Battery unit anomaly detection method, device, equipment and medium

Abstract

The invention relates to the technical field of data processing, can be applied to business scenes of battery management, financial science and technology, medical health and the like, and discloses a battery unit anomaly detection method, device, equipment and medium, and the method comprises the steps: obtaining static attribute data and operation state parameters of a plurality of battery units, and generating a homogenized battery unit group through dynamic clustering; performing dynamic reference analysis by using the internal resistance data, identifying abnormal battery units and generating a preliminary abnormal battery unit set; classifying the abnormal battery units by using a time sequence analysis model, and generating and confirming an abnormal battery unit list; and generating a battery unit maintenance schedule based on the confirmation list and a preset maintenance knowledge base. According to the invention, through combination of the dynamic reference analysis model and the time sequence analysis model, the abnormal state of the battery unit is accurately identified, a potential problem battery can be found in time, the efficiency and accuracy of a battery monitoring system are improved, and the reliability of the battery at a critical moment is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a battery cell abnormality detection method, device, equipment and storage medium. BACKGROUND

[0002] In the field of monitoring and management of backup lead-acid batteries for data center UPS, the traditional monitoring method mainly relies on setting fixed thresholds for alarm triggering. This method identifies the health status of the battery by monitoring the operating state of the battery, such as internal resistance, voltage, temperature, and especially the change in internal resistance. However, due to the fixed setting of the threshold value and the lack of dynamic adjustment capability, simply relying on this method often cannot timely identify the battery in the early stage of degradation. This technical problem is particularly evident in the application of data centers. In a large-scale battery pack, as the number of batteries increases, the efficiency of manual analysis and monitoring decreases significantly, making it easy to miss the early degradation of the battery. Especially when the internal resistance starts to change, without timely and accurate monitoring means, it is often impossible to identify the trend of gradual degradation of the battery, and only when the internal resistance suddenly rises to the threshold value will an alarm be triggered, at which time the battery capacity has often decreased significantly, or even lost normal working ability.

[0003] In the field of financial technology business, similar monitoring and management problems also exist. Financial institutions usually rely on battery packs in the system to ensure the power supply of critical infrastructure, especially in emergency situations. However, the traditional battery monitoring method is still mainly based on simple threshold alarms, lacking in-depth analysis and dynamic prediction of the battery health status. The degradation of the battery state is often gradual, and if there is a lack of accurate trend identification, potential battery failures cannot be detected in advance, which may lead to the inability to restore financial services in critical situations, thereby affecting the stability and business continuity of the system.

[0004] In the field of medical and health services, many hospitals and medical equipment also rely on lead-acid batteries as emergency power supplies. However, the traditional battery monitoring method has the same problem in medical facilities, especially in hospital life support equipment and data recording systems. The change in internal resistance directly affects the running time and stability of the equipment, and if the monitoring system cannot accurately identify the early degradation of the battery, it may lead to the inability of the equipment to provide power support in critical situations, affecting the continuity and safety of medical services. The traditional internal resistance threshold alarm system not only cannot capture the small changes in internal resistance in time, but also when the internal resistance changes to the alarm threshold, the battery has often entered the degradation stage and cannot be recovered in time.

[0005] In summary, the prior art generally has the problem of fixed threshold setting and low efficiency of manual analysis, which cannot provide accurate and real-time battery health state evaluation. Especially in large-scale applications, manual intervention cannot effectively solve the problem of early identification of potential battery failures, which may lead to battery failure at critical moments, thereby affecting system stability and emergency response capability. Therefore, a more intelligent and accurate battery monitoring method is needed to cope with the changing conditions in battery health management and the management needs of large-scale battery groups. SUMMARY

[0006] The main purpose of the present application is to provide a battery cell anomaly detection method, device, equipment and storage medium, aiming to solve the technical problem that the traditional battery monitoring method only relies on fixed threshold alarm and cannot accurately identify the potential problems of the battery in the early degradation stage, leading to the inability to discover the battery degradation in time and affecting the system stability and emergency response capability.

[0007] To achieve the above purpose, the present application provides a battery cell anomaly detection method, comprising:

[0008] Obtaining static attribute data and running state parameters of a plurality of battery cells;

[0009] According to the static attribute data, the plurality of battery cells are dynamically clustered to generate a homogenized battery cell group;

[0010] In the homogenized battery cell group, based on the internal resistance data in the running state parameters, dynamic reference analysis is performed to preliminarily identify abnormal battery cells, forming a preliminary abnormal battery cell set;

[0011] For each battery cell in the preliminary abnormal battery cell set, a multivariate time series analysis is performed using a time series analysis model to classify the preliminarily identified abnormal battery cells and generate a confirmed abnormal battery cell list;

[0012] Based on the confirmed abnormal battery cell list and a preset maintenance knowledge base, a battery cell maintenance schedule is generated.

[0013] Further, to achieve the above purpose, the present application provides a battery cell anomaly detection device, comprising:

[0014] The data acquisition module is used for obtaining static attribute data and running state parameters of a plurality of battery cells;

[0015] The clustering analysis module is used for dynamically clustering the plurality of battery cells according to the static attribute data to generate a homogenized battery cell group;

[0016] A reference analysis module is configured to perform dynamic reference analysis based on internal resistance data in the operating state parameters within the homogeneous battery cell group, to preliminarily identify abnormal battery cells and form a preliminary abnormal battery cell set;

[0017] A time series analysis module is configured to perform multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set using a time series analysis model, to classify the preliminarily identified abnormal battery cells and generate a confirmed abnormal battery cell list.

[0018] A maintenance plan generation module is configured to generate a battery cell maintenance plan table based on the confirmed abnormal battery cell list and a preset maintenance knowledge base.

[0019] Further, to achieve the above-mentioned purpose, the present application also provides a computer device, which comprises a memory, a processor and a battery cell anomaly detection program stored in the memory and executable on the processor, and the battery cell anomaly detection program implements the steps of the battery cell anomaly detection method when executed by the processor.

[0020] Further, to achieve the above-mentioned purpose, the present application also provides a computer readable storage medium, which stores a battery cell anomaly detection program, and the battery cell anomaly detection program implements the steps of the battery cell anomaly detection method when executed by a processor.

[0021] Beneficial effects: The present application relates to the technical field of data processing, and can be applied to business scenarios such as battery management, financial technology and medical health, and discloses a battery cell anomaly detection method, device, equipment and medium, which comprises: obtaining static attribute data and operating state parameters of a plurality of battery cells, generating a homogeneous battery cell group through dynamic clustering; performing dynamic reference analysis based on internal resistance data within the homogeneous battery cell group, preliminarily identifying abnormal battery cells and generating a preliminary abnormal battery cell set; performing multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set using a time series analysis model, and generating a confirmed abnormal battery cell list; and generating a battery cell maintenance plan table based on the confirmed abnormal battery cell list and a preset maintenance knowledge base. The present application accurately identifies the abnormal state of the battery cell by combining dynamic reference analysis and a time series analysis model, avoids misjudgment caused by a fixed threshold in the traditional method, can timely discover potential problem batteries, improves the efficiency and accuracy of the battery monitoring system, and guarantees the reliability of the battery at critical moments. BRIEF DESCRIPTION OF DRAWINGS

[0022] The present application will be further described below with reference to the accompanying drawings and embodiments, in which:

[0023] Figure 1A schematic diagram of an application environment of a battery cell abnormality detection method in an embodiment of the present application;

[0024] Figure 2 A flowchart of an embodiment of the battery cell abnormality detection method of the present application;

[0025] Figure 3 A schematic diagram of functional modules of a preferred embodiment of the battery cell abnormality detection device of the present application;

[0026] Figure 4 A schematic diagram of a computer device in an embodiment of the present application;

[0027] Figure 5 Another schematic diagram of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0028] It should be understood that the specific embodiments described herein are merely illustrative of the present application and do not limit the present application.

[0029] The battery cell abnormality detection method provided by the embodiments of the present application can be applied in an application environment such as Figure 1 , wherein a user end communicates with a service end through a network. The service end can obtain static attribute data and running state parameters of a plurality of battery cells through the user end, generate a homogenized battery cell group through dynamic clustering, perform dynamic benchmark analysis based on internal resistance data in the homogenized battery cell group, preliminarily identify abnormal battery cells and generate a preliminary abnormal battery cell set, use a time series analysis model to perform multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set, and generate a confirmed abnormal battery cell list. Based on the confirmed abnormal battery cell list and a preset maintenance knowledge base, a battery cell maintenance schedule is generated. The present application accurately identifies the abnormal state of the battery cell by combining dynamic benchmark analysis and a time series analysis model, avoids the misjudgment caused by a fixed threshold in the traditional method, can timely discover potential problem batteries, improves the efficiency and accuracy of the battery monitoring system, and guarantees the reliability of the battery at critical moments. The user end can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices. The service end can be implemented by an independent server or a server cluster composed of multiple servers. The present application will be described in detail below through specific embodiments.

[0030] Please refer to Figure 2 , Figure 2 A flowchart of an embodiment of the battery cell abnormality detection method provided by the present application. It should be noted that although a logical sequence is shown in the flowchart, in some cases, the steps shown or described can be performed in an order different from that herein.

[0031] AsFigure 2 As shown, the battery cell anomaly detection method proposed in this invention includes the following steps:

[0032] S10, acquire static attribute data and operating status parameters of multiple battery cells;

[0033] In this embodiment, when acquiring static attribute data and operational status parameters of multiple battery cells, the static attribute data and operational status parameters play different roles. Static attribute data is the battery's fundamental information, providing insights into its basic configuration and usage history. This data is acquired by the battery management system during the production or installation phase, typically by reading battery tags or querying a database. Operational status parameters, on the other hand, are data collected during actual operation, usually through sensors and monitoring systems in real time. Internal resistance, voltage, temperature, and charge / discharge status are important indicators of battery health, especially changes in internal resistance, which are closely related to battery aging and capacity degradation. These parameters for each battery cell can be obtained using existing sensor devices and real-time monitoring systems. The battery management system periodically collects and stores this data to ensure its usability in subsequent analyses.

[0034] After acquiring the data, standardization and data cleaning can unify data from different sources, making them more consistent and comparable in subsequent analyses. For example, battery internal resistance data may vary at different measurement times and with different measuring devices. Therefore, these data need to be standardized to eliminate the influence of different devices and measurement conditions. Standardization typically employs methods such as z-score standardization or min-max standardization to ensure fair comparison of data between different battery cells.

[0035] In practical applications, acquiring this data typically requires the use of existing battery management systems (BMS) and monitoring equipment. BMS usually integrates sensors that can collect real-time operating status parameters of the battery cells. The battery management system is also responsible for preliminary processing of this data, including data cleaning and preprocessing, to ensure data accuracy and consistency.

