Lithium battery intelligent management system and method and medium

Abstract

The invention relates to a lithium battery intelligent management system and method and a medium, and relates to the technical field of battery management. The lithium battery intelligent management system comprises an intelligent battery management terminal used for collecting and cleaning battery operation parameters and battery position data and transmitting the parameters and the data to a cloud computing platform; the cloud computing platform is used for receiving, integrating and intelligently analyzing the battery operation parameters and the battery position data, outputting a battery multi-dimensional scene state in combination with a multi-dimensional intelligent judgment model, and generating an abnormal scene solving strategy; the user terminal is used for receiving the abnormal scene solving strategy and the battery digital file and supporting work order reservation, progress query and self-service processing; the after-sales service terminal receives the maintenance work order and the battery alarm information and uploads the actual working condition, the actual detection report and the actual abnormality solving strategy of the battery; and comparing multi-path output results through a multi-dimensional intelligent judgment model, calculating a scene difference degree to verify consistency, and triggering an alarm when the difference exceeds a limit, so that high-precision intelligent judgment and fault early warning of the battery state are realized.

Description

Technical Field

[0001] This application relates to the field of battery management technology, and in particular to a lithium battery intelligent management system, method and medium. Background Technology

[0002] With the widespread application of lithium batteries in electric vehicles, energy storage systems, and other fields, the demand for lithium battery management and maintenance is increasing. Traditional lithium battery management systems (BMS) mainly focus on real-time monitoring and basic protection functions of the battery.

[0003] Currently, lithium battery after-sales management is mainly achieved through regular manual maintenance and offline testing, with technicians conducting on-site checks of the battery status.

[0004] Existing patents disclose an intelligent BMS management method for lithium batteries, including: collecting lithium battery charging and discharging parameters, constructing a database, and storing the collected lithium battery charging and discharging parameters; traversing the lithium battery charging and discharging parameters stored in the database, analyzing the changing trends of the lithium battery charging and discharging state, and generating a lithium battery charging and discharging state change trend map. This invention generates a lithium battery charging and discharging state change trend map through historical charging and discharging parameter analysis, thus providing visualized monitoring and management conditions for lithium battery charging and discharging state monitoring. Furthermore, based on further analysis of the lithium battery charging and discharging parameters, the invention performs multi-faceted monitoring of the lithium battery charging and discharging state, and based on the monitoring results, further provides recommendations for lithium battery charging disconnection operations, providing charging disconnection operation prompts to lithium battery users, making the overall charging process of the lithium battery healthier, thereby maintaining the overall lifespan of the lithium battery.

[0005] The existing technical solutions mentioned above have the following drawbacks: 1. Traditionally, battery data is exported and analyzed using professional software. Offline data analysis of lithium batteries is not timely and lacks comprehensive analysis of battery status and scenario-based decision support, resulting in low maintenance efficiency, high cost and poor user experience. It is impossible to achieve intelligent, automated and precise after-sales maintenance of lithium batteries. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this application is to provide a lithium battery intelligent management system, method, and medium. By implementing AI-powered intelligent after-sales maintenance in lithium battery after-sales management, the system intelligently analyzes scenarios such as battery damage, battery life, and battery theft. When a user triggers a corresponding scenario, the AI ​​recommends customer service and provides the corresponding solution, significantly improving the security, intelligence level, and operational efficiency of the battery management system.

[0007] This was achieved using the following technical solutions: In a first aspect, this application provides a lithium battery intelligent management system, comprising: The intelligent battery management terminal is used to collect and clean battery operating parameters and battery location data, and transmit them to the cloud computing platform; The cloud computing platform is used to receive, integrate, and intelligently analyze battery operating parameters and battery location data, and combine them with a multi-dimensional intelligent judgment model to output the multi-dimensional scenario status of the battery and generate abnormal scenario resolution strategies. The user terminal is used to receive abnormal scenario resolution strategies and battery digital profiles, and supports work order scheduling, progress inquiry and self-service processing. The after-sales service terminal is used to receive repair work orders and battery alarm information, and to upload the actual battery operating conditions, actual test reports, and actual fault resolution strategies.

[0008] Furthermore, the intelligent battery management terminal includes: The sensor module is used to monitor the battery's operation, collect battery operating parameters, locate and track the battery's real-time position, and obtain battery position data. The data processing module is used to filter and denoise, remove duplicate values, fill missing values, and normalize battery operating parameters and battery location data according to timestamps to obtain standard operating parameters and standard location data. The data encryption module is used to encrypt standard operating parameters and standard location data to obtain encrypted operating parameters and encrypted location data. The communication module is used to verify the communication protocol with the cloud computing platform and establish a data transmission channel to transmit encrypted operating parameters and encrypted location data.

[0009] Furthermore, cloud computing platforms include: The data receiving module is used to receive and decrypt encrypted operating parameters and encrypted location data according to the data transmission channel to obtain secure operating parameters and secure location data; The data integration module is used to integrate and store safe operating parameters, safe location data, battery production information and historical maintenance records based on the battery code, forming a digital battery profile; The intelligent analysis module is used to deconstruct and analyze the battery digital profile and identify its status based on a multi-dimensional intelligent judgment model, and output the battery's multi-dimensional scene status. The intelligent decision-making module is used to analyze the multi-dimensional scenario status of the battery, extract abnormal scenario information, and match and generate abnormal scenario resolution strategies. The feedback learning module is used to compare and optimize the multidimensional intelligent judgment model based on the actual battery operating conditions, actual test reports, and actual anomaly resolution strategies to obtain the optimal multidimensional judgment model.

