BMS (Battery Management System) charging management method of battery with echelon utilization of battery core

Abstract

The invention provides a battery cell echelon utilization battery BMS charging management method comprising the following steps: obtaining historical usage record data of each battery, the historical usage record data comprising charging and discharging times, charging and discharging current values, voltage values and temperature values, and constructing a battery state original data set; according to the standardized battery state data set, calculating and establishing a battery health assessment model, a battery capacity attenuation rate, an internal resistance growth rate and charge and discharge efficiency, and calculating according to a preset weight ratio to obtain a battery health score; based on the health scores of the batteries, grouping the batteries by adopting a clustering algorithm, and generating battery grouping data according to a preset threshold value; and after each charging is completed, updating the original data set of the battery state, recalculating the health score of the battery, and dynamically updating the grouping data of the battery.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a BMS charging management method for a battery cell recycling battery. Background Art

[0002] Second-use batteries face many technical challenges during the charging process. First, due to the capacity differences of the batteries themselves, and the different capacity decay rates of each battery as the usage time increases, the capacity of each battery in the battery pack varies. Secondly, during long-term use, the internal resistance of the battery will gradually increase, the charging efficiency will decrease, and the full charge time will be extended. Furthermore, during repeated charging and discharging, the battery is prone to overcharge and over-discharge, which accelerates battery aging and shortens its service life. Finally, the batteries in the second-use scenario usually have complex sources and harsh operating environments, which makes it difficult to control the battery status and even more difficult to accurately estimate the capacity. The above problems are intertwined, making the charging monitoring of second-use batteries face huge challenges. How to track the capacity changes of each battery in real time and realize accurate charging control based on the dynamic estimation results of the capacity to avoid overcharging and over-discharging and extend the battery life is a key technical problem that needs to be solved urgently. This requires comprehensive consideration of the historical data and real-time status of the battery, in-depth analysis of the intrinsic relationship between the battery health status and the charging characteristics, and the use of advanced algorithm models to achieve dynamic estimation of battery capacity and intelligent regulation of the charging process. Summary of the invention

[0003] The present invention provides a BMS charging management method for a battery cell cascade utilization battery, which mainly includes:

[0004] The historical usage record data of each battery is obtained, and the historical usage record data includes the number of charge and discharge times, charge and discharge current values, voltage values ​​and temperature values, and the original data set of battery status is constructed; the original data set of battery status is preprocessed to remove abnormal values ​​of voltage value, temperature value and current value that exceed the preset range, and the data after removing the abnormal values ​​is normalized to generate a standardized battery status data set, and the standardized battery status data set is used as training data to establish and train a battery status classification model; based on the standardized battery status data set, a battery health assessment model is established, and the battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency are predicted, and the battery health score is calculated according to the preset weight ratio; based on the battery health score, the clustering algorithm is used to group the batteries, and the battery grouping data is generated according to the preset threshold; in the charging stage, according to the battery grouping data, the charging current value of the batteries in different groups is determined, and the real-time voltage value and temperature value of the battery during the charging process are obtained. If the voltage value or temperature value exceeds the preset threshold, the charging is interrupted; after each charging is completed, the original data set of battery status is updated, the battery health score is recalculated, and the battery grouping data is dynamically updated.

[0005] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0006] The present invention discloses a BMS charging management method for a battery with a second-life battery cell. By collecting and analyzing the historical battery usage data, a battery status classification model and a health assessment model are established. The method first preprocesses and standardizes the raw data, then trains the classification model using a machine learning algorithm, and calculates the battery health score based on indicators such as the capacity decay rate. Furthermore, the present invention uses a clustering algorithm to group the batteries, and dynamically adjusts the charging strategy according to the grouping situation during the charging process, and monitors the voltage and temperature in real time to ensure safety. After each charge, the system updates the data and re-evaluates the battery status, thereby realizing intelligent management of the battery throughout its life cycle. This method not only improves the battery's efficiency and life, but also enhances the safety of the charging process, providing a reliable management solution for large-scale battery applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 The present invention is a flow chart of a BMS charging management method for a battery cell recycling battery.

[0008] Figure 2 It is a schematic diagram of a BMS charging management method for a battery with a second-life battery cell according to the present invention.

[0009] Figure 3 It is another schematic diagram of a BMS charging management method for a battery with a second-life battery cell according to the present invention. DETAILED DESCRIPTION

[0010] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0011] like Figure 1-3 In this embodiment, a BMS charging management method for a battery cell recycling battery may specifically include:

[0012] S101, obtaining historical usage record data of each battery, the historical usage record data including charge and discharge times, charge and discharge current values, voltage values ​​and temperature values, and constructing a battery status original data set.