[0036] During implementation, the first step is to equip each battery cell with appropriate sensors to collect real-time operating status parameters. These sensors typically include internal resistance sensors, voltage sensors, temperature sensors, and charge / discharge status sensors. Data from these sensors can be transmitted to the battery management system (BMS) for storage and processing via wired or wireless means. Static attribute data is usually obtained by reading battery tags or retrieving it from a database. All data acquisition processes must comply with the standards and specifications of the BMS to ensure data accuracy and real-time performance.

[0037] After data collection, it is common to perform data cleaning on the raw data collected. Data cleaning includes steps such as removing outliers, filling in missing values, etc. to ensure the quality of the data used for subsequent analysis. This process can be done by setting a reasonable threshold range to filter the data and remove data points that do not conform to the actual situation. For example, voltage and internal resistance values may have abnormal values due to equipment failure or transmission errors, which must be cleaned at this time.

[0038] After data cleaning, the next step is to standardize the operating state parameters. This step is to ensure the consistency of the data when analyzing, especially when the parameters of different battery units differ in dimension, standardization can eliminate such differences and make the data of different battery units comparable. Standardization usually uses z-score standardization or min-max standardization method to ensure that different data ranges are converted to the same scale.

[0039] Example: In a battery management system, obtaining static attribute data and operating state parameters of multiple battery units is the basis for battery health monitoring and maintenance. For example, in a data center, the operating state parameters of each UPS backup lead-acid battery include internal resistance, voltage, temperature and charge-discharge state. These data are collected in real time by the battery management system (BMS) for judging the health status of the battery. By obtaining these parameters and combining static attribute data such as battery brand, model, capacity and service time, the performance of the battery can be analyzed in detail. In the early stage of battery performance degradation, by monitoring the trend of internal resistance change in real time, potential problems of the battery can be identified in advance, and maintenance or replacement can be performed to avoid discovering the problem when the battery fails, ensuring that the UPS system in the data center can provide emergency power supply when power is cut off.

[0040] In the battery management system in medical devices, the performance of each medical device battery is crucial for the stable operation of the device. By obtaining the static attribute data (such as battery model, capacity) and operating state parameters (such as battery voltage, temperature, internal resistance, etc.) of the battery unit, the medical device maintenance team can monitor the health status of the battery in real time and identify potential aging problems. For example, for the battery used in a cardiac pacemaker, by monitoring the internal resistance change, temperature fluctuation and other parameters of the battery, abnormalities in the battery can be discovered in time to prevent the device from failing to provide power supply at a critical moment. In this way, the service life of the device can be extended, and replacement plans can be made in advance, thereby improving the safety and treatment effect of patients.

[0041] In a financial data center, UPS batteries are critical equipment to ensure uninterrupted operation of the system. By obtaining static attribute data and operating state parameters of battery units, financial institutions can implement battery health monitoring to ensure that they can provide sufficient power when the system is powered off, avoiding downtime accidents caused by battery failure. For example, by monitoring the internal resistance, temperature, voltage and other data of the battery in real time, the health trend of the battery can be analyzed, potential problems can be identified in advance, and maintenance strategies can be adjusted according to the data change rule. Financial institutions can develop more accurate battery maintenance plans based on these monitoring data to reduce the risk of battery failure and ensure the continuous availability of financial services to avoid the impact of sudden events caused by battery failure on service quality.

[0042] The present embodiment can lay a foundation for subsequent dynamic clustering analysis and anomaly detection by obtaining and cleaning static attribute data and operating state parameters of multiple battery units. Through reasonable data preprocessing, errors and abnormalities in the collection process can be eliminated to ensure the quality and reliability of the data. This processing method not only improves the accuracy of subsequent clustering analysis and anomaly detection, but also reduces the possibility of human intervention and false positives, thereby greatly improving the efficiency of battery monitoring and maintenance.

[0043] S20, dynamically clustering the multiple battery units according to the static attribute data to generate a group of homogeneous battery units;

[0044] In the present embodiment, the operation of dynamically clustering multiple battery units to generate a group of homogeneous battery units is aimed at grouping battery units based on their static attribute data to facilitate more accurate performance analysis, fault prediction and maintenance decision-making. Static attribute data refers to parameters that do not change over time, such as battery brand, model, capacity specification and service life, which provide basic information about the battery itself.

[0045] In this operation, the static attribute data needs to be preprocessed first. This includes encoding and standardizing the static attributes of each battery unit, such as brand, model, capacity and service life. Through encoding and standardization, the differences between different battery models, brands, etc. can be eliminated, so that the subsequent clustering operation can be performed on the same scale.

[0046] Next, a clustering operation is performed based on the processed static attribute data. The purpose of the clustering operation is to divide the battery cells into multiple groups based on similarity, so that the battery cells within the same group are very similar in certain dimensions. For these battery cells, unified monitoring, maintenance, and fault diagnosis can be performed. Common clustering methods include the K-means clustering algorithm, the DBSCAN algorithm, etc. Here, we use distance measurement to cluster the battery cells. By optimizing the grouping of the battery static attribute data based on distance, we can ensure that the battery performance and feature similarity within the homogeneous battery cell group are maximized.

[0047] After the clustering operation, further verification is needed to ensure that the batteries within each battery cell group are uniform in function. This step is to ensure the accuracy of the health monitoring and maintenance solutions based on these battery cell groups. By performing attribute consistency verification operations, we can ensure that the battery cells within the homogeneous battery cell group have no significant deviation in performance, thereby improving the accuracy of analysis and fault diagnosis.

[0048] In actual implementation, the preprocessing of static attribute data includes encoding all categorical data (such as brand, model) into numerical data. Numerical data (such as capacity, service life) needs to be standardized. Standardization methods (such as Z-score standardization or Min-Max standardization) can be used to convert data to a unified scale, avoiding the impact of large value range differences on clustering results.

[0049] The choice of clustering method can be adjusted according to actual conditions. For small data sets, the classic K-means clustering method can be selected, with appropriate cluster number (K value) set. For larger data sets, or when you want to find arbitrary shape clustering structure, you can use the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, which does not need to set the number of clusters in advance, and can better handle noise data.

[0050] After the clustering results are obtained, attribute consistency verification is needed. For example, by calculating the average value of the internal resistance, temperature, voltage, etc. of the batteries within each battery cell group, the attribute difference of the batteries within the group is checked. If there is a large difference in the attributes of the batteries within a group, the group may not be suitable as a homogeneous group and needs to be adjusted or re-divided.

[0051] Example: In a data center battery management system, the brand, model, capacity, and service life of each UPS battery unit are static attributes that directly affect the health and performance of the battery. After clustering battery units based on static attribute data, the data center can group batteries with similar characteristics together for unified monitoring and maintenance. If an abnormal internal resistance occurs in a battery pack, the system can immediately determine the health status of all similar batteries in the pack and quickly perform maintenance or replacement, improving overall operational efficiency and avoiding sudden downtime.

[0052] In medical health device battery management, clustering battery units using static attribute data can help identify potential risks of battery failure in advance. In this way, the battery can be monitored while the device is running normally to ensure that the device battery is in the best working condition and avoid device failure or medical service interruption due to battery failure.

[0053] In the financial field, by clustering the static attributes of the battery, the battery units can be more effectively managed at multiple key data points. For example, the capacity and service life of the battery are the main factors affecting battery performance. By grouping these battery units, the financial data center can better monitor the health of the battery and ensure the reliability and stability of the uninterruptible power supply system.

[0054] This embodiment can classify battery units with similar performance into the same group through dynamic clustering of static attribute data, thereby improving the efficiency and accuracy of subsequent analysis. In data center, medical health and financial field, etc. Application scenarios, through clustering operations, more accurate battery health monitoring and fault diagnosis can be performed, thereby optimizing battery maintenance plans, reducing system failure risks, and ensuring the continuous and reliable operation of critical equipment.

[0055] S30, in the homogenized battery unit group, based on the internal resistance data in the running state parameter, dynamic benchmark analysis is performed to preliminarily identify abnormal battery units and form a preliminary abnormal battery unit set;

[0056] In this embodiment, in the homogenized battery unit group, the operation of dynamic benchmark analysis based on internal resistance data in the running state parameter aims to identify abnormal battery units and form a preliminary abnormal battery unit set according to the change of battery internal resistance. Internal resistance is an important parameter for evaluating battery health, and an increase in internal resistance will directly affect the charging and discharging efficiency and service life of the battery, so dynamic benchmark analysis of internal resistance is a key step in battery health management.

[0057] First, before performing dynamic benchmark analysis, it is necessary to obtain real-time performance information of the battery unit through the internal resistance data in the running state parameters. Internal resistance data is an important indicator reflecting the health status of the battery, so it is necessary to collect the internal resistance value of the battery in real time through a high-frequency monitoring system. Through the collected data, the internal resistance change curve of the battery can be generated, which can be used for comparative analysis and benchmark construction.

[0058] The core of dynamic benchmark analysis is to generate a dynamic benchmark curve based on the real-time collected internal resistance data. The benchmark curve is the ideal internal resistance range obtained by statistical analysis of the internal resistance data of the battery unit within a certain period of time. Through comparative analysis of the benchmark curve, it can be determined whether the internal resistance change of the battery is abnormal, and further identify those battery units whose internal resistance values deviate from the normal range. This analysis process can be compared among multiple battery units to ensure that each battery unit can be reasonably positioned in the group.

[0059] When initially identifying abnormal battery units, the criteria generally include internal resistance increase, frequency of internal resistance values exceeding the benchmark range, and duration of abnormal fluctuations. These factors help identify battery units that gradually degrade in performance or suddenly fail under certain special operating conditions.

[0060] The key operation of generating a preliminary set of abnormal battery units is to collect abnormal battery units after determining the internal resistance abnormal battery units to form a set for subsequent fault analysis, monitoring and maintenance work. The set not only includes internal resistance abnormal battery units, but also may include some other potential fault units identified by the abnormal detection model.

[0061] In the implementation process, first, the high-frequency data acquisition module continuously monitors the internal resistance data of each battery unit. These data can be obtained in real time through sensors and uploaded to the battery management system for processing. In the data processing link, first, the internal resistance data of each battery unit is preprocessed, including removing abnormal noise to ensure the accuracy of the data.

[0062] Next, the system will generate a dynamic benchmark curve based on the internal resistance data of each battery unit based on a certain time window. The construction of this benchmark curve usually uses a sliding window algorithm to smooth the data, so that the fluctuations of historical data will not have a large impact on the benchmark curve. By calculating the mean and standard deviation of the internal resistance of the battery in each time window, the generated benchmark curve will reflect the normal performance fluctuation range of the battery.

[0063] Then, the system compares the current internal resistance data of each battery cell with the reference curve, and uses a set threshold to determine which battery cells have internal resistance beyond the normal range. These out-of-range batteries will be marked as abnormal battery cells and will be part of the preliminary abnormal battery cell set. At this time, the data-driven anomaly detection algorithm can also further analyze the internal resistance fluctuation pattern to identify those battery cells that show progressive degradation or sudden failure characteristics.

