A method for improving the cycle performance of lithium-ion batteries based on multi-feature analysis

Abstract

The present invention relates to the technical field of battery management, and specifically discloses a method for improving the cycle performance of lithium-ion batteries based on multi-feature analysis. By real-time monitoring key parameters such as the capacity, internal resistance, charge and discharge efficiency, and temperature fluctuation of the battery, and using advanced algorithms such as dynamic time warping, Kalman filtering, fast Fourier transform, and Haar wavelet transform to calculate the corresponding eigenvalues, combined with machine learning models of gradient boosting decision tree and support vector machine, the present invention can accurately evaluate the risk of electrolyte decomposition and provide early warning. Once a potential risk is detected, the system automatically activates a PID controller to precisely regulate the charge and discharge process and temperature, ensuring that the battery operates under optimal conditions. This comprehensive monitoring and regulation mechanism not only effectively prevents safety hazards such as electrolyte decomposition, but also significantly extends the battery life and improves the overall safety of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and particularly to a method for improving the cycle performance of lithium-ion batteries based on multi-feature analysis. Background Art

[0002] With the rapid development of the renewable energy and electric vehicle industries, lithium-ion batteries, as an efficient and environmentally friendly energy storage device, have been widely used. However, during long-term use, lithium-ion batteries face problems such as capacity decay, increased internal resistance, and electrolyte decomposition, which directly affect the cycle performance and service life of the batteries. Especially under complex working conditions, the battery's operating environment (such as temperature fluctuations) and operating methods (such as charge and discharge efficiency) have a significant impact on the battery's health status. Therefore, how to extend the battery life, improve its safety and reliability through effective monitoring and management means has become the focus of current research. Traditional battery management systems (BMS) usually rely on simple threshold alarm mechanisms and cannot comprehensively and accurately evaluate the battery health status and its change trend.

[0003] The existing technologies have the following deficiencies:

[0004] Most traditional methods use fixed thresholds to determine whether the battery is in a normal working state. This method is difficult to capture the subtle changes in the battery health status and is prone to misjudgment or missed judgment. Secondly, the existing technologies perform poorly in dealing with complex and variable operating conditions. For example, the impacts of temperature fluctuations and changes in charge and discharge efficiency on battery health have not been fully considered and quantified. In addition, existing systems usually do not have predictive maintenance functions and cannot early warn of potential risks, making maintenance measures often lag behind, increasing the probability of battery failure. Finally, most existing solutions lack intelligent data analysis tools and cannot effectively utilize advanced technologies such as machine learning to improve the overall efficiency of the battery management system. These deficiencies limit the effectiveness of the battery management system in practical applications, and there is an urgent need for a more intelligent and accurate method to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for improving the cycle performance of lithium-ion batteries based on multi-feature analysis to solve the problems in the above background.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A method for improving the cycle performance of lithium-ion batteries based on multi-feature analysis includes the following steps:

[0008] S1: During the monitoring period of the battery, monitor and analyze the change degrees of the battery's capacity and internal resistance in real time, and judge whether the electrolyte decomposes according to the monitoring and analysis results;

[0009] S2: According to the evaluation results, obtain the charge-discharge efficiency and temperature fluctuation degree of the battery in real time during the monitoring period, and analyze the influence degree of the charge-discharge efficiency and temperature fluctuation degree on the decomposition of the electrolyte;

[0010] S3: According to the analysis result of the influence degree, perform real-time regulation on the charge-discharge efficiency and temperature of the battery to improve the cycle performance of the battery.

[0011] As a further solution of the present invention: The judgment of whether the electrolyte decomposes specifically includes:

[0012] During the monitoring period of the battery, obtain the capacity data and internal resistance data of the battery in real time according to the time series. According to the change degree of the capacity data and internal resistance data of the battery, calculate the battery capacity characteristic value and the battery internal resistance characteristic value respectively. Use the battery capacity characteristic value and the battery internal resistance characteristic value as the input of the machine learning model. The output of the model is the decomposition score. Judge whether the decomposition score is greater than or equal to the preset threshold. If so, the electrolyte of the corresponding battery decomposes. If not, the electrolyte of the corresponding battery does not decompose.

