters into a multi-scale attention bidirectional long short-term memory network to perform battery performance detection to obtain a battery health status value.

9. An electronic device, characterized in that: The electronic device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory so that the electronic device executes the performance detection method of the lithium battery as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the performance detection method of the lithium battery according to any one of claims 1 to 7 is implemented.

Citation Information

Cited By

  • Lithium battery sensitivity test method and system

    CN119165374A

  • Wireless charging power adjusting method and device, equipment and storage medium

    CN119275974A

  • Wireless Charging Power Adjustment Method, Device, Equipment and Storage Medium

    CN119275974B

  • Performance optimization test method and system for new energy battery

    CN120089833A

  • Performance optimization test method and system for new energy batteries

    CN120089833B