[0064] Example: In a data center battery management system, the internal resistance change of the battery is a key factor in evaluating battery health. When the internal resistance value of the battery exceeds the preset reference value, the system will automatically mark it as an abnormal battery cell and perform subsequent analysis. Through this method, the operation and maintenance team of the data center can monitor the health status of the battery in real time, identify those batteries with declining performance in a timely manner, and take preventive measures such as maintenance or replacement to ensure that the UPS system can always provide reliable power supply.

[0065] In medical health devices, the stability of the battery is crucial for the continuous operation of the device. Through dynamic reference analysis of internal resistance data, the system can identify battery cells that pose potential failure risks in a timely manner, thereby avoiding interruptions in medical device operation due to battery failure and ensuring normal operation of the device at critical times.

[0066] In the financial sector, UPS batteries used by important institutions such as banks and insurance companies need to ensure long-term stable operation. Through real-time monitoring and dynamic reference analysis, financial institutions can identify problematic battery cells in advance, develop reasonable maintenance plans, reduce the risk of sudden downtime, and ensure the continuity of critical services.

[0067] This embodiment can more accurately identify battery cells that may fail due to degradation, overuse, or environmental factors by monitoring the internal resistance changes of the battery in real time and generating a dynamic reference curve. It can help maintenance personnel identify and address potentially risky batteries before the problem becomes serious, avoiding device downtime due to battery failure, thereby effectively improving the efficiency and reliability of battery management.

[0068] S40, for each battery cell in the preliminary abnormal battery cell set, use a time series analysis model for multivariate time series analysis to classify the preliminary identified abnormal battery cells and generate a confirmed abnormal battery cell list;

[0069] In this embodiment, when processing each battery unit in the preliminary abnormal battery unit set, the operation of adopting a time series analysis model for multivariate time series analysis is to help distinguish and confirm abnormal battery units by jointly analyzing the historical data of multiple battery performance parameters such as internal resistance, voltage, temperature, etc. Compared with the traditional method of relying on single indicator fault detection, the time series analysis model can more comprehensively and accurately identify the classification of abnormal battery units.

[0070] First, it is necessary to extract the multi-dimensional historical data of each battery unit from the preliminary abnormal battery unit set. The historical data of the battery unit includes internal resistance change, charging and discharging condition, voltage fluctuation, temperature change, etc. These data are often collected by real-time monitoring system, and these data of each battery unit form a multi-dimensional time series data sequence. Input these historical data sequences into the time series analysis model, the model can accurately judge the state of the battery unit is normal, deteriorated or already failed by learning and analyzing the dynamic change pattern of these data.

[0071] The core role of the time series analysis model is to classify based on time series data, especially when dealing with the historical data of multiple battery units, it can effectively identify those battery units that show abnormal fluctuations or long-term decline trends. Time series models such as long short-term memory (LSTM) network or convolutional neural network (CNN) can capture long-term dependencies and short-term fluctuation patterns in data when processing time series data, thereby distinguishing abnormal states caused by multiple factors such as internal resistance, temperature, voltage, etc.

[0072] During model training, the labeled normal battery unit and abnormal battery unit data are used, and the trained model can predict the future health status of the battery unit according to its real-time data. By analyzing the time series data of each battery unit, the model classifies it into "normal", "deteriorated" or "failed" categories, and finally forms a list of confirmed abnormal battery units.

[0073] In the implementation process, first, from the preliminary set of abnormal battery units, the historical operation data of each battery unit is obtained, especially the information such as internal resistance change, voltage fluctuation and temperature change. These data can be obtained in real time through the data acquisition module in the battery management system (BMS), and cleaned and standardized through the preprocessing algorithm to ensure data quality. Next, these processed time series data are input into the time series analysis model. The time series analysis model such as LSTM network can capture the long and short term dependence relationship and abnormal fluctuation pattern in the time series by modeling the historical data. In the model training stage, a part of historical data (including battery data labeled as "normal" and "abnormal") is usually used for supervised learning to learn the rules in the data by optimizing the loss function. The trained time series analysis model can be used for real-time battery unit state classification. Whenever the data of a battery unit is updated, the system will automatically classify it using the time series analysis model and classify it as "normal" or "abnormal". The state of each battery unit after analysis will be added to the list of confirmed abnormal battery units, helping the operation and maintenance team to make further processing decisions.

[0074] Example: In the data center UPS battery management, the data such as internal resistance, temperature and voltage of the battery are collected in real time by the monitoring system and transmitted to the battery management system. The historical data of each battery unit is input into the time series analysis model for training, and the model can classify the health status of the battery according to these data. For example, when the internal resistance and temperature values fluctuate abnormally at the same time, the model can identify that these fluctuations are caused by battery degradation and mark the battery as abnormal, and add it to the list of confirmed abnormal battery units in time. Through this method, the operation and maintenance team of the data center can identify the deteriorating battery in advance and maintain or replace it before the battery fails seriously, avoiding sudden downtime and ensuring the efficient operation of the data center.

[0075] In the field of medical equipment, battery is the key power source of medical equipment, and it is crucial to ensure the normal operation of the battery. Through multivariate time series analysis, medical equipment managers can identify potential battery failures in advance and replace problem batteries in time to avoid equipment downtime due to battery problems and ensure power supply during patient treatment.

[0076] In the financial field, the battery management system of banks and insurance companies also faces the same problem. Through accurate battery state classification, battery degradation can be identified in advance to avoid sudden power failure and ensure the continuous operation of critical business.

[0077] The embodiment can accurately classify abnormal battery units through a multivariate time series analysis model, significantly improving the identification accuracy of abnormal battery units. Compared with traditional detection methods based on a single internal resistance indicator or manual threshold, the method can not only identify battery units that have deteriorated early but have not reached the warning threshold, but also accurately identify problem batteries when battery performance deviates in multiple dimensions. In this way, the operation and maintenance team can handle the problem before the battery fails, effectively extend the service life of the battery, reduce downtime risks, and improve the intelligent level of battery management.

[0078] S50, generating a battery unit maintenance schedule based on the confirmed abnormal battery unit list and the preset maintenance knowledge base.

[0079] In the embodiment, generating a battery unit maintenance schedule based on the confirmed abnormal battery unit list and the preset maintenance knowledge base is a key operation to ensure timely maintenance and replacement of battery units. The core of this step is to combine the health status data of the battery units and the existing maintenance strategy library to develop a structured and executable maintenance plan.

[0080] First, the health status of each battery unit in the confirmed abnormal battery unit list has been classified and identified as "abnormal" or "failed". For each abnormal battery unit, its static attribute data, including brand, model, capacity, service time, and other information, need to be associated first. These static attribute data provide detailed background information for the battery units, which helps to determine the maintenance needs and priorities of each battery unit.

[0081] Next, the maintenance strategy matching operation is performed in combination with the existing maintenance strategies in the preset maintenance knowledge base. The preset maintenance knowledge base usually contains standard operating procedures (SOP) for battery maintenance, maintenance cycles, processing priorities, and measures for various faults. For example, when the internal resistance of some battery units changes beyond the set threshold, the system can find appropriate measures from the maintenance knowledge base, such as deep discharge, adjusting the charging strategy, or replacing the battery.

[0082] Based on the matched maintenance strategies, the system next fills in the corresponding data fields according to the preset maintenance schedule template to generate a battery unit maintenance schedule. The maintenance schedule usually includes the identifier of each battery unit, the health status, the maintenance measures to be performed, the maintenance priority, the execution time, and the responsible person, etc. These data not only help to optimize the maintenance process of the battery units, but also improve work efficiency and ensure that problem battery units are handled in a timely manner.

[0083] In the implementation process, first, the identifier of each battery unit is obtained from the list of abnormal battery units, and the static attribute data thereof is retrieved from the database according to the identifiers. These static attribute data include but are not limited to the brand, model, capacity and service life of the battery, which are basic information and are crucial for subsequent determination of the maintenance priority and applicable strategy of the battery. Next, in combination with the preset maintenance knowledge base, the system automatically matches the maintenance strategy of each abnormal battery unit. The maintenance strategy includes various possible processing measures, such as adjusting the charging mode, deep discharging, heating, discharging cycle, replacing the battery, etc. These strategies in the maintenance knowledge base are closely related to the static attributes and abnormal state of the battery unit. For example, if the internal resistance value of the battery is high and the number of charging times is also high, the system may recommend replacing the battery. After matching the appropriate maintenance strategy, the system fills in the data according to the preset maintenance schedule template to generate a battery unit maintenance schedule. The contents in the schedule include the identifier of the battery unit, the abnormal type, the maintenance priority, the executed maintenance measure, the expected execution time, the person in charge, etc. Each data in the schedule is automatically filled according to the actual situation of the battery unit, and is sorted according to the priority and execution time of the maintenance strategy. In this way, the battery management personnel can clearly view the battery units that need to be maintained, and ensure that each battery unit can be timely and properly processed.

[0084] Example: In the UPS battery management of data center, the internal resistance change, temperature fluctuation and voltage change data of the battery unit are monitored and stored in real time by the battery management system (BMS). When the internal resistance of the battery unit exceeds the threshold value, the system automatically identifies the battery as "abnormal" and extracts the corresponding maintenance strategy from the preset maintenance knowledge base according to the static attribute data of the battery, such as the brand, model, capacity, service life, etc. For some relatively aged batteries, the system may recommend deep discharging, while for batteries with a significantly increased internal resistance value, the system may recommend replacing the battery. Finally, the system automatically generates a maintenance schedule, listing all the battery units to be maintained and the corresponding maintenance measures and execution time, ensuring that the operation and maintenance personnel can timely and accurately process each abnormal battery unit.

[0085] In the battery management of medical devices, the internal resistance change and temperature data of the battery are collected in real time by sensors. If the battery has an abnormal internal resistance, the system will generate a maintenance plan according to the static attribute data of the battery, such as the service life and working environment. For example, for a long-term unused backup battery, the system may recommend regular charging and discharging cycles to restore its capacity. The generated maintenance schedule helps device managers to maintain the battery in a timely manner, preventing the device from shutting down due to battery failure, and thus ensuring the normal operation of the device.

[0086] In the financial field, especially in the management of backup batteries in places such as banks, the health of the battery directly affects the stability of the key system. Through the automatic generation of maintenance plans, the system can effectively monitor the battery and provide timely maintenance recommendations, avoiding service interruptions caused by battery failures and ensuring business continuity.

[0087] The embodiment automatically generates a battery unit maintenance schedule by combining the abnormal state of the battery unit with the strategy in the maintenance knowledge base, thereby reducing manual intervention and ensuring timely maintenance and replacement of the battery unit. The maintenance plan can be automatically generated based on the health of the battery, improving the accuracy and efficiency of battery management while reducing unnecessary maintenance work. The structured output of the maintenance plan also makes the battery management work more standardized and efficient, greatly reducing the impact of battery failure on equipment and systems.