[0013] As a further solution of the present invention: The process of obtaining the battery capacity characteristic value is as follows:

[0014] Collect the real-time battery capacity data of the battery during the monitoring period to form a time series. At the same time, select the time series of a reference battery as the capacity change mode in the healthy state;

[0015] Construct a local distance matrix for storing the distance between each pair of corresponding points in the target sequence and the reference sequence, which is obtained by calculating the square of the difference between two points;

[0016] Initialize a cumulative distance matrix of the same size with the same starting value; for the first row of the cumulative distance matrix, each element is obtained by adding the cumulative distance of the same column in the previous row to the current local distance; for the first column of the cumulative distance matrix, each element is obtained by adding the cumulative distance of the same row in the previous column to the current local distance. The remaining positions are obtained by adding the current local distance to the minimum cumulative distance among the three adjacent positions on its left, above, and upper left. Starting from the lower right corner of the cumulative distance matrix, trace back to the upper left corner along the direction with the minimum cumulative distance to determine an optimal alignment path connecting the starting point and the end point. The final battery capacity characteristic value is obtained by calculating the average value of all cumulative distances on the optimal alignment path.

[0017] As a further solution of the present invention: The process of obtaining the battery internal resistance characteristic value is as follows:

[0018] Set the initial state estimate to the internal resistance value of the battery measured for the first time, and set an initial error covariance matrix. At each time step, predict the internal resistance value at the current moment based on the state estimate of the previous step, and at the same time update the predicted error covariance matrix. Obtain the internal resistance of the battery at the current moment, denoted as the actual measurement value, and calculate the Kalman gain based on the predicted value and the actual measurement value, update the state estimate, and at the same time update the error covariance matrix. Repeat the prediction and update steps until the monitoring period ends to obtain a series of smoothed internal resistance estimates, and calculate the standard deviation of these smoothed internal resistance values as the battery internal resistance characteristic value.

[0019] As a further solution of the present invention: the establishment process of the machine learning model is as follows:

[0020] Obtain the battery capacity characteristic value and the battery internal resistance characteristic value of the battery during the monitoring period, construct the battery capacity characteristic value and the battery internal resistance characteristic value into a comprehensive feature vector as the input of the machine learning model, use minimizing the error between the predicted decomposition score and the actual decomposition score of the battery as the training objective to train the machine learning model, use historical battery data to train the machine learning model, and according to the trained machine learning model, output the decomposition score of the battery. The machine learning model is a gradient boosting decision tree.

[0021] As a further solution of the present invention: specifically analyzing the influence degree of the charge-discharge efficiency and the temperature fluctuation degree on the decomposition of the electrolyte includes:

[0022] During the monitoring period of the battery, obtain the charge-discharge efficiency data and the temperature data in real time according to the time series. According to the fluctuation ranges of the charge-discharge efficiency and the temperature, calculate the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value respectively. Calculate the influence factor according to the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value, and compare the influence factor with the preset threshold to determine whether the influence factor is greater than or equal to the preset threshold. If so, a serious influence is generated; if not, a slight influence is generated.

[0023] As a further solution of the present invention: the process of obtaining the charge-discharge efficiency characteristic value is as follows:

[0024] During the monitoring period of the battery, obtain the charge-discharge efficiency data in real time according to the time series. Apply the fast Fourier transform to the time series data of the charge-discharge efficiency data to convert the time domain signal into a frequency domain signal, select the cut-off frequency, which is used to distinguish the important low-frequency components from the relatively unimportant high-frequency components, and calculate the sum of the energies of all frequency components below the cut-off frequency; calculate the proportion of the low-frequency energy in the total energy to obtain the charge-discharge efficiency characteristic value.

[0025] As a further solution of the present invention: the process of obtaining the temperature fluctuation characteristic value is as follows:

[0026] During the monitoring period of the battery, in time series, the temperature data of the battery is obtained in real time, and the Haar wavelet transform is performed on the temperature data to decompose the temperature data of the battery into approximation coefficients and detail coefficients. The approximation coefficients are obtained by calculating the average value of adjacent data points, and the detail coefficients are obtained by calculating the difference between adjacent data points.