[0010] By adopting the above technical solutions, real-time acquisition and cleaning of battery operating parameters are achieved through the integration of sensor data fusion and adaptive filtering algorithms, and encrypted transmission protocols are used to ensure secure data upload to the cloud. The cloud computing platform uses multi-dimensional intelligent judgment models (such as state recognition and decision tree strategy matching based on machine learning) to conduct in-depth analysis of the battery digital profile and output scenario status and solution strategies. At the same time, the model is continuously optimized through feedback learning algorithms, forming a closed-loop management system from terminal perception, cloud decision-making to after-sales feedback, which significantly improves the security, intelligence level and operation and maintenance efficiency of the battery management system.

[0011] Furthermore, the intelligent analysis module includes: The information deconstruction unit is used to deconstruct and aggregate the battery digital profile according to the data verification dimensions to construct health detection data, fault identification data, and scenario classification data; The model building unit is used to iteratively update the model based on multi-dimensional historical battery data and deep learning algorithms to generate a multi-dimensional intelligent judgment model. The comprehensive analysis unit is used to perform multidimensional analysis on health detection data, fault identification data, and scene classification data based on different recognition network branches of the multidimensional intelligent judgment model, and to determine the multidimensional scene status of the battery.

[0012] By adopting the above technical solution, multi-dimensional data aggregation of battery digital archives is performed through the information deconstruction unit, and iterative training and comprehensive analysis are carried out using a deep learning-based multi-dimensional neural network architecture. This enables high-precision intelligent judgment of battery health, faults and scenario status, significantly improving the comprehensiveness and accuracy of battery status assessment.

[0013] Furthermore, the model building unit includes: The data separation layer is used to separate multi-dimensional historical battery data to obtain battery production data, battery maintenance data, battery operation data, carrier location data, and battery positioning data. The multidimensional aggregation layer is used to aggregate battery production data, battery maintenance data, battery operation data, carrier location data, and battery positioning data according to the multidimensional transmission channels to obtain health datasets, fault datasets, and scenario datasets. The feature extraction layer is used to extract features from the health dataset based on the first extraction branch using two 3*3 convolutional blocks and one average pooling block, to obtain the health feature matrix. Based on the second extraction branch, features are extracted from the fault dataset using three 1*1 convolutional blocks and one max pooling block to obtain the fault feature matrix; Based on the third extraction branch, features are extracted from the scene dataset using two 1*1 convolutional blocks, one 3*3 convolutional block, and one adaptive pooling block to obtain the scene feature matrix; The branch training layer is used to perform several iterations of training on the health feature matrix based on the first training branch, using two parallel residual blocks and combining a time attention mechanism, to obtain the health weight matrix. Based on the second training branch, the fault feature matrix is ​​cross-trained using one inverted residual block and two parallel darknet blocks combined with a global attention mechanism to obtain the fault weight matrix; Based on the third training branch, the scene feature matrix is ​​iteratively trained using two InceptionV2 blocks and one residual block combined with a spatial attention mechanism to obtain the scene weight matrix; The aggregation training layer is used to aggregate and iterate the scene weight matrix, health weight matrix, and fault weight matrix using three InceptionV4 blocks combined with a local attention mechanism to obtain the initial comprehensive weight matrix. The identification output layer is used to comprehensively identify multi-dimensional historical battery data by combining two fully connected blocks, one random deactivated block and a multi-label classification loss function with an initial comprehensive weight matrix, and output the remaining battery life, battery health, battery failure mode, failure risk level and scene attribution. The comparison and judgment layer is used to compare the remaining battery life, battery health, battery failure mode, failure risk level and scenario attribution based on the preset multi-dimensional battery labels. If either the remaining battery life or the battery health is different from the corresponding battery label, it indicates that the health weight matrix has not met the standard. The preset iterative constraint parameters are corrected, and the current health weight matrix is ​​iteratively trained again until both are the same. If either the battery fault mode or the fault risk level differs from the corresponding fault label, it indicates that the fault weight matrix has not met the standard. The iteration constraint parameters are then corrected, and the current fault weight matrix is ​​iterated and trained again until both are the same. If the scene attribution differs from the corresponding scene label, it indicates that the scene weight matrix has not met the standard. The iteration constraint parameters are then corrected, and the current scene weight matrix is ​​iterated and trained again until both are the same. If at least two items in the above weight matrix fail to meet the standard, the corresponding failure degree is calculated, and the corresponding weight correction coefficient is generated. The update optimization layer is used to correct the multi-label classification loss function according to the weight correction coefficient, obtain the optimal weight matrix, and obtain a multi-dimensional intelligent judgment model.

[0014] By adopting the above technical solution, a multi-branch convolutional neural network is used to extract health, fault and scene features by combining residual and Inception blocks respectively. Attention mechanism and closed-loop iterative optimization algorithm are used to train and correct the weight matrix of each branch. Finally, a multi-task intelligent judgment model that can accurately predict battery life, health, fault mode and scene attribution at the same time is constructed, realizing one-stop high-precision assessment of battery status.

[0015] Furthermore, the comprehensive analysis unit includes: The branch detection layer is used to detect and identify health detection data through the health weight matrix of the multi-dimensional intelligent judgment model, and output the remaining battery life and battery health. By using the fault weight matrix of the multidimensional intelligent judgment model, the fault identification data is predicted, and the battery fault mode and fault risk level are output. By using the scene weight matrix of the multidimensional intelligent judgment model, scene classification data is classified and identified, and the scene attribution is output. The comprehensive detection layer is used to analyze and detect the battery digital profile through the optimal weight matrix of the multi-dimensional intelligent judgment model, and output a multi-dimensional detection parameter set. The analysis and judgment layer is used to compare and judge the remaining battery life, battery health, battery failure mode, failure risk level and scenario attribution based on the multi-dimensional detection parameter group, and calculate the scenario difference. If the scene difference is greater than or equal to the preset scene judgment threshold, the current scene is determined to be correct, and a battery alarm message is generated.