[0013] S102, preprocessing the original battery status data set, removing abnormal values ​​of voltage, temperature and current values ​​that exceed a preset range, normalizing the data after removing the abnormal values, and generating a standardized battery status data set; using the standardized battery status data set as training data, establishing and training a battery status classification model.

[0014] Obtain the original data set of battery status, determine whether the voltage value, temperature value and current value are abnormal according to the preset range, and mark them as abnormal values ​​if they exceed the preset range; traverse each data, remove abnormal data according to the abnormal value mark, and obtain the battery status data set after removing the abnormal value. Obtain the battery status data set after removing the abnormal value, and use the maximum and minimum value normalization method to normalize the voltage value, temperature value and current value. The normalization formula is: x'=(x-min) / (max-min), where x is the original value, min and max are the minimum and maximum values ​​of the feature respectively, and x' is the normalized value, ranging from [0, 1]; construct a standardized battery status data set based on the normalized voltage value, temperature value and current value. Use the K-means clustering algorithm to perform cluster analysis on the standardized battery status data set, and determine the optimal number of clusters k by the elbow rule or silhouette coefficient method; divide the data into k clusters, each cluster corresponding to a battery status category. A battery status classification model is established based on the CART decision tree algorithm. The standardized battery status data set is used as training data, and the category label of the clustering result is used as the training target to train the battery status classification model. The hyperparameters of the decision tree are optimized through the grid search method to improve the classification performance of the model. The established battery status classification model is applied to actual scenarios. For the newly collected battery status data, outliers are first removed and normalized, and then input into the trained decision tree model to determine the status category of the battery, realizing real-time monitoring and prediction of the battery status.

[0015] Exemplarily, first, the original data is detected for abnormal values. The preset voltage range is 5V to 2V, the temperature range is 0℃ to 45℃, and the current range is -5A to 5A. Assume that a set of battery data is collected: voltage 5V, temperature 25℃, current 3A. Since the voltage value 5V exceeds the preset range, it is marked as an abnormal value. Other temperature values ​​and current values ​​are within the preset range and are considered normal values. The data containing the abnormal voltage value is removed to obtain the battery status data set after the abnormal value is removed. This is done to avoid the interference of abnormal data on subsequent analysis results and to ensure the accuracy of the model. Next, the remaining data is normalized. Assume that the battery status data set after the abnormal value is removed contains the following three data: voltage 8V, temperature 20℃, current 2A; voltage 0V, temperature 30℃, current 1A; voltage 7V, temperature 25℃, current 5A. The maximum value of the voltage is 8V and the minimum value is 0V; the minimum value of the temperature is 20℃ and the maximum value is 30℃; the minimum value of the current is 1A and the maximum value is 2A. Taking the first data as an example, the normalized value of voltage is (8-0) / (8-0)=1, the normalized value of temperature is (20-20) / (30-20)=0, and the normalized value of current is (2-1) / (2-1)=1. Similarly, the other data are normalized to obtain a standardized battery state data set. The purpose of normalization is to eliminate the influence of different feature dimensions. For example, the unit of voltage is volt, the unit of temperature is Celsius, and the unit of current is ampere, which are not directly comparable. After normalization, the values ​​of all features are between 0 and 1, so that different features can be compared and calculated. Then, the K-means clustering algorithm is used to perform cluster analysis on the standardized data. The core idea of ​​the K-means algorithm is to divide the data points into k clusters, and each data point belongs to the cluster where the cluster center closest to it is located. The optimal number of clusters k is determined by methods such as the elbow rule or the silhouette coefficient. Assuming that the optimal k value is 3, the data is divided into 3 clusters, representing different battery status categories, such as normal state, low power state, high temperature state, etc. The purpose of cluster analysis is to classify similar battery status data into one category for subsequent training of the classification model. Based on the clustering results, the CART decision tree algorithm is used to establish a battery status classification model. The standardized battery status data set is used as training data, and the category label of the clustering result is used as the training target to train the battery status classification model. The core idea of ​​the decision tree algorithm is to divide the data into different leaf nodes through a series of judgment conditions, and each leaf node represents a category. The classification performance of the model can be improved by optimizing the hyperparameters of the decision tree, such as the maximum depth and the minimum number of leaf node samples, through methods such as grid search. For example, the maximum depth is set to 3 and the minimum number of leaf node samples is set to 5. By adjusting these parameters, the best combination of model parameters is found to achieve the highest classification accuracy of the model.Finally, the trained decision tree model is applied to actual scenarios. For the newly collected battery status data, outliers are first removed and normalized, and then input into the trained decision tree model to determine the battery status category, such as normal state, low power state, high temperature state, etc. In this way, real-time monitoring and prediction of battery status can be achieved, battery anomalies can be discovered in time, and corresponding measures can be taken.

[0016] S103. Based on the standardized battery status data set, a battery health assessment model is established to predict the battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency, and the battery health score is calculated according to a preset weight ratio.