[0027] For each layer of the detail coefficient set, the squares of all its coefficients are summed to obtain the energy value corresponding to that layer.

[0028] The energy values of the detail coefficients of each layer are multiplied by the corresponding preset weights and then summed to obtain the final temperature fluctuation eigenvalue.

[0029] As a further solution of the present invention: the process of obtaining the influencing factor is as follows:

[0030] The charge-discharge efficiency eigenvalue and the temperature fluctuation eigenvalue of the battery are obtained, and a set of historical data is prepared, including the charge-discharge efficiency eigenvalues, temperature fluctuation eigenvalues of multiple samples, and the corresponding electrolyte decomposition labels. The electrolyte decomposition labels include: electrolyte decomposition and no electrolyte decomposition.

[0031] The support vector machine is used to train the historical data set. For the classification task, the linear kernel is selected to map the input features to a high-dimensional space, and the optimal hyperplane is found to distinguish samples of different classes. For the regression task, the support vector regression model is used to predict the continuous value of the decomposition degree. During the training process, the model parameters are optimized through cross-validation, including: regularization parameters and kernel function parameters.

[0032] The trained support vector machine model is used to predict the electrolyte decomposition influencing factor, and it is judged whether the influencing factor is greater than or equal to the preset threshold. If so, it is recorded as a serious influence. If not, it is recorded as a slight influence.

[0033] As a further solution of the present invention: according to the analysis result of the influence degree, the charge-discharge efficiency and temperature of the battery are adjusted in real time, specifically including:

[0034] Based on the serious influence, the PID controller is started to dynamically adjust the working state of the battery. For the charge-discharge efficiency, the PID controller controls the charge-discharge efficiency according to the deviation between the current efficiency and the target efficiency. For the temperature regulation, the PID controller controls the power output of the cooling or heating device according to the deviation between the current temperature and the target temperature, so that the battery temperature is quickly stabilized within the safe range.

[0035] The beneficial effects of the present invention:

[0036] (1)The present invention realizes the accurate assessment of the battery health state by real-time monitoring of key parameters such as the capacity, internal resistance, charge-discharge efficiency, and temperature fluctuation of lithium-ion batteries, and calculating the corresponding eigenvalues using advanced algorithms such as dynamic time warping, Kalman filtering, fast Fourier transform, and Haar wavelet transform. Combining machine learning models such as gradient boosting decision tree and support vector machine, the present invention can accurately predict the risk of electrolyte decomposition, provide early warning, and ensure that measures can be taken as soon as the potential risk appears. Once the detected decomposition risk exceeds the preset threshold, the system automatically activates the PID controller to precisely regulate the charge-discharge process and temperature, optimize the charging current or voltage and the operating parameters of the cooling / heating device, so as to maintain the operation of the battery under optimal working conditions. This multi-level and intelligent monitoring and regulation mechanism not only effectively prevents safety hazards such as electrolyte decomposition, significantly extends the battery life, but also greatly improves the overall safety and reliability of the system. In addition, through the analysis of the influencing factors of charge-discharge efficiency and temperature fluctuation, the present invention provides solid data support for formulating more scientific and reasonable battery management strategies, further enhancing the battery performance and economic benefits. This innovative solution embodies the concept of comprehensive battery health management from data collection, feature extraction, risk assessment to intelligent regulation, laying an important foundation for the realization of efficient, safe, and long-life lithium-ion battery applications.