[0016] By adopting the above technical solution, the battery health, fault and scenario features are extracted layer by layer through the weight matrix of the multi-dimensional intelligent judgment model. After the comprehensive detection layer fuses and analyzes the data, the analysis and judgment layer compares the output results of multiple channels and calculates the scenario difference to verify consistency. Finally, an alarm is triggered only when the difference exceeds the limit, thus realizing high-precision and high-reliability intelligent judgment and fault warning of battery status.

[0017] Secondly, this application also provides a lithium battery intelligent management method, which adopts the following technical solution; A lithium battery intelligent management method includes: The system monitors the battery's operation, collects battery operating parameters, and locates and tracks the battery's real-time position to obtain battery location data. Clean and encrypt battery operating parameters and battery location data to obtain encrypted operating parameters and encrypted location data; Based on the data transmission channel, the received and decrypted encrypted operating parameters and encrypted location data are verified to obtain secure operating parameters and secure location data. Based on the battery code, safety operation parameters, safety location data, battery production information and historical maintenance records are integrated and stored to form a battery digital archive; Based on the multidimensional intelligent judgment model, the battery digital file is deconstructed and analyzed to identify the state, and the multidimensional scene state of the battery is output. Analyze the multi-dimensional scenario states of the battery, extract abnormal scenario information, and match and generate abnormal scenario resolution strategies; The optimal multidimensional intelligent judgment model is obtained by comparing and optimizing the actual battery operating conditions, actual test reports, and actual anomaly resolution strategies.

[0018] By adopting the above technical solution, battery cleaning data is collected through sensor fusion and adaptive filtering algorithms. After encrypted transmission, a data integration algorithm is used to construct a digital battery profile. Then, a cloud-based machine learning model is used for multi-dimensional intelligent analysis and solution generation. Finally, a feedback learning algorithm is used to continuously optimize the model, forming a safe, accurate, and self-evolving intelligent management system for the entire battery life cycle.

[0019] Thirdly, this application also provides a storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the lithium battery intelligent management method as described above.

[0020] In summary, the beneficial technical effects of this application are as follows: Data on battery cleaning is collected through sensor fusion and adaptive filtering algorithms. After encrypted transmission, data integration algorithms are used to construct digital files for the battery. Then, multi-dimensional intelligent analysis is performed through cloud-based machine learning models to generate solutions. Finally, feedback learning algorithms are used to continuously optimize the model, forming a safe, accurate, and self-evolving intelligent management system for the entire battery life cycle. The multidimensional intelligent judgment model extracts battery health, fault and scenario features in layers through the weight matrices of each layer. After fusion and analysis by the comprehensive detection layer, the analysis and judgment layer compares the output results of multiple channels and calculates the scenario difference to verify consistency. Finally, an alarm is triggered only when the difference exceeds the limit, realizing high-precision and high-reliability intelligent judgment and fault warning of battery status. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the lithium battery intelligent management system in this application; Figure 2 This is a schematic diagram of the structure of the multidimensional intelligent judgment model in this application; Figure 3 This is a flowchart illustrating the intelligent lithium battery management method described in this application. Detailed Implementation

[0022] The present application will be further described in detail below with reference to the accompanying drawings.

[0023] Reference Figure 1 The present application discloses a lithium battery intelligent management system, comprising: The intelligent battery management terminal is used to collect and clean battery operating parameters and battery location data, and transmit them to the cloud computing platform; The cloud computing platform is used to receive, integrate, and intelligently analyze battery operating parameters and battery location data, and combine them with a multi-dimensional intelligent judgment model to output the multi-dimensional scenario status of the battery and generate abnormal scenario resolution strategies. The user terminal is used to receive abnormal scenario resolution strategies and battery digital profiles, and supports work order scheduling, progress inquiry and self-service processing. The after-sales service terminal is used to receive repair work orders and battery alarm information, and to upload the actual battery operating conditions, actual test reports, and actual fault resolution strategies.

[0024] Preferably, the intelligent battery management terminal includes: The sensor module is used to monitor the battery's operation, collect battery operating parameters, locate and track the battery's real-time position, and obtain battery position data. The data processing module is used to filter and denoise, remove duplicate values, fill missing values, and normalize battery operating parameters and battery location data according to timestamps to obtain standard operating parameters and standard location data. The data encryption module is used to encrypt standard operating parameters and standard location data to obtain encrypted operating parameters and encrypted location data. The communication module is used to verify the communication protocol with the cloud computing platform and establish a data transmission channel to transmit encrypted operating parameters and encrypted location data.

[0025] Preferably, the cloud computing platform includes: The data receiving module is used to receive and decrypt encrypted operating parameters and encrypted location data according to the data transmission channel to obtain secure operating parameters and secure location data; The data integration module is used to integrate and store safe operating parameters, safe location data, battery production information and historical maintenance records based on the battery code, forming a digital battery profile; The intelligent analysis module is used to deconstruct and analyze the battery digital profile and identify its status based on a multi-dimensional intelligent judgment model, and output the battery's multi-dimensional scene status. The intelligent decision-making module is used to analyze the multi-dimensional scenario status of the battery, extract abnormal scenario information, and match and generate abnormal scenario resolution strategies. The feedback learning module is used to compare and optimize the multidimensional intelligent judgment model based on the actual battery operating conditions, actual test reports, and actual anomaly resolution strategies to obtain the optimal multidimensional judgment model.