[0017] A1. Obtain a standardized battery status data set, preprocess the data set, including removing missing values ​​and outliers, selecting features with a correlation with the battery health status higher than a preset threshold, and normalizing the data set to obtain a preprocessed data set; A2. Based on the preprocessed data set, use the support vector machine algorithm, set the kernel function and penalty coefficient, and train the battery health assessment model; A3. Optimize the hyperparameters of the battery health assessment model through cross-validation and grid search methods; A4. Obtain the status data of the battery to be evaluated, repeat step A1 to preprocess the status data, and obtain preprocessed status data; A5. Input the preprocessed status data into the battery health assessment model to predict the battery capacity decay rate, internal resistance growth rate, and charge and discharge efficiency; A6. A7. Analyze historical battery health data, set the weight ratio of battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency; use weighted average method to obtain a comprehensive battery health score; A8. Associate the battery health score with the battery number and evaluation time information, and store them in the SQLite database; A9. Use the historical battery health assessment data in the SQLite database and use the Prophet library to perform time series prediction; visualize the prediction results and analyze the changing trend of the battery health status; if the predicted result of the battery health status continues to deteriorate and the current battery health score is lower than the preset threshold, it is judged that the battery needs maintenance or replacement; if the predicted result of the battery health status is stable and the current battery health score is higher than the preset threshold, it is judged that the battery is in good health.

[0018] For example, data preprocessing is the first step in building a battery health assessment model. Raw data usually contains missing values, outliers, and irrelevant features, which will affect the accuracy of the model. Take the Pandas library as an example, which provides powerful data cleaning functions. Suppose a battery data set contains data such as voltage, current, and temperature, and some data points are missing. You can use the fillna function of Pandas to fill the missing values ​​with the mean or median. For outliers, assume that the normal range of voltage values ​​is between 2V and 5V, and values ​​outside this range are considered outliers. You can use the conditional filtering function of Pandas to remove outliers. Feature selection is also important. For example, the production date of the battery may be irrelevant to the battery health status and needs to be removed. The Scikit-learn library provides a variety of feature selection methods, such as variance analysis, correlation coefficient analysis, etc. Through feature selection, features with high correlation with the battery health status, such as internal resistance, capacity, etc., can be retained. Finally, in order to eliminate the influence of different feature dimensions, normalization processing is required to scale the data to between 0 and 1. The MinMaxScaler of the Scikit-learn library can achieve this function. Support vector machine (SVM) is a commonly used machine learning algorithm suitable for classification and regression tasks such as battery health assessment. The core idea of ​​SVM is to find an optimal hyperplane to separate data of different categories. The kernel function is the key to SVM, which can map data to a high-dimensional space to make it easier to linearly separable. Common kernel functions include linear kernel, polynomial kernel and radial basis function kernel (RBF). The penalty coefficient C controls the complexity of the model. A larger C value will make the model more inclined to fit the training data and prone to overfitting; a smaller C value will make the model simpler but may underfit. Through methods such as cross-validation and grid search, the optimal kernel function and penalty coefficient can be found to improve the generalization ability of the model. For example, the GridSearchCV function in the Scikit-learn library can be used to search for combinations of different kernel functions and C values ​​and select the parameters with the best performance on the validation set. Evaluating the battery health state requires the same preprocessing operations on the data as the training dataset, such as missing value filling, outlier removal and normalization. By inputting the preprocessed data into the trained battery health assessment model, the predicted values ​​of key indicators such as battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency can be obtained. These indicators reflect the degree of battery aging. For example, the higher the capacity decay rate, the greater the capacity loss of the battery and the worse the battery health status. Battery health status is a comprehensive concept that needs to consider the influence of multiple indicators. In order to comprehensively evaluate the battery health status, the weighted average method can be used. According to the experience of battery health assessment experts and historical data analysis, the weight ratio of each indicator is determined. For example, the weight of capacity decay rate can be set to 5, the weight of internal resistance growth rate to 3, and the weight of charge and discharge efficiency to 2.Assume that the capacity decay rate of a battery is 1, the internal resistance growth rate is 0.5, and the charge and discharge efficiency is 95. Then the health score of the battery is 1*5+0.5*3+95*2=255. Using a database can effectively manage battery health assessment data. SQLite is a lightweight database suitable for storing and managing battery health assessment results. Associate the battery health assessment results with information such as battery number and assessment time, and store them in the SQLite database to facilitate subsequent data analysis and trend prediction. For example, you can create a table named battery_health, which contains fields such as battery number, assessment time, capacity decay rate, internal resistance growth rate, charge and discharge efficiency, and health score. SQL statements can be used to easily query, insert, and update data. Time series prediction can help understand the changing trend of battery health status. The Prophet library is an open source time series prediction library from Facebook, which can handle complex time series patterns such as seasonality and trend changes. Use the battery health score as time series data and use the Prophet library for prediction. For example, you can use the battery health score data for the past year to predict the battery health status for the next three months. By visualizing the prediction results, you can intuitively understand the changing trend of the battery health status and provide decision support for battery maintenance and replacement. Based on the battery health status prediction results and health score, you can determine whether the battery needs maintenance or replacement. Assume that based on historical data and expert experience, the threshold of the health score is set to 7. If the predicted battery health status continues to deteriorate and the current health score is lower than 7, it is recommended to maintain or replace the battery. If the predicted battery health status is stable and the current health score is higher than 7, the battery is considered to be in good health and can continue to be used. Doing so can avoid battery failures, improve battery utilization, and reduce maintenance costs.