[0037] (2)The present invention uses advanced machine learning models of gradient boosting decision tree and support vector machine to comprehensively evaluate the risk of electrolyte decomposition based on multi-dimensional eigenvalues (including the eigenvalues of key parameters such as battery capacity, internal resistance, charge-discharge efficiency, and temperature fluctuation). Through deep learning and pattern recognition technologies, these models can more accurately capture the subtle change trends of the battery health state and provide reliable predictions of future health conditions, thus realizing early warning of potential failures. In particular, through the analysis of the influencing factors of charge-discharge efficiency and temperature fluctuation, the present invention can not only identify the key factors affecting battery performance degradation, but also quantify their influencing degrees, providing a scientific basis for maintenance personnel to take preventive measures in advance. This data-driven method significantly improves the efficiency and accuracy of battery management, making it possible to optimize battery management and maintenance strategies. In addition, the present invention lays a solid foundation for building an intelligent and automated battery management system. Through real-time monitoring and dynamic regulation mechanisms, it ensures that the battery always operates under optimal conditions, greatly improving the overall cycle performance and economic benefits of the battery. Finally, this innovative solution not only extends the battery life, but also reduces the maintenance cost, enhances the reliability and safety of the system, and promotes the development of battery health management technology to a new height. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be further described below with reference to the accompanying drawings.

[0039] Figure 1 It is a specific step flow block diagram of a method for improving the cycling performance of lithium-ion batteries based on multi-feature analysis of the present invention. Specific embodiments

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0041] Please refer to Figure 1 As shown, the present invention is a method for improving the cycling performance of lithium-ion batteries based on multi-feature analysis, including the following steps:

[0042] S1: During the monitoring period of the battery, the changes in the capacity and internal resistance of the battery are monitored and analyzed in real time. According to the monitoring and analysis results, it is judged whether the electrolyte decomposes;

[0043] S2: According to the evaluation results, the charge-discharge efficiency and temperature fluctuation degree of the battery during the monitoring period are obtained in real time, and the influence degree of the charge-discharge efficiency and temperature fluctuation degree on the decomposition of the electrolyte is analyzed;

[0044] S3: According to the analysis results of the influence degree, the charge-discharge efficiency and temperature of the battery are adjusted in real time to improve the cycling performance of the battery.

[0045] In S1, during the monitoring period of the battery, the changes in the capacity and internal resistance of the battery are monitored and analyzed in real time. According to the monitoring and analysis results, it is judged whether the electrolyte decomposes. Specifically, it includes: during the monitoring period of the battery, the capacity data and internal resistance data of the battery are obtained in real time according to the time series. According to the change degree of the capacity data and internal resistance data of the battery, the battery capacity characteristic value and the battery internal resistance characteristic value are calculated respectively. The battery capacity characteristic value and the battery internal resistance characteristic value are used as the input of the machine learning model, and the output of the model is the decomposition score. It is judged whether the decomposition score is greater than or equal to the preset threshold. If so, the electrolyte of the corresponding battery decomposes. If not, the electrolyte of the corresponding battery does not decompose;

[0046] The process of obtaining the battery capacity characteristic value is as follows:

[0047] Collect the real-time battery capacity data of the battery during the monitoring period to form a time series. At the same time, select the time series of a reference battery as the capacity change mode in a healthy or ideal state;

[0048] Construct a local distance matrix to store the distances between each pair of corresponding points in the target sequence and the reference sequence. For each element in the matrix, this distance is obtained by calculating the square of the difference between the two points;

[0049] Initialize a cumulative distance matrix of the same size with the same starting value; for the first row of the cumulative distance matrix, each element is obtained by adding the cumulative distance in the same column of the previous row to the current local distance; for the first column of the cumulative distance matrix, each element is obtained by adding the cumulative distance in the same row of the previous column to the current local distance. For the remaining positions, the value is obtained by adding the current local distance to the cumulative distance of the position with the minimum cumulative distance among its three adjacent positions: the left, above, and upper-left. Starting from the lower-right corner of the cumulative distance matrix, backtrack along the direction that minimizes the cumulative distance to the upper-left corner to determine an optimal alignment path connecting the starting point and the ending point. The points on this path represent how the most similar parts between the two sequences match each other. The final battery capacity eigenvalue is obtained by calculating the average of all cumulative distances on the optimal alignment path;

[0050] The process for obtaining the battery internal resistance eigenvalue is as follows:

[0051] Set the initial state estimate value as the internal resistance value of the battery measured for the first time, and set a relatively large initial error covariance matrix to reflect the uncertainty of the initial estimate. At each time step, predict the internal resistance value at the current moment based on the state estimate value of the previous step, and at the same time update the prediction error covariance matrix to reflect the uncertainty of the predicted value. Obtain the internal resistance of the battery at the current moment, denoted as the actual measurement value, and calculate the Kalman gain based on the predicted value and the actual measurement value. Through the formula Update the state estimate value, and at the same time update the error covariance matrix. Repeat the prediction and update steps until the end of the monitoring period to obtain a series of smoothed internal resistance estimate values. Calculate the standard deviation of these smoothed internal resistance values as the battery internal resistance eigenvalue;

[0052] Among them, represents the th time step, represents the updated state estimate value at the th time step, represents the predicted internal resistance value at the current moment based on the state estimate value at the th time step, represents the th Kalman gain at the th time step, represents the actual measurement value at the

[0053] The process for establishing the machine learning model is as follows:

[0054] Obtain the battery capacity characteristic value and the battery internal resistance characteristic value of the battery within the monitoring period, construct the battery capacity characteristic value and the battery internal resistance characteristic value into a comprehensive characteristic vector, which is used as the input of the machine learning model, and use minimizing the error between the predicted decomposition score and the actual decomposition score of the battery as the training objective to train the machine learning model. Use historical battery data to train the machine learning model, and according to the trained machine learning model, output the decomposition score of the battery. The machine learning model is a gradient boosting decision tree;

[0055] It should be noted that: judge whether the electrolyte of the battery decomposes through the calculated decomposition score of the battery, so as to improve the overall cycle performance of the lithium-ion battery. This method can not only accurately capture the change trend of the battery health state, but also provide an important basis for predicting the future health condition of the battery.

[0056] In S2, according to the evaluation result, obtain the charge-discharge efficiency and the temperature fluctuation degree of the battery within the monitoring period in real time, and analyze the influence degree of the charge-discharge efficiency and the temperature fluctuation degree on the decomposition of the electrolyte, specifically including:

[0057] Within the monitoring period of the battery, obtain the charge-discharge efficiency data and the temperature data in real time according to the time series. According to the fluctuation amplitude of the charge-discharge efficiency and the temperature, calculate the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value respectively. Calculate the influence factor according to the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value, and compare the influence factor with the preset threshold. Judge whether the influence factor is greater than or equal to the preset threshold. If so, a serious influence is generated. If not, a slight influence is generated;

[0058] The process of obtaining the charge-discharge efficiency characteristic value is as follows:

[0059] Within the monitoring period of the battery, obtain the charge-discharge efficiency data in real time according to the time series. Apply the fast Fourier transform to the time series data of the charge-discharge efficiency data to convert the time domain signal into a frequency domain signal, and select the cut-off frequency. The cut-off frequency is used to distinguish important low-frequency components from relatively unimportant high-frequency components, and calculate the sum of the energies of all frequency components below the cut-off frequency; calculate the proportion of the low-frequency energy in the total energy to obtain the charge-discharge efficiency characteristic value;

[0060] The calculation expression of the energy is: Among them, represents the low-frequency energy, represents the th low-frequency component, represents the total number of low-frequency components, represents the frequency domain representation of the charge-discharge efficiency data after the fast Fourier transform;

[0061] The process of obtaining the temperature fluctuation eigenvalue is as follows:

[0062] During the monitoring period of the battery, according to the time series, the temperature data of the battery is obtained in real time. The Haar wavelet transform is performed on the temperature data to decompose the temperature data of the battery into approximation coefficients and detail coefficients. The approximation coefficients are obtained by calculating the average value of adjacent data points, and the detail coefficients are obtained by calculating the difference between adjacent data points.

[0063] For the detail coefficients obtained by each layer of Haar wavelet transform, calculate their energy values. Specifically, for a set of detail coefficients of a certain layer, sum the squares of all its coefficients to obtain the energy value of the corresponding layer.