[0088] The present application relates to the field of data processing technology, which can be applied to battery management, financial technology, medical health and other business scenarios, and discloses a battery unit anomaly detection method, device, equipment and medium, comprising: acquiring static attribute data and running state parameters of a plurality of battery units, and generating homogenized battery unit groups through dynamic clustering; within the homogenized battery unit group, performing dynamic benchmarking based on internal resistance data, preliminarily identifying abnormal battery units and generating a preliminary abnormal battery unit set; for each battery unit in the preliminary abnormal battery unit set, performing multivariate time series analysis using a time series analysis model and generating a confirmed abnormal battery unit list; and generating a battery unit maintenance schedule based on the confirmed abnormal battery unit list and a preset maintenance knowledge base. The present application accurately identifies the abnormal state of the battery unit by combining dynamic benchmarking and a time series analysis model, avoids misjudgment caused by fixed thresholds in traditional methods, can timely discover potential problem batteries, improves the efficiency and accuracy of the battery monitoring system, and ensures the reliability of the battery at critical moments.

[0089] In one embodiment, the above step S10 comprises:

[0090] S101, collecting the running state parameters of each battery unit, including internal resistance, voltage and temperature;

[0091] S102, acquiring the static attribute data of each battery unit, including brand, model, capacity and service time;

[0092] S103, performing a data cleaning operation on the running state parameters to generate cleaned running state parameters;

[0093] S104, performing a time alignment operation on the cleaned running state parameters to generate time-aligned running data;

[0094] S105, performing a statistical aggregation operation on the time-aligned running data to generate running state aggregated data;

[0095] S106, storing the running state aggregated data and the static attribute data.

[0096] In this embodiment, obtaining and processing the static attribute data and running state parameters of the battery are the basic steps to ensure real-time mastery of the battery health status. Specifically, the static attribute data and running state parameters of the battery cell are usually collected in real time by multiple sensors and data acquisition devices. The static attribute data includes brand, model, capacity, and service time, etc. which do not change with time, while the running state parameters such as internal resistance, voltage, and temperature are the dynamic embodiment of the battery health status.

[0097] Firstly, for each battery cell, the obtained running state parameters include internal resistance, voltage, and temperature. Internal resistance is an important indicator reflecting the health status of the battery, and its change directly affects the performance and life of the battery. Voltage and temperature affect the charging and discharging efficiency and safety of the battery. The process of obtaining these parameters is usually completed by sensors built-in the battery cell and external monitoring system, and the data is transmitted and stored through Internet of Things devices.

[0098] Next, the obtained raw running state parameters need to go through data cleaning operation. This operation is mainly to deal with the noise, missing data or outliers that may exist in the collection process, to ensure the accuracy and consistency of the data. The cleaned data is the basis for further analysis and processing, which helps to improve the reliability of the analysis results.

[0099] The cleaned data usually also needs to perform time alignment operation. The purpose of time alignment operation is to ensure that the data collected from different battery cells or sensors are aligned in time, so that the parameter data from different devices or different time points can be effectively compared and aggregated. The time-aligned data provides a unified time basis for subsequent statistical analysis and data mining.

[0100] Statistical aggregation operation is a key step of summarizing and analyzing the time-aligned data. In this process, by statistically processing the data of internal resistance, voltage, and temperature, etc., the aggregated data of each battery cell is generated. These aggregated data include the maximum value, mean value and minimum value per hour, which can reflect the health status and working performance of the battery cell in different time periods. The maximum value, mean value and minimum value respectively provide the extreme range, average level and lowest point of the battery performance fluctuation, which provides a strong basis for the maintenance and replacement decision of the battery.

[0101] Finally, all generated operational state aggregated data and static attribute data will be stored. Data storage can use traditional database systems or cloud platform-based storage systems to facilitate subsequent queries, analysis, and historical data comparison. These data storage will become an important basis for battery monitoring and maintenance decisions.

[0102] The embodiment can achieve accurate monitoring of the health status of the battery cells by acquiring, cleaning, aligning, and aggregating the static attribute data and operational state parameters of multiple battery cells. This series of operations can effectively eliminate noise and inconsistencies in the data, ensuring the reliability and accuracy of the analysis results. Through time alignment and statistical aggregation, not only can the current state of the battery be fully understood, but also the small fluctuations in battery performance can be discovered, helping to identify potential failure risks in a timely manner. In addition, storing the cleaned data and aggregated data provides a comprehensive historical data record, facilitating long-term monitoring and optimization of the battery management system, improving battery usage efficiency and extending the service life of the battery, ultimately reducing sudden power failure events caused by battery failure, and ensuring the reliability of the system.

[0103] In one embodiment, the above step S20 comprises:

[0104] S201, performing category attribute encoding and numerical attribute standardization processing on the static attribute data to generate standardized static attribute data;

[0105] S202, performing feature selection operation on the standardized static attribute data to determine clustering feature data;

[0106] S203, performing distance optimization grouping operation on the clustering feature data to generate initial battery cell groups;

[0107] S204, performing attribute consistency verification operation on the initial battery cell groups to generate homogenized battery cell groups;

[0108] S205, recording the division results of the homogenized battery cell groups.

[0109] In the embodiment, dynamic clustering of multiple battery cells is performed based on static attribute data to group battery cells with similar performance, which helps to improve the efficiency and accuracy of battery management. The core of the clustering operation is to process the static attributes of the battery cells so that battery cells with similar features can be grouped together, facilitating subsequent monitoring, analysis, and maintenance.

[0110] First, the static attribute data is processed with category attribute encoding and numerical attribute standardization. Category attribute encoding refers to converting the category features of battery cells (such as brand, model, etc.) into numerical data that can be processed by computers. For example, the brand attribute can be realized by assigning unique coding values to different brands, while the model, capacity, and other attributes can be directly processed numerically. Numerical attribute standardization is to eliminate the differences between different attributes due to different dimensions, ensuring that each attribute is compared in the same dimension, thereby avoiding interference with the clustering results. For example, capacity is usually measured in mAh (milliampere hours), while voltage is measured in volts (V). Standardization will unify the value range of these numerical attributes.

[0111] Second, the standardized static attribute data is processed with feature selection. Feature selection aims to filter out the most relevant features for clustering from all attribute data, which can improve the accuracy of clustering and reduce computational complexity. Feature selection analyzes the correlation between each attribute and the target variable, or uses algorithms such as principal component analysis (PCA) to remove redundant attributes, ensuring that clustering only relies on the most meaningful data.

[0112] Next, the selected clustering feature data is processed with distance optimization grouping. The purpose of this operation is to group battery cells according to certain distance metrics (such as Euclidean distance or Manhattan distance). The key to distance optimization grouping is to determine which battery cells should be divided into the same group by calculating the similarity between different battery cells. This process can classify battery cells into appropriate groups according to their characteristics, ensuring that the clustering results of battery cells have practical significance.

[0113] Then, the generated initial battery cell groups are processed with attribute consistency verification. This step is a further verification of the preliminary grouping results, ensuring that the battery cells in each group have consistency in attributes. For example, within the same battery cell group, attributes such as capacity and voltage should be similar. If there are significant differences in these attributes within a group, the grouping needs to be adjusted until the attributes of the battery cells in the group meet the consistency requirements.

[0114] Finally, the division results of the homogenized battery cell groups are recorded. The recording of the division results provides a traceable data basis for subsequent analysis, monitoring, and maintenance. By recording the group to which each battery cell belongs, real-time tracking of the battery group can be performed during subsequent monitoring, and battery cells with abnormal performance can be identified in a timely manner.

[0115] In different embodiments, the above clustering operation can be implemented through different technical means. For example, for the static attribute data of the battery cells, a distance optimization grouping operation can be performed using the K-means clustering algorithm in machine learning. The K-means algorithm divides the data points into K clusters through iterative optimization, and the data points in each cluster are as similar as possible. This algorithm performs well in dynamic clustering of battery cells, especially for handling large amounts of battery cell data. In addition, the feature selection operation can be implemented through various methods. In some scenarios, using a feature selection method based on correlation analysis (such as Pearson correlation coefficient or information gain) can effectively identify and filter features closely related to the battery health state, while ignoring irrelevant or redundant features. In other scenarios, dimensionality reduction methods such as PCA can be used to reduce the dimensionality of the feature space, thereby improving the efficiency of subsequent clustering operations.

[0116] In practical applications, the selection of clustering algorithms and the implementation of feature selection can be adjusted according to the different characteristics of the battery cells. For example, in some high-end battery monitoring systems, more complex clustering algorithms (such as hierarchical clustering or DBSCAN) can be used to adapt to more diverse battery cell characteristics. Feature selection operations can also be fine-tuned according to specific device performance to ensure more accurate clustering processes.

[0117] This embodiment can effectively group battery cells according to similar characteristics through dynamic clustering based on static attribute data, thereby providing a solid foundation for subsequent monitoring, maintenance, and optimization. The battery cells in each battery cell group have similar performance characteristics, allowing operations and maintenance personnel to take more accurate management measures based on group characteristics. This clustering operation can significantly improve battery management efficiency, reduce misjudgments and lag reactions caused by the complexity and complexity of data processing in traditional methods, and thus improve the response speed and accuracy of battery monitoring and maintenance.

[0118] In one embodiment, the above step S30 includes:

[0119] S301, based on the internal resistance data in the operating state parameters, performing a reference determination operation on the homogenized battery cell group to generate dynamic reference curve data;

[0120] S302, performing a standard score determination operation on the internal resistance data of each battery cell to generate an internal resistance standard score sequence;

[0121] S303, performing a continuous deviation detection operation on the internal resistance standard score sequence to generate a marked continuous deviation battery cell;

[0122] S304, performing an anomaly identification operation on the marked continuous deviation battery cell to generate a preliminary set of abnormal battery cells.

[0123] S305, updating the dynamic reference curve data based on the internal resistance data in the newly collected operating status parameters.

[0124] In this embodiment, dynamic reference analysis based on internal resistance data is performed on the homogenized battery cell group, which can achieve early abnormal identification and ensure effective monitoring of the health status of the battery group. The change of internal resistance is an important indicator of battery degradation. When the internal resistance of the battery cell exceeds the normal range, it often means that the performance of the battery begins to degrade, and even there is a risk of failure. Therefore, by performing dynamic reference analysis based on internal resistance data, potential battery failure problems can be identified in a timely manner.

[0125] First, based on the internal resistance data in the operating status parameters, the reference determination operation is performed on the homogenized battery cell group. The purpose of this operation is to generate dynamic reference curve data within the battery cell group, which reflects the trend of internal resistance change of the battery cell under normal working conditions. In the process of generating dynamic reference curve, the internal resistance data of the battery is integrated and processed according to its historical performance and current state, thereby providing a reasonable reference for subsequent abnormal identification. In the process of reference determination, methods such as regression analysis or smoothing techniques can be used to ensure that the reference curve can accurately reflect the normal working range of the battery cell.