[0026] In this embodiment, the intelligent BMS terminal integrates a sensor module, a data processing module, and a communication module, which are used to collect parameters such as battery voltage, current, temperature, SOC, SOH, and location information in real time.

[0027] The cloud platform includes a data receiving module, a data storage module, a data integration module, an AI analysis engine, a decision push module, and a feedback learning module; it is used to receive data, integrate and analyze the data to form a complete digital battery profile, and perform AI intelligent analysis on the data to assess the remaining battery life and health status, and push the data to users and after-sales service.

[0028] The user terminal is used to receive scenario analysis reports and solutions, and supports work order scheduling, progress inquiry, and self-service processing.

[0029] The after-sales service terminal is used to receive repair work orders and alarm information, and to upload on-site test data, handling solutions and maintenance results.

[0030] The implementation principle of this embodiment is as follows: The intelligent battery management terminal collects battery operating parameters and location data in real time through the sensor module. After filtering, noise reduction, and normalization by the data processing module, the data is encrypted and transmitted to the cloud computing platform by the data encryption module. After receiving and decrypting the data, the platform integrates production and historical information through the data integration module to construct a digital file for the battery. The intelligent analysis module uses a multi-dimensional intelligent judgment model based on machine learning algorithms to evaluate the battery's health, lifespan, and risk status. The intelligent decision-making module generates anomaly resolution strategies based on these strategies and pushes them to users and after-sales service terminals. Finally, the feedback learning module continuously optimizes the model based on the actual working conditions and processing results from the field feedback, achieving high-precision monitoring and adaptive operation and maintenance of the battery status. Example

[0031] In the management of shared electric bicycles in cities, a smart battery management system (BMS) is integrated into the battery system of each electric bicycle. It collects key battery operating parameters (such as voltage, current, temperature, SOC, SOH, etc.) in real time through sensor modules and tracks the battery's real-time location using positioning functions, forming raw data collection of battery operating parameters and location data. The data processing module then cleans the collected raw data: filtering and denoising the data using timestamp verification, removing duplicate values, filling in missing values, and normalizing the values, ultimately outputting standard operating parameters and standard location data. The data encryption module uses a secure algorithm (such as AES) to encrypt the cleaned standard data, ensuring privacy and security during transmission. The communication module establishes a stable data transmission channel by verifying the communication protocol, uploading the encrypted operating parameters and location data to the cloud computing platform.

[0032] After receiving and decrypting the data, the cloud computing platform verifies data integrity and generates safe operating parameters and safe location data. The data integration module uses the battery code as a unique identifier to link and integrate the safety data with the battery's production information (such as brand, model, and manufacturing date) and historical maintenance records (such as repair time, maintenance personnel, and maintenance content), forming a complete battery digital profile, which serves as the core data source for subsequent analysis. The intelligent analysis module, based on a multi-dimensional intelligent judgment model (such as LSTM and other time-series analysis algorithms), deconstructs and analyzes the battery digital profile, identifying the battery's health status (such as aging level), remaining lifespan, and potential fault risks, and outputs the battery's multi-dimensional scenario status (such as "battery high temperature abnormality requires cooling detection" or "SOH below threshold requires cell replacement").

[0033] The intelligent decision-making module further analyzes multi-dimensional scenario states, extracts key information from abnormal scenarios (such as temperature thresholds and location offsets), matches it with a pre-set abnormal scenario resolution strategy library, and generates specific repair suggestions (such as "it is recommended to prioritize checking the BMS cooling system" or "it is recommended to have a nearby cooperative repair shop handle it"). The feedback learning module continuously receives actual battery operating conditions (such as temperature changes after repair), actual test reports (such as capacity test results), and actual abnormal resolution strategies reported by maintenance personnel uploaded by after-sales service terminals. It compares these with the original model output and optimizes the multi-dimensional intelligent judgment model through incremental learning to improve the accuracy of future state recognition and decision-making.

[0034] User terminals (such as mobile apps or WeChat mini-programs) receive scenario analysis reports and solutions pushed by the cloud computing platform. Users can directly book repair services within the app, check work order progress in real time (e.g., "Under Repair" or "Resolved"), and conduct preliminary troubleshooting of simple problems through self-service functions (e.g., remote diagnostic guidance). After-sales service terminals (such as the management system of repair points) receive repair work orders and battery alarm information assigned by the platform. After completing on-site inspection and maintenance, repair personnel upload the actual test data of the battery (e.g., temperature, voltage curves), the implemented treatment plan (e.g., cell replacement, BMS parameter reset), and maintenance results (e.g., repair status, remaining life update) through this terminal. The cloud platform synchronously updates the battery digital profile and optimizes the model based on the new data, significantly improving the operation and maintenance efficiency and safety of shared electric bicycle batteries.

[0035] Preferably, the intelligent analysis module includes: The information deconstruction unit is used to deconstruct and aggregate the battery digital profile according to the data verification dimensions to construct health detection data, fault identification data, and scenario classification data; The model building unit is used to iteratively update the model based on multi-dimensional historical battery data and deep learning algorithms to generate a multi-dimensional intelligent judgment model. The comprehensive analysis unit is used to perform multidimensional analysis on health detection data, fault identification data, and scene classification data based on different recognition network branches of the multidimensional intelligent judgment model, and to determine the multidimensional scene status of the battery.