[0019] S104: Based on the battery health score, a clustering algorithm is used to group the batteries, and battery grouping data is generated according to a preset threshold.

[0020] Get the health score data of each battery, store it in the MySQL database, and use a table to store the battery number and the corresponding health score; according to the historical battery health score data, use the KMeans function in the Scikit-learn library to select the optimal clustering number parameter through the elbow rule, pre-build a clustering model, and use the Silhouette coefficient to evaluate the quality of the clustering results, and save the clustering model as a pkl file; input the currently acquired battery health score data into the pre-built clustering model for processing, and determine which cluster the health score of each battery belongs to based on the calculation results of the clustering model; set a preset threshold for the battery health score, and use regression analysis to determine the threshold by analyzing the relationship between the historical battery health score and the actual battery life. If the battery health score is higher than the threshold, the battery is divided into a healthy battery group; if the battery health score is lower than the threshold, the battery is The battery is divided into the unhealthy battery group; for the healthy battery group and the unhealthy battery group, the KMeans function in the Scikit-learn library is used for secondary clustering, and the silhouette coefficient and SSE indicators are used to evaluate the clustering effect under different clustering number parameters, and the optimal number of clusters is selected to obtain a more fine-grained battery grouping; the mean and variance of the battery health score in each group are calculated. The higher the mean and the smaller the variance, the higher the similarity within the group. If the mean of a group is lower than the preset threshold, the Z-score method is used to identify outliers by analyzing the distribution of the battery health score in the group, and the outliers are divided into other groups, and the group boundary threshold is dynamically adjusted; the final battery grouping result data is output in the format of a CSV file, which contains the battery number, health score, and group field, and the file is saved in the specified directory for subsequent battery management and maintenance.

[0021] For example, battery health assessment and group management can effectively improve battery utilization and extend service life. First, it is necessary to obtain the health score data of each battery and store it in the database to facilitate subsequent analysis and processing. Here, the MySQL database is selected to create a table to store the battery number and the corresponding health score. For example, if the battery number is Battery001 and its health score is 85, a record (Battery001, 85) is inserted into the database. In order to group the batteries, the KMeans clustering algorithm can be used. Before using the KMeans algorithm, the optimal number of clusters needs to be determined. The elbow rule is a commonly used method to determine the number of clusters. By drawing the SSE (sum of squared errors) curve corresponding to different numbers of clusters and observing the curve change trend, when the change in the number of clusters causes the curve change trend to slow down significantly, the number of clusters at this point is the optimal number of clusters. For example, when the number of clusters increases from 1 to 2, the SSE decreases more; and when the number of clusters increases from 4 to 5, the SSE decreases less. Then, 4 can be selected as the optimal number of clusters. In addition, the Silhouette coefficient can be used to evaluate the quality of the clustering results. The higher the value of the Silhouette coefficient, the better the clustering effect. Save the trained clustering model as a pkl file for subsequent use. By inputting the currently acquired battery health score data into the pre-built clustering model for processing, each battery can be divided into different clusters. For example, the health score of battery Battery001 is 85, which belongs to cluster 1; the health score of battery Battery002 is 60, which belongs to cluster 2. In order to distinguish between healthy batteries and unhealthy batteries, a preset threshold of battery health score needs to be set. The determination of the threshold can be based on historical data and actual experience. For example, by analyzing the relationship between historical battery health scores and actual battery life, it is found that when the battery health score is lower than 70, the actual battery life will be significantly shortened. Therefore, 70 can be set as the preset threshold. According to the preset threshold, the battery can be divided into a healthy battery group and an unhealthy battery group. For example, the health score of battery Battery001 is 85, which is higher than the threshold 70 and belongs to the healthy battery group; the health score of battery Battery002 is 60, which is lower than the threshold 70 and belongs to the unhealthy battery group. In order to further refine the battery grouping, the healthy battery group and the unhealthy battery group can be clustered again. The KMeans algorithm can also be used to select the optimal number of clusters through indicators such as silhouette coefficient and SSE. For example, the healthy battery group can be further divided into two subgroups, and the unhealthy battery group can also be further divided into two subgroups. The mean and variance of the battery health score in each group can be calculated to evaluate the similarity within the group. The higher the mean and the smaller the variance, the higher the similarity within the group.For example, the mean of healthy battery subgroup 1 is 90 and the variance is 5; the mean of healthy battery subgroup 2 is 80 and the variance is 10. This indicates that the battery health status within healthy battery subgroup 1 is more similar. If the mean of a group is lower than the preset threshold, the group needs to be further analyzed. For example, the mean of unhealthy battery subgroup 1 is 65, which is lower than the threshold of 70. By analyzing the distribution of battery health scores within the group, outliers can be identified. For example, the health score of battery Battery003 is 40, which is significantly lower than other batteries. It can be regarded as an outlier and divided into other groups or processed separately. According to the situation of the outlier, the group boundary threshold is dynamically adjusted, for example, the threshold is adjusted to 65. Finally, the battery grouping result data is output in the format of a CSV file and saved in the specified directory. The CSV file contains fields such as battery number, health score, and group to which it belongs. For example, Battery001,85, healthy battery subgroup 1; Battery002,60, unhealthy battery subgroup 1. These data can be used for subsequent battery management and maintenance work. For different battery groups, corresponding charging and discharging strategies and maintenance strategies can be formulated. For example, for healthy battery groups, a higher charging current can be used to shorten the charging time; for unhealthy battery groups, a lower charging current can be used to extend the battery life. In addition, regular inspection plans can be formulated based on the health status of the battery to detect potential problems in time and avoid battery failure. Through these strategies, the battery life can be maximized and the battery utilization efficiency can be improved.