[0126] Next, the standard score determination operation is performed on the internal resistance data of each battery cell. The purpose of this operation is to convert the internal resistance data of each battery cell into standard scores (Z-scores), so as to facilitate the comparison of internal resistance performance of different battery cells. Standard score is obtained by comparing internal resistance data with the average internal resistance in the battery group, and measuring the deviation degree according to the standard deviation. This step helps to eliminate the differences between battery cells caused by data fluctuations, so that the fluctuation degree of internal resistance data is more significantly presented. The generation of standard score sequence plays a key role in subsequent abnormal identification.

[0127] After generating the standard score sequence, the continuous deviation detection operation is performed on the sequence. The core purpose of the continuous deviation detection operation is to mark the battery cells whose internal resistance standard scores continuously exceed the preset threshold for a period of time. Through this operation, battery cells that may have early degradation can be identified, because significant deviation of internal resistance is often a signal of battery degradation or failure. The detection of continuous deviation is based on the time sequence of internal resistance data, ensuring that the deviation is persistent and stable, rather than short-term fluctuation.

[0128] Subsequently, an abnormality identification operation is performed on the battery cells marked as persistent deviation. The purpose of this step is to further confirm whether the battery cells belong to a failure or abnormal state. The abnormality identification operation can use algorithms to evaluate the deviation situation in combination with historical data and performance indicators of the battery cells, to determine whether it constitutes a real anomaly. Common methods include judgment based on statistical models, machine learning models or rule engines. Through this operation, a preliminary set of abnormal battery cells can be generated, providing clear basis for battery management and subsequent maintenance.

[0129] Finally, based on the internal resistance data in the newly collected operating state parameters, the dynamic reference curve data is updated. As the battery cells are used, the internal resistance value may change, so it is necessary to update the reference curve regularly to ensure the timeliness and accuracy of the reference data. The updated reference curve will compare the new internal resistance data to help timely discover new abnormal trends. The update process can be done regularly or when the state of the battery cells changes significantly.

[0130] This embodiment can effectively improve the accuracy and efficiency of battery cell management by performing dynamic reference analysis and abnormality identification based on internal resistance data. By monitoring the changes in internal resistance in real time and comparing them with the dynamic reference, potential faulty or degraded battery cells can be identified early, preventing them from affecting the overall performance of the battery pack. This greatly reduces the lag in traditional manual monitoring and accurately identifies battery cells that need maintenance, reducing unnecessary maintenance costs. In addition, by regularly updating the reference curve, the battery state is continuously tracked, further improving the reliability and stability of the system.

[0131] In one embodiment, the above step S304 includes:

[0132] S3041, for each persistent deviation battery cell, obtain its internal resistance standard score sequence;

[0133] S3042, perform a continuous deviation segment identification operation on the internal resistance standard score sequence to generate a continuous deviation segment set;

[0134] S3043, perform a duration determination operation on each continuous deviation segment to generate a segment duration value;

[0135] S3044, perform an abnormality persistence determination operation based on the segment duration value to generate a persistent abnormality state judgment result and record the generation timestamp of the persistent abnormality state judgment result;

[0136] S3045, perform an abnormality confirmation operation on the persistent abnormality state judgment result to generate a confirmed abnormal battery cell;

[0137] S3046, aggregate all confirmed abnormal battery cells, generate a preliminary abnormal battery cell set, and record the generation timestamp of the preliminary abnormal battery cell set.

[0138] In this embodiment, by marking the continuously deviating battery cells, potential abnormal battery cells can be effectively identified, preventing the overall impact of battery degradation on the system. Changes in internal resistance often reflect the health of the battery, especially during long-term use. Therefore, by performing an abnormal identification operation on the continuously deviating battery cells, those degraded batteries can be discovered in time, ensuring the health of the battery pack.

[0139] First, for each battery cell marked as continuously deviating, its internal resistance standard score sequence is obtained. The purpose of this operation is to obtain the standardized sequence of battery cell internal resistance data, facilitating comparison with the normal range. The internal resistance standard score sequence represents the degree of change in battery internal resistance, and the higher the standard score, the greater the degree of deviation of the battery's internal resistance from the normal value, thus enabling more accurate identification of which battery cells exhibit abnormal internal resistance trends.

[0140] Next, a continuous deviation segment identification operation is performed on the internal resistance standard score sequence. This operation aims to identify segments in the internal resistance standard score sequence that continuously exceed the normal range (i.e., threshold). By identifying these segments, the duration of battery abnormalities can be further analyzed. For example, when the standard score of internal resistance continuously exceeds the threshold, it indicates that the health of the battery may be deteriorating and requires special attention. This step helps filter out occasional, short-lived fluctuations, ensuring that only those with long-term trends are focused on.

[0141] After generating the continuous deviation segment set, a duration determination operation needs to be performed on each continuous deviation segment. The purpose of this operation is to calculate the duration value of each segment, which is usually determined by the difference between the start time and end time of the segment. By quantifying the duration of each deviation segment, the time span of each abnormal trend can be more clearly understood, and further judgment can be made as to whether it meets the criteria for long-term abnormalities. The duration value is an important basis for determining whether an abnormality is persistent, and a longer duration of continuous deviation often means that the battery has a greater risk of failure.

[0142] Based on the segment duration value, an abnormal persistence determination operation is performed. The purpose of this operation is to determine whether these deviation segments meet the criteria for long-term abnormalities. Typically, the duration value is compared with a pre-set threshold, and if the duration exceeds the threshold, it is determined to be an abnormality, otherwise it is a transient fluctuation. By this method, the battery cells that truly have abnormalities can be more accurately identified, avoiding the misjudgment of temporary fluctuations as failures.

[0143] After completing the persistent abnormality determination, an abnormality confirmation operation needs to be performed on the persistent abnormality state determination result. The purpose of the abnormality confirmation operation is to further confirm which battery units are indeed abnormal, rather than errors caused by external factors. For example, some battery units may exhibit greater fluctuations under certain external conditions, but after abnormality confirmation, it is found that the fluctuations are accidental, rather than battery degradation. This operation verifies through the combination of historical data, diagnostic algorithms or other rule engines to ensure that the finally confirmed abnormal battery units have high accuracy.

[0144] Finally, all confirmed abnormal battery units will be aggregated to generate a preliminary abnormal battery unit set and record the generation timestamp. By aggregating all confirmed abnormal battery units, a complete abnormal list can be formed for subsequent processing and analysis. At the same time, recording the timestamp is very important for subsequent trend analysis and historical data backtracking analysis, which can help operation and maintenance personnel trace the generation time of abnormal battery units and analyze their change trends.

[0145] The embodiment can accurately identify abnormal conditions in battery units, especially those that have been in an abnormal state for a long time, by performing an abnormality identification operation on the marked persistent deviation battery units. This avoids the shortcomings of traditional technology that relies solely on threshold alarms, filters out short-term fluctuations, and ensures that only true abnormalities are processed. At the same time, through the recording of the timestamp and the determination of the duration, the reliability and accuracy of the abnormality judgment are improved, providing strong support for battery health monitoring and subsequent maintenance. This method not only improves the accuracy of battery unit abnormality detection, but also effectively reduces false positives, ensuring accurate implementation of maintenance plans and improving data center operation efficiency and battery system stability.

[0146] In one embodiment, the above step S40 comprises:

[0147] S401, for each battery unit in the preliminary abnormal battery unit set, extracting the internal resistance change feature and voltage fluctuation feature in the operating state parameters;

[0148] S402, based on the internal resistance data and temperature data in the operating state parameters, performing a temperature correlation analysis operation to generate a temperature correlation result;

[0149] S403, using a time series analysis model, based on the internal resistance change feature, the voltage fluctuation feature and the temperature correlation result, performing an abnormality classification operation to generate an abnormality classification result;

[0150] S404, performing a false positive filtering operation on the abnormality classification result to remove transient interference abnormality classification results to generate a confirmed abnormal battery unit list.

[0151] In this embodiment, accurately identifying abnormal battery cells is crucial for ensuring the stability of the battery pack. By combining multiple features in the operational state parameters, especially the internal resistance change, voltage fluctuation, and temperature correlation, more accurate abnormal classification can be performed. The core of this process is to process these features through a time series analysis model to identify the real abnormal battery cells and avoid misjudgment caused by short-term fluctuations or external factors.

[0152] First, for each battery cell in the preliminary abnormal battery cell set, the internal resistance change feature and the voltage fluctuation feature in the operational state parameters are extracted. The change in internal resistance can accurately reflect the health status of the battery, and an increase in internal resistance usually means the degradation or deterioration of battery performance. The voltage fluctuation feature can provide information about the performance changes of the battery under different loads. By extracting these features, necessary data support can be provided for subsequent classification, ensuring that battery performance features involved in the classification process are fully considered.

[0153] Next, based on the internal resistance data and temperature data in the operational state parameters, a temperature correlation analysis operation is performed to generate temperature correlation results. Temperature changes usually have a significant impact on battery performance, especially the relationship between internal resistance and temperature, which can reflect the working state of the battery under different temperature environments. By performing correlation analysis, the relationship between internal resistance and temperature can be quantified, providing more dimensional support for abnormal classification, thereby improving the accuracy of abnormal detection.

[0154] The use of a time series analysis model is the core of the abnormal classification operation. In this step, based on the extracted internal resistance change feature, voltage fluctuation feature, and temperature correlation result, the time series analysis model conducts in-depth analysis on the behavior of the battery cell. Time series analysis models such as LSTM, GRU, and other deep learning models can handle the correlation and dependency of time series data. Through analysis of historical data, the model can identify potential abnormal patterns and classify battery cells. The training process of the model ensures the accuracy and reliability of battery cell classification, which can distinguish between real abnormal battery cells and normal battery cells.

[0155] Perform false alarm filtering on the abnormal classification results. The purpose of false alarm filtering is to remove false classification results caused by transient disturbances or short-term fluctuations. Fluctuations in the battery system may be caused by external environmental changes (such as temperature fluctuations) or other non-battery factors, and these errors need to be filtered out during abnormal identification. Through false alarm filtering, the system can more accurately identify long-term stable abnormal states, reducing the need for human intervention and improving the level of automation of detection.