[0064] The energy value of each layer reflects the signal fluctuation amplitude corresponding to that layer. The energy value of the low-frequency layer reflects the fluctuation on a long time scale, and the energy value of the high-frequency layer reflects the fluctuation on a short time scale.

[0065] Sum the product of the energy values of the detail coefficients of each layer and the corresponding preset weights to obtain the final temperature fluctuation eigenvalue.

[0066] The process of obtaining the influencing factor is as follows:

[0067] Obtain the charge-discharge efficiency eigenvalue and temperature fluctuation eigenvalue of the battery. Prepare a set of historical data, including the charge-discharge efficiency eigenvalues, temperature fluctuation eigenvalues, and corresponding electrolyte decomposition labels of multiple samples. The electrolyte decomposition labels include: electrolyte decomposition occurs and electrolyte does not decompose.

[0068] Use the support vector machine to train the historical data set. For the classification task, select the linear kernel to map the input features to a high-dimensional space and find the optimal hyperplane to distinguish different categories of samples. For the regression task, use the support vector regression model to predict the continuous value of the decomposition degree. During the training process, optimize the model parameters including: regularization parameters and kernel function parameters through cross-validation to improve the generalization ability of the model.

[0069] In the trained support vector machine model, analyze the importance of the charge-discharge efficiency eigenvalue and temperature fluctuation eigenvalue for the electrolyte decomposition prediction result.

[0070] For a new test sample, according to its charge-discharge efficiency eigenvalue and temperature fluctuation eigenvalue, use the trained support vector machine model to predict the electrolyte decomposition influencing factor, and judge whether the influencing factor is greater than or equal to the preset threshold. If so, it is recorded as a serious influence. If not, it is recorded as a slight influence.

[0071] In S3, according to the analysis result of the influence degree, perform real-time regulation on the charge-discharge efficiency and temperature of the battery to improve the battery cycle performance, specifically including:

[0072] Based on the serious impact, the PID controller is activated to dynamically adjust the working state of the battery. For the charge-discharge efficiency, the PID controller controls the charge-discharge efficiency according to the deviation between the current efficiency and the target efficiency, adjusts the charge and discharge current or voltage, and optimizes the charge-discharge process to improve the efficiency and reduce heat generation. For temperature regulation, the PID controller controls the power output of the cooling or heating device according to the deviation between the current temperature and the target temperature, so that the battery temperature can be quickly stabilized within the safe range. The PID controller achieves accurate and stable control effects by quickly responding to the deviation through the proportional term, eliminating the steady-state error through the integral term, and suppressing excessive fluctuations through the derivative term. This process ensures that the battery operates in a high-efficiency and safe state, effectively preventing the risk of electrolyte decomposition, extending the battery life, and improving its reliability.

[0073] The working principle of the present invention: During the monitoring period, the capacity and internal resistance data of the battery are obtained and analyzed in real time. The characteristic value of the battery capacity is calculated through dynamic time warping, the characteristic value of the battery internal resistance is obtained by using Kalman filtering, and these characteristic values are input into the gradient boosting decision tree model to output the decomposition score to judge whether the electrolyte decomposes. Secondly, for the charge-discharge efficiency and temperature fluctuations, the fast Fourier transform and Haar wavelet transform are used to calculate the charge-discharge efficiency characteristic value and temperature fluctuation characteristic value respectively, and then the support vector machine model is used to evaluate the influence degree of these characteristics on the electrolyte decomposition to determine whether there is a serious impact. Finally, according to the analysis result of the influence degree, the PID controller is activated to adjust the charge-discharge efficiency and temperature of the battery in real time: for the charge-discharge efficiency, the PID controller adjusts the charge and discharge current or voltage, and optimizes the charge-discharge process to improve the efficiency and reduce heat generation; for the temperature, the PID controller controls the power output of the cooling or heating device to stabilize the battery temperature within the safe range. Through the three functions of proportional, integral, and derivative, the PID controller achieves accurate and stable control effects, ensuring that the battery operates in a high-efficiency and safe state, effectively preventing the risk of electrolyte decomposition, extending the battery life, and improving its reliability. This comprehensive solution can not only accurately capture the change trend of the battery health state, but also provide an important basis for predicting the future health status of the battery, thus comprehensively improving the cycle performance of lithium-ion batteries.