[0022] S105. During the charging phase, the charging current values ​​of batteries in different groups are determined according to the battery grouping data, and the real-time voltage and temperature values ​​of the batteries during the charging process are obtained. If the voltage or temperature value exceeds a preset threshold, the charging is interrupted.

[0023] Obtain battery grouping data, determine the initial charging current value of each battery group, and use it as the reference parameter for charging control; obtain the voltage and temperature values ​​of each battery group in the charging process in real time, compare them with the preset threshold value, and determine whether they exceed the safety range; if the voltage or temperature value of a battery group exceeds the preset threshold, immediately interrupt the charging process of the battery group, record the abnormal data, and trigger the alarm mechanism; for battery groups that do not exceed the threshold, use the PID control algorithm to dynamically adjust the charging current value according to their real-time voltage and temperature values; the PID control algorithm calculates the current adjustment amount according to the voltage and temperature deviation values; use the sliding average filter algorithm to filter the obtained voltage and temperature values Processing: The sliding average filtering algorithm eliminates the influence of short-term noise by calculating the average value of data over a period of time; the filtered voltage and temperature values ​​are used as feature vectors, and whether the battery is faulty is used as a label to build a support vector machine classification model to evaluate the health status of the battery; the charging strategy is optimized according to the evaluation results, and the charging current threshold and charging cut-off voltage parameters are adjusted; the real-time voltage, temperature, charging current key parameters of each group of batteries, and charging abnormalities are uploaded to the cloud database in real time, and SQL statements are used to screen and statistically analyze the uploaded data to discover potential failure modes and optimization opportunities, and the analysis results are fed back to the local control system to dynamically optimize the charging algorithm and strategy.