[0156] The implementation of false alarm filtering operation can be achieved through various means, aiming to effectively distinguish false alarms caused by short-term fluctuations or external interference factors from real abnormal states. First of all, the system needs to set a threshold or standard to identify the characteristics of short-term fluctuations. Generally, these short-term fluctuations are short in duration, small in amplitude, and usually quickly recover to the normal range within a certain period of time. These fluctuations may be caused by external environmental factors such as sudden changes in temperature, humidity changes, transient fluctuations during battery charging, etc. In order to achieve accurate false alarm filtering, the system compares the abnormal classification results with the historical data of the battery, especially with the past fluctuation patterns. By analyzing the historical data of the battery cells, the system can determine which abnormalities are caused by external environment or non-battery factors. The system can use sliding window technology to analyze the fluctuations within a certain period of time, and if the fluctuation duration is shorter than the pre-set threshold and the fluctuation recovery time is fast, it can be judged as false alarm, and then filter out these false alarm data. In addition, combined with multi-factor analysis, it is also helpful to reduce the generation of false alarms. For example, if multiple operating state parameters such as internal resistance, voltage and temperature fluctuate at the same time, and these fluctuations are not associated with the health status of the battery itself (such as aging, wear and tear), the system can classify them as transient interference, rather than real abnormalities. These data will be marked as "non-anomalies", ensuring that only long-term stable abnormal states are identified, avoiding false alarms. In summary, the key to false alarm filtering operation is to remove false classification caused by short-term fluctuations or transient interference through comparison and analysis of historical data, fluctuation duration, and external environmental factors, ensuring that the system can accurately identify long-term stable and real abnormal battery cells, thereby improving the accuracy and reliability of the entire detection process.

[0157] Finally, a list of confirmed abnormal battery cells is generated. This list contains all the battery cells that have been confirmed as abnormal, which need to be followed by maintenance, inspection or replacement operations. Through the generation of this list, the system can provide clear operation guidelines for battery management personnel, helping them to focus resources on the battery cells that need the most attention, thereby optimizing maintenance work and extending the service life of the battery pack.

[0158] This embodiment can more accurately identify abnormal battery cells by using a time series analysis model to comprehensively analyze internal resistance changes, voltage fluctuations and temperature correlations, especially in the early degradation stage. When internal resistance data, temperature fluctuations and voltage changes all point to changes in battery health, the system can efficiently screen out battery cells that need attention. In addition, the false alarm filtering mechanism can reduce false positives caused by short-term fluctuations, improving the accuracy of abnormal identification, thereby effectively extending the service life of the battery and reducing maintenance costs.

[0159] In one embodiment, the above step S50 comprises:

[0160] S501, for each battery unit in the list of confirmed abnormal battery units, associate its static attribute data, and generate an abnormal battery detailed list;

[0161] S502, based on the preset maintenance knowledge base, perform maintenance strategy matching operation on the abnormal battery detailed list, and generate maintenance strategy matching result;

[0162] S503, according to the preset maintenance schedule template, perform plan filling operation on the maintenance strategy matching result, and generate battery unit maintenance schedule.

[0163] In this embodiment, generating the maintenance schedule is a key step to ensure that the battery unit is maintained in a timely and effective manner. This process relies on the list of confirmed abnormal battery units and the existing maintenance knowledge base to provide personalized maintenance strategies and ultimately form a clear maintenance schedule.

[0164] First, each battery unit in the list of confirmed abnormal battery units is associated with the static attribute data of the battery unit. Static attribute data refers to basic information recorded at the time of battery factory shipment or initial deployment, which typically includes brand information to distinguish products from different manufacturers, product model information to distinguish different technical specifications, product capacity information to reflect the rated capacity parameters of the battery, and service duration information to describe the cumulative working period or time since the battery was put into use. The above information is derived from data records formed by the system during battery deployment or routine inspection. The association operation refers to data retrieval and binding of abnormal objects in the list of confirmed abnormal battery units with static attribute data in the database based on the unique identification of the battery unit, forming an abnormal battery detailed list. This list not only contains the basic attribute information of each abnormal battery, but also includes extended information such as the abnormal type, abnormal level, detection time, and historical maintenance records of the battery, facilitating the use of reference in the subsequent maintenance decision-making process.

[0165] After obtaining the detailed list of abnormal batteries, a maintenance strategy matching operation is performed based on a preset maintenance knowledge base. The maintenance knowledge base refers to a set of standardized rules pre-established in the system, covering standard maintenance measures and strategy information corresponding to different brands, models, capacities, battery types, and abnormal situations. The content of this knowledge base is usually derived from a multi-dimensional information set including technical manuals provided by equipment manufacturers, industry standards, historical maintenance data, expert experience rules, etc. The maintenance strategy matching operation is to compare the information in the detailed list of abnormal batteries with the maintenance rules in the maintenance knowledge base through logical judgment and rule matching, to determine the specific maintenance measures. For example, if a certain model of battery shows abnormal internal resistance in the detection and the service length has approached the upper limit, the system can automatically match the strategy of recommending replacement; if the battery has slight voltage fluctuations and other parameters are within the controllable range, the system matches the suggestion of regular observation or adjustment of the charging and discharging strategy. The matching result forms a maintenance strategy matching result data structure, including the specific maintenance measures, priority arrangement, and operation suggestions for each abnormal battery unit.

[0166] After completing the strategy matching, the system performs a plan filling operation according to a preset maintenance schedule template to generate a battery unit maintenance schedule. The maintenance schedule template refers to a structured data table or electronic document framework that defines various necessary information fields for maintenance tasks, such as battery unit number, abnormal type, maintenance strategy, operation steps, execution time node, responsible personnel information, execution status, etc. During the plan filling operation, the system fills in the corresponding maintenance measures for each abnormal battery unit into the corresponding fields of the maintenance schedule according to the maintenance strategy matching result, ensuring complete information, clear logic, and unified structure. The final battery unit maintenance schedule has clear task arrangement, operation specification, and execution standard, facilitating the operation and maintenance team to efficiently perform maintenance tasks, reducing the risk of omission, and improving the overall management level.

[0167] Example: In the field of battery management, for the operation and maintenance needs of data center UPS backup lead-acid batteries, for a certain data center room, a backup battery pack containing 12000 lead-acid batteries is configured. The system continuously collects the operating state parameters of all battery units, including the internal resistance data, voltage data, and temperature data of each battery unit, as well as the static attribute data of each battery unit, including brand, model, rated capacity, and service length. The operating state parameters are transmitted in real time through the dynamic environment monitoring system of the data center, and the static attribute data is automatically read through the equipment archives of the operation and maintenance system.

[0168] After the data collection is completed, the system performs data cleaning operation on the running state parameters, eliminates abnormal values and invalid points in the data transmission process, and generates continuous and effective cleaned running state parameters. For the cleaned data, the system unifies the time alignment processing according to the monitoring time axis of the battery operation, and ensures that the monitoring data of all battery units are synchronized for analysis with the same time reference. The data after time alignment is further calculated for the maximum value, mean value and minimum value of the internal resistance data, the maximum value, mean value and minimum value of the voltage data, and the maximum value, mean value and minimum value of the temperature data of each battery unit according to the statistical period of each hour, to form the running state aggregated data, and the aggregated data is stored in the battery state management database together with the static attribute data.

[0169] For the stored data, the system encodes the static attribute data according to the category attribute, converts the brand and model into numerical category index, and at the same time standardizes the capacity and service time data, to ensure that the feature parameters are processed in the same numerical scale. After standardization, the static attribute data is selected by the system to retain the brand, model, capacity and service time which are highly related to the battery health state as clustering feature data. Based on the clustering feature data, the system uses distance-optimized grouping method to dynamically cluster all battery units, and automatically divides the battery units with high data similarity and consistent battery attributes into initial battery unit groups. The attribute consistency verification is performed on the initial battery unit groups to ensure that each group of batteries maintains high consistency in brand, model, capacity and other key attributes. After verification, the homogenized battery unit groups are generated, and the division results are recorded in the battery grouping management table.

[0170] In each homogenized battery unit group, the system selects the internal resistance data of the corresponding battery, performs reference determination operation, periodically calculates the internal resistance mean value and standard deviation based on historical data, and generates dynamic reference curve data as the comparison reference of internal resistance change within the group. For the internal resistance data of each battery unit, the system further calculates the internal resistance standard score sequence to measure the deviation degree of the battery internal resistance at each time point relative to the reference curve. Through continuous deviation detection, the system automatically identifies the battery units whose internal resistance standard score continuously exceeds the set threshold in multiple consecutive time periods, and marks them as continuously deviated battery units.

[0171] For the marked continuous deviation battery unit, the system acquires its complete internal resistance standard score sequence, performs a continuous deviation segment identification operation, locates all continuous deviation time periods, and counts the duration of each continuous deviation segment. The system compares the duration of each segment with the preset abnormal duration judgment condition, generates a continuous abnormal state judgment result, and records the timestamp of the judgment result in real time. The system inputs the continuous abnormal state judgment result into the abnormal confirmation operation, confirms whether it belongs to a long-term internal resistance abnormality by cross- verifying historical data and the current state. All confirmed abnormal battery units are automatically summarized by the system to form a preliminary abnormal battery unit set, and the set generation time point is recorded.

[0172] For each battery unit in the preliminary abnormal battery unit set, the system extracts its operating state parameters, calculates the internal resistance change rate and voltage fluctuation amplitude as the time series features of the battery. In combination with the internal resistance data and temperature data of the battery unit, the system performs temperature correlation analysis, generates a temperature correlation result, and reflects the coupling change trend of the battery internal resistance and temperature. The above internal resistance change features, voltage fluctuation features, and temperature correlation results are input into the trained time series analysis model to perform abnormal classification operation. The system classifies the abnormal batteries and determines whether they belong to real degraded batteries or false positive interference batteries. For the abnormal classification result, the system further applies a false positive filtering mechanism to dynamically analyze parameters such as abnormal duration, fluctuation law, and correlation with environmental changes, automatically excludes false positive abnormalities caused by short-term external interference or temperature transients, and finally generates a confirmed abnormal battery unit list.

[0173] Based on the confirmed abnormal battery unit list, the system associates the static attribute data of each battery unit to generate an abnormal battery detailed list, which records detailed information such as the brand, model, capacity, service duration, abnormal type, and abnormal occurrence time of each abnormal battery. The system calls the maintenance knowledge base and matches the parameters in the abnormal battery detailed list with the maintenance rules in the maintenance knowledge base to generate a maintenance strategy matching result, determines the processing priority and maintenance recommendations for different battery units, such as deep discharge, parameter adjustment, or battery replacement, etc. Finally, the system fills the confirmed abnormal battery unit list and the matched maintenance strategy into the standardized maintenance schedule according to the maintenance schedule template, generates a battery unit maintenance schedule containing battery number, abnormal type, maintenance measures, operation steps, execution time node, and responsible person, etc., and provides it to the operation and maintenance team as a direct reference for subsequent on-site maintenance. The entire process realizes the full-process closed-loop automated management of data center lead-acid batteries from data acquisition, clustering analysis, abnormal detection, classification confirmation, maintenance decision-making, to schedule generation.

[0174] In the field of medical health, for the management of backup lead-acid batteries widely used in hospital backup power supply systems, in the configuration of large hospital central power supply systems, there are multiple UPS backup battery groups, which directly affect the continuous power supply capability of operating rooms, emergency equipment and intensive care equipment during main power failure. The hospital power management system continuously collects the operating state parameters of all backup battery units, including internal resistance data, voltage data and temperature data, and obtains the static attribute information of each battery unit, including brand, model, capacity and service length, etc. The operating state parameters are transmitted in real time through the hospital equipment monitoring platform, and the static attribute information is imported synchronously through the power asset management system.