[0036] In this embodiment, when an electric bicycle's battery triggers an alarm due to abnormal high temperature, the information deconstruction unit of the intelligent analysis module first deconstructs the battery's digital profile, extracting historical operating parameters (such as temperature curves and charging records), location data (such as areas with frequent high temperatures), and maintenance records (such as whether the battery cells have been replaced). This data is then aggregated to generate health detection data (such as current State of Health), fault identification data (such as potential cooling system malfunctions), and scenario classification data (such as usage patterns under high-temperature environments). The model building unit, based on massive amounts of historical battery data (such as the degradation patterns of similar batteries under high temperatures) and deep learning algorithms (such as LSTM time-series prediction), iteratively updates the multi-dimensional intelligent judgment model, improving its ability to identify complex scenarios. The comprehensive analysis unit, through different recognition network branches of the model, analyzes the health status (such as remaining lifespan), fault risk (such as the probability of cooling system failure), and scenario characteristics (such as the correlation between high-temperature environments and riding habits). Ultimately, it determines whether the battery is in a multi-dimensional scenario state of "abnormal high temperature requiring cooling detection" or "external environment overheating requiring user to store in a dark place," and generates targeted solutions (such as pushing high-temperature warnings, recommending nearby repair shops, or guiding users to adjust their parking locations).

[0037] In this embodiment, the intelligent BMS terminal collects parameters such as battery voltage, current, temperature, SOC, SOH, and location information in real time, and transmits the data to the cloud platform via 4G / 5G network with encryption.

[0038] The cloud platform receives terminal data and integrates it with battery production information, historical maintenance records, etc., to form a complete digital archive of the battery. Distributed storage technology is used to ensure data security and traceability.

[0039] The AI ​​analytics engine performs multi-dimensional analysis of battery data: it uses a battery health model to assess the remaining lifespan and health status of the battery; it identifies potential fault modes and risk levels through a fault prediction algorithm; and it determines the current after-sales scenario (damage, lifespan, theft, etc.) based on a scenario classifier.

[0040] Based on the analysis results, the system automatically triggers the corresponding after-sales process, pushing a scenario analysis report to the user and after-sales service personnel, and providing them with relevant solutions. The user can then choose to handle the issue themselves or have it handled by after-sales service personnel based on the recommended solutions.

[0041] In the event of battery damage, users can schedule a service appointment, generate a repair work order, and send it to the nearest service center. For scenarios involving lifespan degradation, optimized usage suggestions are generated and pushed to users, with after-sales service personnel providing online guidance and follow-up. In the event of theft, the anti-theft mechanism is automatically triggered, and alarm information is simultaneously sent to the user and after-sales service personnel.

[0042] After receiving the scenario analysis report and solution, after-sales service personnel perform on-site maintenance on the lithium battery; they also upload actual test reports and handling solutions based on the actual operating conditions of the lithium battery. The system collects after-sales handling result data and compares and analyzes it with AI-pushed data to continuously optimize the AI ​​model, improve the accuracy of fault prediction and the efficiency of after-sales handling.

[0043] Preferably, refer to Figure 2 The model building units include: The data separation layer is used to separate multi-dimensional historical battery data to obtain battery production data, battery maintenance data, battery operation data, carrier location data, and battery positioning data. The multidimensional aggregation layer is used to aggregate battery production data, battery maintenance data, battery operation data, carrier location data, and battery positioning data according to the multidimensional transmission channels to obtain health datasets, fault datasets, and scenario datasets. In this embodiment, battery production data: Health: Production data (such as raw material batches and process parameters) is the basis for assessing the initial quality of the battery and predicting its lifespan. For example, if the purity of the raw materials in a batch of batteries is insufficient, it may lead to accelerated capacity decay.

[0044] Faults: Production defects (such as poor welding) may directly cause short circuits or thermal runaway. Problem batches can be quickly located through production data.

[0045] After-sales scenario: Production data is used for quality traceability. For example, if the battery failure rate of a certain model is high, the same batch of batteries can be recalled to reduce losses.

[0046] Battery maintenance data: Health status: Maintenance data (such as charge / discharge cycles, temperature records) reflects the actual wear and tear of the battery. For example, a battery with more than 500 cycle cycles may have less than 80% capacity.

[0047] Faults: Maintenance data can provide early warnings of potential faults, such as a battery with a persistently high temperature, which may indicate a risk of thermal runaway.

[0048] After-sales scenario: Maintenance data is used for fault analysis. For example, if a battery catches fire, historical temperature data can be retrieved to determine whether it was caused by overcharging.

[0049] Battery operating data: Health status: Operating data (such as voltage and current) reflects the battery status in real time. For example, large voltage fluctuations in a battery may indicate capacity degradation.

[0050] Fault: The operating data can trigger an alarm in real time. For example, if the current of a certain battery suddenly increases, a short circuit may occur.

[0051] After-sales scenario: Operational data is used for accident tracing. For example, if a vehicle spontaneously combusts, data from 30 seconds before the incident (sampling interval ≤ 1 second) can be retrieved to analyze the cause.

[0052] Vehicle location data: After-sales scenario: Location data is used to quickly locate faulty vehicles, such as a vehicle with a battery failure, which can navigate to the nearest repair shop.

[0053] Battery location data: After-sales scenario: Location data is used for battery recycling and traceability. For example, if a retired battery needs to be reused or a stolen battery is recovered, its location can be found through location data.