[0024] Exemplarily, after the battery grouping is completed, the initial charging current value of each group of batteries needs to be determined. For example, the initial charging current of the healthy battery group is set to 2A, and the initial charging current of the unhealthy battery group is set to 1A. This is because healthy batteries can withstand a larger charging current, while unhealthy batteries require a gentler charging method to avoid further damage to the battery. During the charging process, the voltage and temperature of each group of batteries need to be monitored in real time. Assume that the preset voltage threshold is 2V and the temperature threshold is 40°C. If the voltage of a group of batteries exceeds 2V or the temperature exceeds 40°C, the charging of the group of batteries is stopped immediately. For example, if the voltage of a battery in a healthy battery group reaches 3V, the charging of all batteries in the group is stopped immediately, and the abnormal data is recorded, and the alarm mechanism is triggered to notify relevant personnel to handle it. This can prevent battery damage or safety accidents caused by overcharging. For battery groups whose voltage and temperature do not exceed the threshold, the PID control algorithm is used to dynamically adjust the charging current. For example, the current voltage of a healthy battery group is 1V, the temperature is 35°C, and the target voltage is 2V. The PID controller will calculate an adjustment amount, such as increasing the charging current by 1A, based on the difference between the current voltage and the target voltage, the difference between the temperature and the target temperature, and historical data. Through PID control, the charging process can be made smoother and voltage and temperature fluctuations can be avoided. In order to improve data accuracy and reliability, the collected voltage and temperature values ​​are processed using a sliding average filtering algorithm. For example, the sliding window size is set to 5, that is, the average value of the voltage and temperature values ​​of the last five sampling points is taken as the current value. This can effectively filter out short-term noise and make the data smoother, thereby improving the accuracy of PID control. Assuming that the five consecutive voltage values ​​collected by a battery pack are 10V, 12V, 0.8V, 11V, and 0.9V, respectively, the voltage value after sliding average filtering is 10V. The filtered voltage and temperature values ​​are used as feature vectors, and whether the battery is faulty is used as a label to construct a support vector machine (SVM) classification model to evaluate the health status of the battery. For example, batteries with voltage values ​​higher than 15V and temperature values ​​lower than 38°C are marked as "healthy", and other cases are marked as "unhealthy". Use the existing battery data to train the SVM model, and then use the trained model to classify the real-time data to determine whether the battery is faulty. Based on the evaluation results of the SVM model, the charging strategy can be dynamically optimized. For example, if a group of batteries is judged to be "unhealthy", its charging current threshold and charging cut-off voltage are lowered to protect the battery and extend its service life. Conversely, if a group of batteries is judged to be "healthy", the charging current threshold can be appropriately increased to shorten the charging time. All key parameters and abnormal conditions are uploaded to the cloud database in real time. For example, the voltage, temperature, current and abnormal condition records of each group of batteries are uploaded to the cloud database every second. SQL statements can be used to perform statistical analysis on these data.For example, you can query the average voltage, maximum temperature, and number of abnormal conditions of a group of batteries over a period of time. By analyzing this data, you can find potential failure modes and optimization opportunities. For example, if you find that the average voltage of a group of batteries continues to drop, it may indicate that the health of the group of batteries is deteriorating and needs to be maintained or replaced in time. Feedback the analysis results to the local control system to dynamically adjust the charging algorithm and strategy, thereby improving charging efficiency and safety and extending battery life.

[0025] S106. After each charging is completed, the battery status original data set is updated, the battery health score is recalculated, and the battery grouping data is dynamically updated.

[0026] Obtain the charging completion signal to trigger the battery status data update process; obtain the latest battery status raw data from the battery management system and update the battery status data set; use the preset battery health assessment model and the updated battery status data set as input to calculate the health score of each battery; compare the battery health score with the preset threshold to determine the health level of the battery; if the battery health level changes, trigger the battery grouping adjustment process; according to the current health level of the battery, dynamically adjust it to the corresponding battery group; use the K-means clustering algorithm to cluster the updated battery status data set to obtain a new battery grouping result; write the battery health score and grouping result into the relational database to complete the dynamic update and grouping adjustment of the battery status.

[0027] Exemplarily, after obtaining the charging completion signal, the battery status is updated and evaluated. This signal is usually sent by the battery management system (BMS), indicating that the battery is fully charged or has reached a preset charging state. First, the latest battery status raw data is obtained from the battery management system (BMS). These data include key parameters such as voltage, current, temperature, and internal resistance of each battery. For example, the voltage of battery No. 1 can be obtained from the BMS as 2V, the temperature is 25°C, and the internal resistance is 10mΩ; the voltage of battery No. 2 is 15V, the temperature is 26°C, and the internal resistance is 12mΩ. These data are obtained to fully understand the current state of the battery. Next, the latest data obtained is updated to the battery status data set. This data set stores the historical and current status information of all batteries. The data set is updated to maintain the timeliness and accuracy of the information and provide a basis for subsequent health assessment and grouping adjustment. Then, each battery is scored for health according to the preset battery health assessment model. Here, the support vector machine (SVM) model is taken as an example. The SVM model is trained based on historical data, and it can evaluate the health status of the battery based on various parameters. For example, with parameters such as voltage, temperature, and internal resistance as input features, the SVM model will output a score between 0 and 1, with a higher score representing a healthier battery. Assume that the health score of battery No. 1 is 9, the health score of battery No. 2 is 8, and so on. The calculated health score is compared with the preset health threshold to determine the health level of the battery. For example, if the preset health threshold is 8, then batteries with scores above 8 are classified as "healthy" and batteries with scores below 8 are classified as "unhealthy". Therefore, battery No. 1 belongs to the "healthy" level, battery No. 2 also belongs to the "healthy" level, and if battery No. 3 has a score of 7, it belongs to the "unhealthy" level. If the health level of the battery changes, the battery group adjustment process is triggered. This is to ensure that batteries with different health states are divided into different groups for management so as to adopt different charging strategies and protection measures. For example, if battery No. 1 has always been at the "healthy" level, there is no need to adjust the grouping. But if the health level of battery No. 3 changes from "healthy" to "unhealthy", it needs to be adjusted from the "healthy" battery group to the "unhealthy" battery group. The K-means clustering algorithm is used to regroup the updated battery status data set. The K-means algorithm clusters the batteries into different groups based on their various parameters. For example, the battery voltage, temperature, internal resistance and other parameters can be used as clustering features to divide the batteries into three groups: "healthy", "sub-healthy" and "unhealthy". This allows for more detailed management of batteries in different states. Finally, the battery health score and grouping results are written into the relational battery status database MySQL. For example, the health score of battery No. 1 is 9 and the group "healthy" is written to the database; the health score of battery No. 2 is 8 and the group "healthy" is written to the database; the health score of battery No. 3 is 7 and the group "unhealthy" is written to the database.This is done to store battery status information persistently for easy query and analysis. Through this process, dynamic updates and grouping adjustments of battery status can be achieved, thereby better managing batteries, improving charging efficiency and safety, and extending battery life. The information in the database can also be used for subsequent data analysis, such as counting the number of batteries with different health levels, analyzing the changing trend of battery health status, etc., to provide data support for further optimizing charging strategies.