[0175] During data collection, the hospital power management system first removes outliers from the operating state parameters, removes invalid data caused by monitoring abnormalities or device transmission interruptions, and generates continuous valid data sequences. The cleaned data is uniformly aligned with the time reference to ensure that the monitoring data of all battery units is recorded synchronously at consistent time nodes. The data after time alignment is counted every hour, and the maximum, average and minimum values of the internal resistance, voltage and temperature of each battery unit in each monitoring period are calculated to form detailed operating state aggregated data and store it simultaneously with the battery static attribute information.

[0176] In the hospital battery data processing link, the system performs category coding and numerical standardization processing on the static attribute information, and the brand and model are converted into coded data, and the capacity and service length are normalized. The system selects clustering features from the standardized data and extracts attributes that have a greater impact on battery operating status for subsequent grouping. Based on the feature data, the system applies an optimized distance algorithm to divide battery units with similar operating status and consistent attributes into initial battery unit groups. Then the system checks whether the initial grouping maintains high consistency in key attributes such as brand, model and capacity, and generates homogenized battery unit groups after verification, and records the grouping attribution of all batteries.

[0177] For each homogenized battery unit group, the hospital power management system analyzes the historical changes of the internal resistance data, generates a dynamic reference curve based on the monitoring period, and forms a real-time reference standard reflecting the battery health baseline. The current internal resistance data of each battery unit is compared with the dynamic reference curve, and the system automatically calculates the internal resistance standard score sequence to identify whether the internal resistance continuously deviates from the reference range. Through analysis of the internal resistance standard score sequence, the system continuously detects the continuous deviation state and automatically marks the battery units with long-term deviation of internal resistance.

[0178] For the marked persistent deviation battery unit, the system further extracts its internal resistance standard score sequence, identifies the time segment of internal resistance persistent deviation in segments, and counts the duration of each deviation segment. The system determines whether the deviation time exceeds the threshold according to the persistent anomaly standard set by the hospital backup battery management requirement, forms the persistent anomaly state judgment result, and records the time node of anomaly confirmation. The system confirms the anomaly of all persistent anomaly battery units, eliminates short-term fluctuations or external factors, and obtains a preliminary anomaly battery unit set and records the generation time.

[0179] The hospital power management system extracts the internal resistance change rate and voltage fluctuation characteristics of each battery for the preliminary anomaly battery unit set, identifies the correlation between battery internal resistance and environmental temperature fluctuation through temperature correlation analysis, and inputs the above characteristics into the hospital customized time series analysis model to classify and identify whether it is a real degraded battery or a short-term abnormal disturbance. The false alarm filtering mechanism automatically excludes false alarms caused by environmental temperature fluctuations, equipment load changes and other non-battery factors, and forms a confirmed anomaly battery unit list.

[0180] Based on the confirmed anomaly battery unit list, the hospital power management system automatically associates the static attribute information of each abnormal battery to generate an abnormal battery detailed list, including battery model, capacity, anomaly type, and anomaly time. The system matches the anomaly battery detailed list with the hospital battery maintenance knowledge base to determine the priority and maintenance plan for different battery types, such as on-site deep discharge test, battery replacement, and battery pack re-adjustment. The system fills the confirmed anomaly battery information and maintenance strategy into the hospital battery maintenance schedule to clearly define the maintenance tasks, time arrangement, responsible person and operation steps for each abnormal battery. This schedule is used to guide the hospital logistics power team to efficiently and orderly perform backup battery maintenance tasks, ensuring the continuous power supply capability of key equipment in the hospital when the main power fails. The entire process realizes the automation and intelligentization of hospital backup battery management, reduces the burden of manual analysis, and improves the response speed of battery anomaly detection and the rationality of maintenance decision.

[0181] In the field of financial technology, for the uninterruptible power supply system supporting data centers, core machine rooms or financial transaction platforms, the above battery anomaly detection and maintenance method can effectively ensure the continuity and data security of financial business systems, and avoid high-risk events such as transaction interruption and system downtime caused by battery degradation or failure.

[0182] In the data center of large financial institutions, multiple groups of lead-acid batteries are deployed for UPS backup systems, and these battery units directly determine the emergency power supply capacity of financial business systems in the event of a power supply failure. The management system obtains the operating state parameters of all battery units in real time through the dynamic environment monitoring network of the financial data center, including internal resistance data, voltage data, and temperature data. At the same time, the system synchronously manages the static attribute data of each battery unit, covering brand, model, capacity, and service length, etc. These data are regularly updated and archived by the asset management system of the financial computer room.

[0183] In the data acquisition phase, the system first performs data cleaning on all operating state parameters, eliminating noise data, abnormal sampling values, and incomplete records to ensure the accuracy of subsequent analysis. The cleaned data is uniformly aligned with the time reference to form structured time series, ensuring strict time synchronization of the monitoring data of each battery unit. Based on the aligned data, the system calculates the maximum, average, and minimum values of the internal resistance, voltage, and temperature of each battery unit in each hour period, generates stable and reliable operating state aggregation data, and stores it together with the static attribute data to the data management platform of the financial institution.

[0184] The system performs category attribute encoding and numerical attribute standardization operations on battery static attribute data, converting brands and models into a unified encoding format, and standardizing capacity and service length parameters into highly comparable numerical expressions. Through feature selection, the system extracts key influencing factors to participate in the subsequent dynamic clustering process. The battery units of the financial data center are divided into initial battery unit groups based on the optimized distance algorithm, and the system checks the consistency of brands, models, and capacities within each group to ensure that the batteries within the group are highly homogeneous, ultimately forming homogeneous battery unit groups and recording the grouping attribution information of each battery unit.

[0185] For homogeneous battery unit groups, the system performs dynamic benchmarking analysis based on internal resistance data in the operating state parameters. By analyzing historical data, the system constructs a dynamic benchmark curve for each group, reflecting the normal internal resistance fluctuation range. The system automatically calculates the internal resistance standard score sequence, compares the current internal resistance level with the dynamic benchmark curve, and identifies battery units that deviate from the normal range in real time, and marks the continuous deviation state.

[0186] For continuously deviating battery units, the system further obtains the internal resistance standard score sequence, identifies continuous deviation segments, calculates the duration of each segment, and determines whether the internal resistance deviation is long-term and stable based on the continuous anomaly determination logic. Based on the abnormal state judgment result and the timestamp information, the system identifies abnormal battery units and forms a preliminary abnormal battery unit set.

[0187] The system extracts the internal resistance change feature and the voltage fluctuation feature for each battery in the preliminary abnormal battery unit set, comprehensively evaluates the abnormal state of the battery by combining the temperature correlation analysis, and determines the authenticity of the abnormality by using the multivariate feature through the time series analysis model to avoid short-term fluctuations or environmental factors interference. The false alarm filtering mechanism further eliminates the misjudgment caused by temperature changes, load adjustment and other external reasons to form a high-accuracy confirmed abnormal battery unit list.

[0188] The system generates a detailed list containing brand, model, capacity, abnormal type and time information based on the confirmed abnormal battery unit list and the association of the static attribute data of each battery. Based on the maintenance knowledge base of the financial data center, the system performs strategy matching on the detailed list of abnormal batteries to determine the priority, maintenance method and operation steps, and forms a customized maintenance strategy. Finally, the system converts the strategy matching result into a structured battery unit maintenance schedule table by combining the standardized maintenance schedule template, and clearly defines the maintenance time, operator and technical requirements of each battery.

[0189] This process realizes intelligent abnormal detection and maintenance strategy formulation of UPS backup batteries in the field of financial technology, ensures the stable operation of key facilities such as financial transaction systems, core data storage devices and business continuity platforms in the case of main power failure, reduces the risk of business interruption, data damage and economic loss caused by battery performance degradation, and improves the overall operation and management efficiency and system stability of financial institutions.

[0190] This embodiment generates a detailed battery list by combining the confirmed abnormal battery unit list with static attribute data, providing an accurate basis for subsequent maintenance work. The strategy matching operation based on the pre-set maintenance knowledge base enables each abnormal battery unit to receive a customized maintenance strategy, improving the relevance and efficiency of maintenance. By using a standardized template to fill in the maintenance schedule table, the uniformity and operability of the maintenance work are ensured, reducing the need for manual intervention and improving the automation level of battery management, ultimately achieving efficient and accurate maintenance of batteries.

[0191] In one embodiment, a battery unit abnormality detection device is provided, which corresponds to the battery unit abnormality detection method described in the above embodiments. Referring to Figure 3 , Figure 3 The functional module schematic diagram of a preferred embodiment of the battery unit abnormality detection device of the present application is shown. The data acquisition module 10, the clustering analysis module 20, the reference analysis module 30, the time series analysis module 40 and the maintenance schedule generation module 50. The detailed description of each functional module is as follows:

[0192] The data acquisition module 10 is used to acquire the static attribute data and the running state parameters of a plurality of battery units;

[0193] a clustering analysis module 20, configured to perform dynamic clustering on the plurality of battery cells according to the static attribute data, to generate a homogenized battery cell group;

[0194] a benchmark analysis module 30, configured to perform dynamic benchmark analysis on the homogenized battery cell group based on internal resistance data in the operating state parameters, to preliminarily identify abnormal battery cells, and to form a preliminary abnormal battery cell set;

[0195] a time series analysis module 40, configured to perform multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set using a time series analysis model, to classify the preliminarily identified abnormal battery cells, and to generate a confirmed abnormal battery cell list;

[0196] a maintenance plan generation module 50, configured to generate a battery cell maintenance plan table based on the confirmed abnormal battery cell list and a preset maintenance knowledge base.

[0197] In an embodiment, the data collection module 10 is specifically configured to:

[0198] collect operating state parameters of each battery cell, including internal resistance, voltage and temperature;

[0199] collect static attribute data of each battery cell, including brand, model, capacity and service time length;

[0200] perform a data cleaning operation on the operating state parameters, to generate cleaned operating state parameters;

[0201] perform a time alignment operation on the cleaned operating state parameters, to generate time-aligned operating data;

[0202] perform a statistical aggregation operation on the time-aligned operating data, to generate operating state aggregated data;

[0203] store the operating state aggregated data and the static attribute data.

[0204] In an embodiment, the clustering analysis module 20 is specifically configured to:

[0205] perform category attribute encoding and numerical attribute standardization processing on the static attribute data, to generate standardized static attribute data;

[0206] perform feature selection operation on the standardized static attribute data, to determine clustering feature data;

[0207] perform distance optimization grouping operation on the clustering feature data, to generate an initial battery cell group;

[0208] perform attribute consistency verification operation on the initial battery cell group, to generate a homogenized battery cell group;

[0209] record the division result of the homogeneous battery cell group.