[0074] The above formulas are all dimensionless and take their numerical values for calculation. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0075] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, or magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0076] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context before and after.

[0077] It should be understood that in various embodiments of the present application, the magnitudes of the sequence numbers of the above processes do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0078] The above has described in detail one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the application of the present invention should still fall within the scope covered by the present invention.

Claims

1. A method for improving the cycle performance of lithium-ion batteries based on multi-feature analysis, characterized in that, It includes the following steps: S1: During the monitoring period of the battery, the capacity and internal resistance change degree of the battery are monitored and analyzed in real time. According to the monitoring and analysis results, it is judged whether the electrolyte decomposes. Specifically, it includes: During the monitoring period of the battery, according to the time series, the capacity data and internal resistance data of the battery are obtained in real time. According to the change degree of the capacity data and internal resistance data of the battery, the battery capacity characteristic value and the battery internal resistance characteristic value are calculated respectively. The battery capacity characteristic value and the battery internal resistance characteristic value are used as the input of the machine learning model, and the output of the model is the decomposition score. It is judged whether the decomposition score is greater than or equal to the preset threshold. If so, the electrolyte of the corresponding battery decomposes. If not, the electrolyte of the corresponding battery does not decompose; S2: According to the evaluation results, the charge-discharge efficiency and temperature fluctuation degree of the battery during the monitoring period are obtained in real time, and the influence degree of the charge-discharge efficiency and temperature fluctuation degree on the electrolyte decomposition is analyzed. Specifically, it includes: During the monitoring period of the battery, according to the time series, the charge-discharge efficiency data and temperature data are obtained in real time. According to the fluctuation amplitude of the charge-discharge efficiency and temperature, the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value are calculated respectively. The influence factor is calculated according to the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value, and the influence factor is compared with the preset threshold. It is judged whether the influence factor is greater than or equal to the preset threshold. If so, a serious impact is generated. If not, a slight impact is generated; The obtaining process of the influence factor is as follows: Obtain the charge-discharge efficiency characteristic value and the temperature fluctuation characteristic value of the battery, and prepare a set of historical data, including the charge-discharge efficiency characteristic values, temperature fluctuation characteristic values of multiple samples and the corresponding electrolyte decomposition labels. The electrolyte decomposition labels include: the electrolyte decomposes and the electrolyte does not decompose; Use the support vector machine to train the historical data set. For the classification task, select the linear kernel to map the input features to a high-dimensional space and find the optimal hyperplane to distinguish different categories of samples; for the regression task, use the support vector regression model to predict the continuous value of the decomposition degree. During the training process, optimize the model parameters through cross-validation, including: the regularization parameter and the kernel function parameter; Use the trained support vector machine model to predict the electrolyte decomposition influence factor, and judge whether the influence factor is greater than or equal to the preset threshold. If so, it is recorded as a serious impact. If not, it is recorded as a slight impact; S3: According to the analysis results of the influence degree, the charge-discharge efficiency and temperature of the battery are regulated in real time to improve the battery cycle performance.

2. A method for improving the cycle performance of a lithium-ion battery based on multi-feature analysis according to claim 1, characterized in that, The obtaining process of the battery capacity characteristic value is as follows: Collect the real-time battery capacity data of the battery during the monitoring period to form a time series, and at the same time select the time series of a reference battery as the capacity change mode in the healthy state; Construct a local distance matrix to store the distance between each pair of corresponding points in the target sequence and the reference sequence, which is obtained by calculating the square of the difference between two points; Initialize a cumulative distance matrix of the same size with the same starting value. For the first row of the cumulative distance matrix, each element is obtained by adding the cumulative distance in the same column of the previous row to the current local distance. For the first column of the cumulative distance matrix, each element is obtained by adding the cumulative distance in the same row of the previous column to the current local distance. For the remaining positions, the element is obtained by adding the local distance at the current position to the cumulative distance of the position with the minimum cumulative distance among its three adjacent positions: the position to the left, the position above, and the position diagonally above to the left. Starting from the lower right corner of the cumulative distance matrix, backtrack along the direction that minimizes the cumulative distance to the upper left corner to determine an optimal alignment path connecting the starting point and the ending point. The final battery capacity eigenvalue is obtained by calculating the average value of all cumulative distances on the optimal alignment path.