[0028] 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 replace some of the technical features therein by equivalents. 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 BMS charging management method for a battery cell recycling battery, characterized in that: The method comprises: The historical usage record data of each battery is obtained, and the historical usage record data includes the number of charge and discharge times, charge and discharge current values, voltage values ​​and temperature values, and the original data set of battery status is constructed; the original data set of battery status is preprocessed to remove abnormal values ​​of voltage value, temperature value and current value that exceed the preset range, and the data after removing the abnormal values ​​is normalized to generate a standardized battery status data set, and the standardized battery status data set is used as training data to establish and train a battery status classification model; based on the standardized battery status data set, a battery health assessment model is established, and the battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency are predicted, and the battery health score is calculated according to the preset weight ratio; based on the battery health score, the clustering algorithm is used to group the batteries, and the battery grouping data is generated according to the preset threshold; in the charging stage, according to the battery grouping data, the charging current value of the batteries in different groups is determined, and the real-time voltage value and temperature value of the battery during the charging process are obtained. If the voltage value or temperature value exceeds the preset threshold, the charging is interrupted; after each charging is completed, the original data set of battery status is updated, the battery health score is recalculated, and the battery grouping data is dynamically updated.

2. The method according to claim 1, characterized in that The data preprocessing is performed on the original battery status data set, abnormal values ​​of voltage, temperature and current values ​​exceeding a preset range are eliminated, and the data after the abnormal values ​​are eliminated are normalized to generate a standardized battery status data set; Using the standardized battery status dataset as training data, a battery status classification model is established and trained, including: Obtain the original data set of battery status, and determine whether the voltage value, temperature value, and current value are abnormal according to the preset range. If they exceed the preset range, they are marked as abnormal values; Traverse each piece of data, remove abnormal data according to the abnormal value mark, and obtain the battery status data set after removing the abnormal value; Obtain the battery status data set after removing abnormal values, and use the maximum and minimum normalization method to normalize the voltage, temperature and current values. The normalization formula is: x'=(x-min) / (max-min), where x is the original value, min and max are the minimum and maximum values ​​of the original value, respectively, and x' is the normalized value, ranging from [0, 1]; Construct a standardized battery status data set based on the normalized voltage value, temperature value and current value; The K-means clustering algorithm is used to perform cluster analysis on the standardized battery status data set, and the optimal number of clusters k is determined by the elbow rule or silhouette coefficient method; Divide the data into k clusters, each cluster corresponds to a battery status category; A battery status classification model is established based on the CART decision tree algorithm. The standardized battery status dataset is used as training data, and the category label of the clustering result is used as the training target to train the battery status classification model. Optimize the hyperparameters of the decision tree through the grid search method to improve the classification performance of the model; The established battery status classification model is applied to actual scenarios. For the newly collected battery status data, outliers are first eliminated and normalized, and then input into the trained decision tree model to determine the status category of the battery, thereby realizing real-time monitoring and prediction of the battery status.

3. The method according to claim 1, characterized in that The battery health assessment model is established based on the standardized battery status data set, and the battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency are predicted, and the battery health score is calculated according to the preset weight ratio, including: A1. Obtain a standardized battery status data set and preprocess the data set, including removing missing values ​​and outliers, selecting features with a correlation with the battery health status higher than a preset threshold, and performing normalization to obtain a preprocessed data set; A2. Based on the preprocessed data set, the support vector machine algorithm is used to set the kernel function and penalty coefficient to establish and train the battery health assessment model; A3. Optimize the hyperparameters of the battery health assessment model through cross-validation and grid search methods; A4, obtaining the status data of the battery to be evaluated, repeating step A1 to preprocess the status data to obtain preprocessed status data; A5. Input the pre-processed state data into the battery health assessment model to predict the battery capacity decay rate, internal resistance growth rate and charge and discharge efficiency; A6. According to the analysis of battery health history data, the weight ratio of battery capacity attenuation rate, internal resistance growth rate and charge and discharge efficiency is set, and the weighted average method is used to obtain a comprehensive battery health score; A7. Associate the battery health score with the battery number and evaluation time information and store them in a SQLite database; A8. Use the historical battery health assessment data in the SQLite database and the Prophet library to perform time series forecasting; visualize the forecast results and analyze the changing trend of the battery health status; If the predicted result of the battery health status continues to deteriorate and the current battery health score is lower than the preset threshold, it is determined that the battery needs maintenance or replacement; If the prediction result of the battery health status is stable and the current battery health score is higher than the preset threshold, the battery health status is judged to be good.