[0054] The feature extraction layer is used to extract features from the health dataset based on the first extraction branch using two 3*3 convolutional blocks and one average pooling block, to obtain the health feature matrix. Based on the second extraction branch, features are extracted from the fault dataset using three 1*1 convolutional blocks and one max pooling block to obtain the fault feature matrix; Based on the third extraction branch, features are extracted from the scene dataset using two 1*1 convolutional blocks, one 3*3 convolutional block, and one adaptive pooling block to obtain the scene feature matrix; The branch training layer is used to perform several iterations of training on the health feature matrix based on the first training branch, using two parallel residual blocks and combining a time attention mechanism, to obtain the health weight matrix. Based on the second training branch, the fault feature matrix is ​​cross-trained using one inverted residual block and two parallel darknet blocks combined with a global attention mechanism to obtain the fault weight matrix; Based on the third training branch, the scene feature matrix is ​​iteratively trained using two InceptionV2 blocks and one residual block combined with a spatial attention mechanism to obtain the scene weight matrix; The aggregation training layer is used to aggregate and iterate the scene weight matrix, health weight matrix, and fault weight matrix using three InceptionV4 blocks combined with a local attention mechanism to obtain the initial comprehensive weight matrix. The identification output layer is used to comprehensively identify multi-dimensional historical battery data by combining two fully connected blocks, one random deactivated block and a multi-label classification loss function with an initial comprehensive weight matrix, and output the remaining battery life, battery health, battery failure mode, failure risk level and scene attribution. The comparison and judgment layer is used to compare the remaining battery life, battery health, battery failure mode, failure risk level and scenario attribution based on the preset multi-dimensional battery labels. If either the remaining battery life or the battery health is different from the corresponding battery label, it indicates that the health weight matrix has not met the standard. The preset iterative constraint parameters are corrected, and the current health weight matrix is ​​iteratively trained again until both are the same. If either the battery fault mode or the fault risk level differs from the corresponding fault label, it indicates that the fault weight matrix has not met the standard. The iteration constraint parameters are then corrected, and the current fault weight matrix is ​​iterated and trained again until both are the same. If the scene attribution differs from the corresponding scene label, it indicates that the scene weight matrix has not met the standard. The iteration constraint parameters are then corrected, and the current scene weight matrix is ​​iterated and trained again until both are the same. If at least two items in the above weight matrix fail to meet the standard, the corresponding failure degree is calculated, and the corresponding weight correction coefficient is generated. The update optimization layer is used to correct the multi-label classification loss function according to the weight correction coefficient, obtain the optimal weight matrix, and obtain a multi-dimensional intelligent judgment model.

[0055] In this embodiment, the data separation layer first decomposes historical battery data into production data (such as cell batches), maintenance data (such as replacement records), operation data (such as charge and discharge curves), carrier location data (such as vehicle GPS trajectory), and battery positioning data (such as installation location). The multidimensional aggregation layer generates a health dataset (integrating production date, cycle count, etc.), a fault dataset (fusing abnormal logs such as voltage jumps and temperature changes), and a scene dataset (combining location information and environmental temperature and humidity) through independent channels. In the feature extraction layer: the first branch uses double 3×3 convolutions and average pooling to extract feature matrices such as capacity decay trends from the health dataset; the second branch uses three sets of 1×1 convolutions and max pooling to extract short-circuit / overheating feature patterns from the fault dataset; and the third branch uses a combination of 1×1 and 3×3 convolutions and adaptive pooling to separate scene features such as high-frequency use in high-temperature areas from the scene dataset. The branch training layers respectively: iteratively train the health feature matrix with residual blocks and temporal attention (capturing the temporal correlation of lifespan decay); cross-train fault features with inverted residual blocks and Darknet blocks in conjunction with global attention (identifying multi-component coupled faults); and iteratively train scene features with InceptionV2 and residual blocks in conjunction with spatial attention (analyzing the distribution of geographic hot zones). The aggregation training layer fuses three types of weight matrices through an InceptionV4 block and applies local attention to optimize the initial comprehensive weights; the identification output layer outputs the remaining lifetime, SOH value, fault type (such as cooling failure), risk level (such as high risk), and scene attribution (such as "high temperature dense short distance") through a fully connected layer and random deactivation operation. The comparison and judgment layer compares the output value with the measured label. If there is a deviation in lifespan / SOH, the health branch iteration constraint is corrected. If the fault type / level is incorrect, the fault branch parameters are adjusted. If the scenario classification is out of sync, the scenario branch is optimized. If multiple criteria are not met, the weight correction coefficient is calculated. Finally, the update and optimization layer optimizes the loss function based on this coefficient to generate the optimal weight matrix and completes the model iteration.

[0056] Preferably, the comprehensive analysis unit includes: The branch detection layer is used to detect and identify health detection data through the health weight matrix of the multi-dimensional intelligent judgment model, and output the remaining battery life and battery health. By using the fault weight matrix of the multidimensional intelligent judgment model, the fault identification data is predicted, and the battery fault mode and fault risk level are output. By using the scene weight matrix of the multidimensional intelligent judgment model, scene classification data is classified and identified, and the scene attribution is output. The comprehensive detection layer is used to analyze and detect the battery digital profile through the optimal weight matrix of the multi-dimensional intelligent judgment model, and output a multi-dimensional detection parameter set. The analysis and judgment layer is used to compare and judge the remaining battery life, battery health, battery failure mode, failure risk level and scenario attribution based on the multi-dimensional detection parameter group, and calculate the scenario difference. If the scene difference is greater than or equal to the preset scene judgment threshold, the current scene is determined to be correct, and a battery alarm message is generated.