3. A method for improving the cycle performance of a lithium-ion battery based on multi-feature analysis according to claim 1, characterized in that, The process for obtaining the battery internal resistance eigenvalue is as follows: Set the initial state estimate value as the internal resistance value of the battery measured for the first time, and set an initial error covariance matrix. At each time step, predict the internal resistance value at the current moment based on the state estimate value of the previous step, and at the same time update the predicted error covariance matrix. Obtain the internal resistance of the battery at the current moment, denoted as the actual measurement value, and calculate the Kalman gain based on the predicted value and the actual measurement value. Update the state estimate value and at the same time update the error covariance matrix. Repeat the prediction and update steps until the end of the monitoring period to obtain a series of smoothed internal resistance estimate values. Calculate the standard deviation of these smoothed internal resistance values as the battery internal resistance eigenvalue.

4. A method for improving the cycle performance of a lithium-ion battery based on multi-feature analysis according to claim 1, characterized in that, The process for establishing the machine learning model is as follows: Obtain the battery capacity eigenvalue and the battery internal resistance eigenvalue of the battery during the monitoring period. Construct the battery capacity eigenvalue and the battery internal resistance eigenvalue into a comprehensive feature vector as the input of the machine learning model. Use minimizing the error between the predicted decomposition score and the actual decomposition score of the battery as the training objective to train the machine learning model. Use historical battery data to train the machine learning model. According to the trained machine learning model, output the decomposition score of the battery. The machine learning model is a gradient boosting decision tree.

5. A method for improving the cycle performance of a lithium-ion battery based on multi-feature analysis according to claim 1, characterized in that, The process for obtaining the charge-discharge efficiency eigenvalue is as follows: During the monitoring period of the battery, obtain the charge-discharge efficiency data in real time according to the time series. Apply the fast Fourier transform to the time series data of the charge-discharge efficiency data to convert the time-domain signal into a frequency-domain signal. Select the cut-off frequency, which is used to distinguish important low-frequency components from relatively unimportant high-frequency components. Calculate the sum of the energies of all frequency components below the cut-off frequency. Calculate the proportion of the low-frequency energy in the total energy to obtain the charge-discharge efficiency eigenvalue.

6. A method for improving the cycle performance of a lithium-ion battery based on multi-feature analysis according to claim 1, characterized in that, The process for obtaining the temperature fluctuation eigenvalue is as follows: During the monitoring period of the battery, obtain the temperature data of the battery in real time according to the time series. Perform Haar wavelet transform on the temperature data to decompose the temperature data of the battery into approximation coefficients and detail coefficients. The approximation coefficients are obtained by calculating the average value of adjacent data points, and the detail coefficients are obtained by calculating the difference between adjacent data points. For each set of detail coefficients of each layer, sum the squares of all its coefficients to obtain the energy value of the corresponding layer. Sum the energy values of the detail coefficients of each layer after multiplying them by the corresponding preset weights to obtain the final temperature fluctuation eigenvalue.

7. A method for improving the cycle performance of a lithium-ion battery based on multi-feature analysis according to claim 1, characterized in that According to the impact degree analysis result, the charge and discharge efficiency and temperature of the battery are regulated in real time, specifically including: Based on a serious impact, start the PID controller to dynamically adjust the working state of the battery. For the charge and discharge efficiency, the PID controller controls the charge and discharge efficiency according to the deviation between the current efficiency and the target efficiency. For temperature regulation, the PID controller controls the power output of the cooling or heating device according to the deviation between the current temperature and the target temperature, so that the battery temperature quickly stabilizes within a safe range.

Images