4. The method according to claim 3, characterized in that The method of grouping batteries based on battery health scores using a clustering algorithm and generating battery grouping data according to a preset threshold includes: Get the health score data of each battery and store it in a MySQL database, using a table to store the battery number and the corresponding health score; Based on the historical battery health score data, the KMeans function in the Scikit-learn library is used to select the optimal clustering number parameter through the elbow rule, a clustering model is pre-built, and the Silhouette coefficient is used to evaluate the quality of the clustering results. The clustering model is saved as a pkl file; The currently acquired battery health score data is input into the pre-built clustering model for processing, and the cluster to which the health score of each battery belongs is determined according to the calculation result of the clustering model; A preset threshold value of the battery health score is set. By analyzing the relationship between the historical battery health score and the actual battery service life, the threshold value is determined by using a regression analysis method. If the battery health score is higher than the threshold value, the battery is classified into a healthy battery group. If the battery health score is lower than the threshold value, the battery is classified into an unhealthy battery group. For healthy battery groups and unhealthy battery groups, the KMeans function in the Scikit-learn library is used for secondary clustering. The silhouette coefficient and SSE index are used to evaluate the clustering effect under different cluster number parameters, and the optimal cluster number is selected to obtain a more fine-grained battery grouping. Calculate the mean and variance of the battery health score in each group. The higher the mean and the smaller the variance, the higher the similarity within the group. If the mean of a group is lower than the preset threshold, analyze the distribution of the battery health score in the group, use the Z-score method to identify outliers, and divide them into other groups, and dynamically adjust the group boundary threshold; The final battery grouping result data is output in the format of a CSV file, which contains the battery number, health score, and grouping field. The file is saved in a specified directory for subsequent battery management and maintenance work.

5. The method according to claim 1, characterized in that In the charging stage, the charging current values ​​of batteries in different groups are determined according to the battery grouping data, and the real-time voltage value and temperature value of the battery during the charging process are obtained. If the voltage value or the temperature value exceeds a preset threshold, the charging is interrupted, including: Obtain battery grouping data, determine the initial charging current value of each battery group, and use it as a reference parameter for charging control; Obtain the voltage and temperature of each battery pack in real time during the charging process, compare them with the preset thresholds, and determine whether they exceed the safety range; If the voltage or temperature of a battery group exceeds the preset threshold, the charging process of the battery group will be interrupted immediately, and the abnormal data will be recorded, and the alarm mechanism will be triggered at the same time; For battery packs that do not exceed the threshold, the PID control algorithm is used to dynamically adjust the charging current value according to the real-time voltage and temperature values; The PID control algorithm calculates the current adjustment amount based on the deviation between voltage and temperature; The obtained voltage and temperature values ​​are filtered using a sliding average filtering algorithm, which eliminates the influence of short-term noise by calculating the average value of data over a period of time. The filtered voltage and temperature values ​​are used as feature vectors, and whether the battery is faulty is used as a label to build a support vector machine classification model to evaluate the health status of the battery. Optimize the charging strategy based on the evaluation results and adjust the charging current threshold and charging cut-off voltage parameters; The real-time voltage, temperature, charging current and charging abnormalities of each battery group are uploaded to the cloud database in real time. SQL statements are used to filter and statistically analyze the uploaded data to identify potential failure modes and optimization opportunities. The analysis results are fed back to the local control system to dynamically optimize the charging algorithm and strategy.

6. The method according to claim 4, characterized in that After each charging is completed, the original data set of battery status is updated, the battery health score is recalculated, and the battery grouping data is dynamically updated, including: Get the charging completion signal and trigger the battery status data update process; Obtain the latest raw data of battery status from the battery management system and update the battery status data set; Using the preset battery health assessment model, the updated battery status data set is used as input to calculate the health score of each battery; Compare the battery health score with the preset threshold to determine the battery health level; If the battery health level changes, the battery grouping adjustment process is triggered; Dynamically adjust the battery to the corresponding battery group based on its current health level; The K-means clustering algorithm is used to cluster the updated battery status data set to obtain a new battery grouping result; The battery health score and grouping results are written into a relational database to dynamically update the battery status and adjust the grouping.

Images