[0210] In an embodiment, the reference analysis module 30 is specifically configured to:

[0211] perform a reference determination operation on the homogeneous battery cell group based on the internal resistance data in the operating state parameters, to generate dynamic reference curve data;

[0212] perform a standard score determination operation on the internal resistance data of each battery cell, to generate an internal resistance standard score sequence;

[0213] perform a continuous deviation detection operation on the internal resistance standard score sequence, to generate a marked continuous deviation battery cell;

[0214] perform an anomaly identification operation on the marked continuous deviation battery cell, to generate a preliminary anomaly battery cell set;

[0215] update the dynamic reference curve data based on the internal resistance data in the newly collected operating state parameters.

[0216] In an embodiment, the reference analysis module 30 is specifically configured to:

[0217] for each marked continuous deviation battery cell, obtain its internal resistance standard score sequence;

[0218] perform a continuous deviation segment identification operation on the internal resistance standard score sequence, to generate a continuous deviation segment set;

[0219] perform a duration determination operation on each continuous deviation segment, to generate a segment duration value;

[0220] perform an anomaly persistence determination operation based on the segment duration value, to generate a continuous anomaly state judgment result, and record a generation time stamp of the continuous anomaly state judgment result;

[0221] perform an anomaly confirmation operation on the continuous anomaly state judgment result, to generate a confirmed anomaly battery cell;

[0222] aggregate all confirmed anomaly battery cells, to generate a preliminary anomaly battery cell set, and record a generation time stamp of the preliminary anomaly battery cell set.

[0223] In an embodiment, the time sequence analysis module 40 is specifically configured to:

[0224] for each battery cell in the preliminary anomaly battery cell set, extract internal resistance change characteristics and voltage fluctuation characteristics in the operating state parameters;

[0225] Based on the internal resistance and temperature data in the operating status parameters, perform a temperature correlation analysis to generate temperature correlation results.

[0226] Using a time-series analysis model, an anomaly classification operation is performed based on the internal resistance change characteristics, the voltage fluctuation characteristics, and the temperature correlation results to generate anomaly classification results.

[0227] Perform a false alarm filtering operation on the anomaly classification results, remove transient interference anomaly classification results, and generate a list of confirmed abnormal battery cells.

[0228] In one embodiment, the maintenance plan generation module 50 is specifically used for:

[0229] For each battery cell in the confirmed abnormal battery cell list, associate its static attribute data to generate a detailed list of abnormal batteries;

[0230] Based on a preset maintenance knowledge base, a maintenance strategy matching operation is performed on the detailed list of abnormal batteries to generate maintenance strategy matching results.

[0231] According to the preset maintenance plan template, the plan filling operation is performed on the maintenance strategy matching result to generate the battery cell maintenance plan.

[0232] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides determination and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external user terminals via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a battery cell anomaly detection method on the server side.

[0233] In one embodiment, a computer device is provided, which may be a user terminal, and its internal structure diagram may be as follows: Figure 5The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide determination and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external server through a network connection. The computer program is executed by the processor to implement the functions or steps of the user side of the battery cell anomaly detection method

[0234] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the following steps:

[0235] Obtain static attribute data and operating state parameters of a plurality of battery cells;

[0236] Dynamically cluster the plurality of battery cells according to the static attribute data to generate homogeneous battery cell groups;

[0237] Within the homogeneous battery cell groups, perform dynamic benchmarking based on internal resistance data in the operating state parameters to preliminarily identify abnormal battery cells and form a preliminary abnormal battery cell set;

[0238] For each battery cell in the preliminary abnormal battery cell set, perform multivariate time series analysis using a time series analysis model to classify the preliminarily identified abnormal battery cells and generate a confirmed abnormal battery cell list;

[0239] Based on the confirmed abnormal battery cell list and a preset maintenance knowledge base, generate a battery cell maintenance schedule.

[0240] In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, and the computer program is executed by a processor to implement the following steps:

[0241] Obtain static attribute data and operating state parameters of a plurality of battery cells;

[0242] Dynamically cluster the plurality of battery cells according to the static attribute data to generate homogeneous battery cell groups;

[0243] Within the homogeneous battery cell groups, perform dynamic benchmarking based on internal resistance data in the operating state parameters to preliminarily identify abnormal battery cells and form a preliminary abnormal battery cell set;

[0244] performing multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set using a time series analysis model to classify the preliminarily identified abnormal battery cells, to generate a list of confirmed abnormal battery cells;

[0245] generating a battery cell maintenance schedule based on the list of confirmed abnormal battery cells and a preset maintenance knowledge base.

[0246] It should be noted that the functions or steps described above with respect to the computer readable storage medium or the computer device can correspond to the related descriptions of the server side and the user side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0247] Those skilled in the art can understand that all or part of the processes in the foregoing method embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the foregoing method embodiments. In the embodiments provided in the present application, any reference to memory, storage, database or other medium can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0248] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified. In actual applications, the above functions can be completed by different functional units or modules according to needs, i.e. the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0249] It should be explained that if the software tools or components of other companies appear in the embodiments of the present application, they are only used for example introduction and do not represent actual use. The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A battery cell abnormality detection method characterized by comprising: The method comprises the following steps: Obtain static attribute data and running state parameters of a plurality of battery units; Perform dynamic clustering on the plurality of battery units according to the static attribute data to generate homogeneous battery unit groups; Perform dynamic benchmark analysis based on internal resistance data in the running state parameters within the homogeneous battery unit groups to preliminarily identify abnormal battery units and form a preliminary abnormal battery unit set; Perform multivariate time series analysis on each battery unit in the preliminary abnormal battery unit set using a time series analysis model to classify the preliminarily identified abnormal battery units and generate a confirmed abnormal battery unit list; Generate a battery unit maintenance schedule based on the confirmed abnormal battery unit list and a preset maintenance knowledge base.

2. The battery cell abnormality detection method according to claim 1, wherein Obtaining static attribute data and running state parameters of a plurality of battery units comprises: Collecting running state parameters including internal resistance, voltage and temperature of each battery unit; Obtaining static attribute data including brand, model, capacity and service time of each battery unit; Performing data cleaning operation on the running state parameters to generate cleaned running state parameters; Performing time alignment operation on the cleaned running state parameters to generate time-aligned running data; Performing statistical aggregation operation on the time-aligned running data to generate running state aggregated data; Storing the running state aggregated data and the static attribute data.

3. The battery cell abnormality detection method according to claim 1, wherein Performing dynamic clustering on the plurality of battery units according to the static attribute data to generate homogeneous battery unit groups comprises: Performing category attribute encoding and numerical attribute standardization processing on the static attribute data to generate standardized static attribute data; Performing feature selection operation on the standardized static attribute data to determine clustering feature data; Performing distance optimization grouping operation on the clustering feature data to generate initial battery unit groups; Performing attribute consistency verification operation on the initial battery unit groups to generate homogeneous battery unit groups; Recording the division results of the homogeneous battery unit groups.

4. The battery cell abnormality detection method according to claim 1, wherein Performing dynamic benchmark analysis based on internal resistance data in the running state parameters within the homogeneous battery unit groups to preliminarily identify abnormal battery units and form a preliminary abnormal battery unit set comprises: Performing benchmark determination operation on the homogeneous battery unit groups based on the internal resistance data in the running state parameters to generate dynamic benchmark curve data; Performing standard score determination operation on the internal resistance data of each battery unit to generate internal resistance standard score sequence; Performing continuous deviation detection operation on the internal resistance standard score sequence to generate marked continuous deviation battery units; Performing abnormal identification operation on the marked continuous deviation battery units to generate a preliminary abnormal battery unit set; Updating the dynamic benchmark curve data based on newly collected internal resistance data in the running state parameters.

5. The battery cell abnormality detection method according to claim 4, wherein Performing abnormal identification operation on the marked continuous deviation battery units to generate a preliminary abnormal battery unit set comprises: Obtaining the internal resistance standard score sequence of each marked continuous deviation battery unit; Performing continuous deviation segment identification operation on the internal resistance standard score sequence to generate a continuous deviation segment set; performing a duration determination operation on each continuous deviation segment to generate a segment duration value; performing an abnormality persistence determination operation based on the segment duration value to generate a persistent abnormality status determination result and record a generation timestamp of the persistent abnormality status determination result; performing an abnormality confirmation operation on the persistent abnormality status determination result to generate a confirmed abnormal battery cell; aggregating all the confirmed abnormal battery cells to generate a preliminary abnormal battery cell set and record a generation timestamp of the preliminary abnormal battery cell set.

6. The battery cell abnormality detection method according to claim 1, wherein performing multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set using a time series analysis model to classify the preliminarily identified abnormal battery cells, and generating a confirmed abnormal battery cell list, including: extracting an internal resistance change feature and a voltage fluctuation feature in the operating state parameters for each battery cell in the preliminary abnormal battery cell set; performing a temperature correlation analysis operation based on the internal resistance data and the temperature data in the operating state parameters to generate a temperature correlation result; performing an abnormality classification operation based on the internal resistance change feature, the voltage fluctuation feature, and the temperature correlation result using a time series analysis model to generate an abnormality classification result; performing a false alarm filtering operation on the abnormality classification result to remove transient interference abnormality classification results and generate a confirmed abnormal battery cell list.

7. The battery cell abnormality detection method according to claim 1, wherein generating a battery cell maintenance schedule based on the confirmed abnormal battery cell list and a preset maintenance knowledge base, including: associating the static attribute data of each battery cell in the confirmed abnormal battery cell list to generate an abnormal battery detailed list; performing a maintenance strategy matching operation on the abnormal battery detailed list based on the preset maintenance knowledge base to generate a maintenance strategy matching result; performing a plan filling operation on the maintenance strategy matching result according to a preset maintenance schedule template to generate a battery cell maintenance schedule.

8. A battery cell abnormality detection device characterized by comprising: The battery cell anomaly detection device includes: a data acquisition module configured to acquire static attribute data and operating state parameters of a plurality of battery cells; a clustering analysis module configured to perform dynamic clustering on the plurality of battery cells based on the static attribute data to generate a homogenized battery cell group; a reference analysis module configured to perform dynamic reference analysis based on internal resistance data in the operating state parameters within the homogenized battery cell group to preliminarily identify abnormal battery cells and form a preliminary abnormal battery cell set; a time series analysis module configured to perform multivariate time series analysis on each battery cell in the preliminary abnormal battery cell set using a time series analysis model to classify the preliminarily identified abnormal battery cells and generate a confirmed abnormal battery cell list; a maintenance schedule generation module configured to generate a battery cell maintenance schedule based on the confirmed abnormal battery cell list and a preset maintenance knowledge base.

9. A computer device, comprising: The computer device includes a memory, a processor, and a battery cell anomaly detection program stored on the memory and executable on the processor, and the battery cell anomaly detection program, when executed by the processor, implements the steps of the battery cell anomaly detection method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a battery cell abnormality detection program that, when executed by the processor, implements the steps of the battery cell abnormality detection method of any one of claims 1-7.

Images