[0057] In this embodiment: When the battery of an electric bicycle triggers an alarm due to abnormal high temperature, the branch detection layer of the comprehensive analysis unit first analyzes the battery's historical health detection data (such as SOH and cycle count) through the health weight matrix, and outputs the current remaining battery life (e.g., expected usable cycles reduced by 30%) and health (e.g., SOH below 80%). Simultaneously, it analyzes fault identification data (e.g., sudden temperature curve changes, voltage anomalies) through the fault weight matrix, predicting the fault mode (e.g., cooling system failure) and risk level (e.g., high risk). Then, it classifies and identifies scene classification data (e.g., geographical location, ambient temperature) through the scene weight matrix, outputting the scene affiliation (e.g., "high-temperature dense short-distance riding area"). The comprehensive detection layer then calls the optimal weight matrix of the multi-dimensional intelligent judgment model to perform deep analysis of the battery's digital archive (including production information, historical maintenance, and real-time parameters), generating a multi-dimensional detection parameter set containing remaining life, health, fault mode, etc. The analysis and judgment layer compares these parameters with preset labels and calculates the scenario difference (e.g., the actual remaining lifespan deviates from the predicted value by ≥15%). If the difference exceeds the threshold, the current scenario is determined to be correct and a battery alarm message is generated (e.g., "High temperature environment accelerates aging, it is recommended to check the cooling system as soon as possible"), triggering the user terminal's work order appointment and the after-sales service terminal's repair response.

[0058] Reference Figure 3 This application discloses a lithium battery intelligent management method, comprising: S1: Monitor the battery's operation process, collect battery operating parameters, locate and track the battery's real-time position, and obtain battery position data; S2: Clean and encrypt battery operating parameters and battery location data to obtain encrypted operating parameters and encrypted location data; S3: Based on the data transmission channel, verify the received and decrypted encrypted operating parameters and encrypted location data to obtain secure operating parameters and secure location data; S4: Based on the battery code, integrate and store safe operating parameters, safe location data, battery production information and historical maintenance records to form a battery digital archive; S5: Based on the multi-dimensional intelligent judgment model, analyze and identify the battery digital file, and output the multi-dimensional scene status of the battery; S6: Analyze the multi-dimensional scene status of the battery, extract abnormal scene information, and match and generate abnormal scene resolution strategies; S7: Based on the actual battery operating conditions, actual test reports, and actual anomaly resolution strategies, the multi-dimensional intelligent judgment model is compared and optimized to obtain the optimal multi-dimensional judgment model.

[0059] The implementation principle of this embodiment is as follows: battery data is collected and uploaded in real time through sensors and encrypted transmission algorithms. Digital archives are constructed in the cloud based on data integration and machine learning algorithms, and multi-dimensional intelligent analysis is performed to generate anomaly resolution strategies. Finally, the judgment model is continuously optimized through closed-loop feedback algorithms, thereby realizing safe monitoring and adaptive intelligent operation and maintenance throughout the entire battery life cycle.

[0060] This application discloses a storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the lithium battery intelligent management method as described above.

[0061] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A lithium battery intelligent management system, characterized in that, include: The intelligent battery management terminal is used to collect and clean battery operating parameters and battery location data, and transmit them to the cloud computing platform; The cloud computing platform is used to receive, integrate, and intelligently analyze the battery operating parameters and battery location data, and combine them with a multi-dimensional intelligent judgment model to output the battery multi-dimensional scenario status and generate abnormal scenario resolution strategies.

2. The lithium battery intelligent management system according to claim 1, characterized in that, The intelligent battery management terminal includes: The sensor module is used to monitor the battery's operation, collect battery operating parameters, locate and track the battery's real-time position, and obtain battery position data. The data processing module is used to filter and denoise, remove duplicate values, fill missing values, and normalize the battery operating parameters and battery location data according to the timestamp to obtain standard operating parameters and standard location data. The data encryption module is used to encrypt the standard operating parameters and the standard location data to obtain encrypted operating parameters and encrypted location data; The communication module is used to verify the communication protocol with the cloud computing platform and establish a data transmission channel to transmit the encrypted operating parameters and the encrypted location data.

3. The lithium battery intelligent management system according to claim 1, characterized in that, The cloud computing platform includes: The data receiving module is used to receive and decrypt encrypted operating parameters and encrypted location data according to the data transmission channel to obtain secure operating parameters and secure location data; The data integration module is used to integrate and store the safe operating parameters, the safe location data, the battery production information and historical maintenance records according to the battery code, to form a battery digital archive; The intelligent analysis module is used to deconstruct and analyze the battery digital file and identify its status based on a multi-dimensional intelligent judgment model, and output the battery multi-dimensional scene status. The intelligent decision-making module is used to analyze the multi-dimensional scenario state of the battery, extract abnormal scenario information, and match and generate abnormal scenario resolution strategies. The feedback learning module is used to compare and optimize the multidimensional intelligent judgment model based on the actual battery operating conditions, actual test reports, and actual anomaly resolution strategies to obtain the optimal multidimensional judgment model.

4. The lithium battery intelligent management system according to claim 3, characterized in that, The intelligent analysis module includes: The information deconstruction unit is used to deconstruct and aggregate the battery digital profile according to the data verification dimensions to construct health detection data, fault identification data, and scenario classification data; The model building unit is used to iteratively update the model based on multi-dimensional historical battery data and deep learning algorithms to generate a multi-dimensional intelligent judgment model. The comprehensive analysis unit is used to perform multidimensional analysis on the health detection data, the fault identification data, and the scene classification data based on different recognition network branches of the multidimensional intelligent judgment model, and to determine the multidimensional scene status of the battery.

5. The lithium battery intelligent management system according to claim 4, characterized in that, The model building unit includes: The data separation layer is used to separate multi-dimensional historical battery data to obtain battery production data, battery maintenance data, battery operation data, carrier location data, and battery positioning data. The multidimensional aggregation layer is used to aggregate the battery production data, battery maintenance data, battery operation data, carrier location data and battery positioning data according to the multidimensional transmission channels to obtain a health dataset, a fault dataset and a scenario dataset. The feature extraction layer is used to extract features from the health dataset based on the first extraction branch using two 3*3 convolutional blocks and one average pooling block to obtain a health feature matrix. Based on the second extraction branch, features are extracted from the fault dataset using three 1*1 convolutional blocks and one max pooling block to obtain a fault feature matrix; Based on the third extraction branch, features are extracted from the scene dataset using two 1*1 convolutional blocks, one 3*3 convolutional block, and one adaptive pooling block to obtain the scene feature matrix; The branch training layer is used to perform several iterations of training on the health feature matrix based on the first training branch using two parallel residual blocks and a time attention mechanism to obtain the health weight matrix. According to the second training branch, the fault feature matrix is ​​cross-trained using one inverted residual block and two parallel darknet blocks combined with a global attention mechanism to obtain the fault weight matrix. According to the third training branch, the scene feature matrix is ​​iteratively trained using two InceptionV2 blocks and one residual block combined with a spatial attention mechanism to obtain the scene weight matrix; The aggregation training layer is used to aggregate and iterate the scene weight matrix, the health weight matrix, and the fault weight matrix using three InceptionV4 blocks combined with a local attention mechanism to obtain an initial comprehensive weight matrix.

6. The lithium battery intelligent management system according to claim 5, characterized in that, The model building unit also includes: The identification output layer is used to comprehensively identify multi-dimensional historical battery data by combining the initial comprehensive weight matrix with two fully connected blocks, one random deactivated block and a multi-label classification loss function, and output the remaining battery life, battery health, battery failure mode, failure risk level and scene attribution. The comparison and judgment layer is used to compare the remaining battery life, battery health, battery failure mode, failure risk level and scenario attribution based on preset multi-dimensional battery tags. If either the remaining battery life or the battery health is different from the corresponding battery label, it indicates that the health weight matrix has not met the standard. The preset iterative constraint parameters are corrected, and the current health weight matrix is ​​iteratively trained again until both are the same. If either the battery failure mode or the failure risk level differs from the corresponding failure label, it indicates that the failure weight matrix has not met the standard. The iterative constraint parameters are then corrected, and the current failure weight matrix is ​​iteratively trained again until both are the same. If the scene attribution differs from the corresponding scene label, it indicates that the scene weight matrix has not met the standard. The iterative constraint parameters are then corrected, and the current scene weight matrix is ​​iteratively trained again until both are the same. If at least two items in the above weight matrix fail to meet the standard, the corresponding failure degree is calculated, and the corresponding weight correction coefficient is generated. An update optimization layer is used to correct the multi-label classification loss function according to the weight correction coefficient, obtain the optimal weight matrix, and acquire a multi-dimensional intelligent judgment model.

7. The lithium battery intelligent management system according to claim 6, characterized in that, The comprehensive analysis unit includes: The branch detection layer is used to detect and identify health detection data through the health weight matrix of the multi-dimensional intelligent judgment model, and output the remaining battery life and battery health. The fault weight matrix of the multidimensional intelligent judgment model is used to predict the fault identification data and output the battery fault mode and fault risk level. The scene weight matrix of the multidimensional intelligent judgment model is used to classify and identify scene classification data and output the scene attribution. The comprehensive detection layer is used to analyze and detect the battery digital file through the optimal weight matrix of the multi-dimensional intelligent judgment model, and output a multi-dimensional detection parameter set. The analysis and judgment layer is used to compare and judge the remaining battery life, battery health, battery failure mode, failure risk level and scenario attribution based on the multi-dimensional detection parameter group, and calculate the scenario difference degree. If the scene difference is greater than or equal to the preset scene judgment threshold, the current scene is determined to be correct, and a battery alarm message is generated.

8. The lithium battery intelligent management system according to claim 1, characterized in that, The intelligent management system also includes: The user terminal is used to receive abnormal scenario resolution strategies and battery digital profiles, and supports work order scheduling, progress inquiry and self-service processing. The after-sales service terminal is used to receive repair work orders and battery alarm information, and to upload the actual battery operating conditions, actual test reports, and actual fault resolution strategies.

9. A lithium battery intelligent management method, applied to the system described in any one of claims 1-8, characterized in that, include: The system monitors the battery's operation, collects battery operating parameters, and locates and tracks the battery's real-time position to obtain battery location data. The battery operating parameters and battery location data are cleaned and encrypted to obtain encrypted operating parameters and encrypted location data; Based on the data transmission channel, the encrypted operating parameters and the encrypted location data are received and decrypted to obtain the secure operating parameters and secure location data. Based on the battery code, the safe operating parameters, the safe location data, battery production information, and historical maintenance records are integrated and stored to form a battery digital archive; Based on the multidimensional intelligent judgment model, the battery digital file is deconstructed, analyzed and identified in terms of state, and the battery multidimensional scene state is output. The battery's multi-dimensional scenario state is analyzed, abnormal scenario information is extracted, and abnormal scenario resolution strategies are matched and generated. The multidimensional intelligent judgment model is optimized by comparing and analyzing the actual battery operating conditions, actual test reports, and actual anomaly resolution strategies to obtain the optimal multidimensional judgment model.

10. A storage medium storing at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the lithium battery intelligent management method as described in claim 9.

Images