CN118275903B - Battery performance test method based on data analysis - Google Patents
Battery performance test method based on data analysis Download PDFInfo
- Publication number
- CN118275903B CN118275903B CN202410712934.1A CN202410712934A CN118275903B CN 118275903 B CN118275903 B CN 118275903B CN 202410712934 A CN202410712934 A CN 202410712934A CN 118275903 B CN118275903 B CN 118275903B
- Authority
- CN
- China
- Prior art keywords
- data
- battery performance
- model
- support vector
- vector machine
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000011056 performance test Methods 0.000 title claims abstract description 34
- 238000007405 data analysis Methods 0.000 title claims abstract description 16
- 238000012360 testing method Methods 0.000 claims abstract description 57
- 238000012706 support-vector machine Methods 0.000 claims abstract description 55
- 238000013145 classification model Methods 0.000 claims abstract description 37
- 238000005457 optimization Methods 0.000 claims abstract description 25
- 230000002159 abnormal effect Effects 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000004140 cleaning Methods 0.000 claims abstract description 11
- 238000004458 analytical method Methods 0.000 claims description 31
- 230000008569 process Effects 0.000 claims description 24
- 238000005516 engineering process Methods 0.000 claims description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 15
- 238000012544 monitoring process Methods 0.000 claims description 14
- 238000007781 pre-processing Methods 0.000 claims description 14
- 230000005540 biological transmission Effects 0.000 claims description 13
- 230000008859 change Effects 0.000 claims description 13
- 230000008030 elimination Effects 0.000 claims description 12
- 238000003379 elimination reaction Methods 0.000 claims description 12
- 238000002790 cross-validation Methods 0.000 claims description 11
- 238000007689 inspection Methods 0.000 claims description 10
- 239000000463 material Substances 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 7
- 238000007726 management method Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 239000002245 particle Substances 0.000 claims description 5
- 230000005856 abnormality Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 230000036314 physical performance Effects 0.000 claims description 4
- 238000012552 review Methods 0.000 claims description 4
- 230000002596 correlated effect Effects 0.000 claims description 3
- 230000000875 corresponding effect Effects 0.000 claims description 3
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 claims description 3
- 238000007637 random forest analysis Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 230000004044 response Effects 0.000 abstract description 7
- 238000010998 test method Methods 0.000 abstract description 2
- 238000003860 storage Methods 0.000 description 11
- 230000007774 longterm Effects 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000013500 data storage Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000012827 research and development Methods 0.000 description 3
- 238000013144 data compression Methods 0.000 description 2
- 238000013523 data management Methods 0.000 description 2
- 238000013079 data visualisation Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000014759 maintenance of location Effects 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000003139 buffering effect Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 230000008713 feedback mechanism Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000019637 foraging behavior Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 238000011031 large-scale manufacturing process Methods 0.000 description 1
- 229910001416 lithium ion Inorganic materials 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000013450 outlier detection Methods 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
Landscapes
- Secondary Cells (AREA)
Abstract
The application particularly relates to a battery performance test method based on data analysis, which comprises the following steps: constructing a dynamic data set, synchronizing and archiving data in real time, cleaning and checking quality of battery performance data, automatically identifying and eliminating abnormal values of the battery performance data, and filling missing battery performance data; training a support vector machine classification model capable of accurately distinguishing battery performance levels; and (3) automatic parameter adjustment and model self-optimization, and testing the battery performance by using a support vector machine classification model after the automatic parameter adjustment and model self-optimization. The application constructs a highly integrated, automatic and intelligent battery performance test system, and solves the limitations of the traditional test method in the aspects of efficiency, precision, response speed and cost control.
Description
Technical Field
The invention belongs to the field of batteries, and particularly relates to a battery performance testing method based on data analysis.
Background
Under the current rapid development background of new energy science and technology and the electric automobile industry, accurate testing and analysis of battery performance become key factors for ensuring product safety, prolonging service life and improving user experience. With the continuous progress of battery technology, new energy storage solutions including lithium ion batteries and solid state batteries are attracting attention, and these technological innovations place higher demands on battery performance tests. Traditional battery testing methods often rely on static laboratory testing and manual data analysis, and are difficult to meet the requirements of rapid and accurate evaluation of battery performance in large-scale production and diversified application scenes. In recent years, the integration application of technologies such as big data analysis, internet of things (IoT), cloud computing, artificial Intelligence (AI) and the like brings revolutionary changes to battery performance testing. Particularly, the data analysis technology can reveal the subtle variation trend of the battery performance, predict potential faults and optimize the battery design and manufacturing flow by integrating massive historical and real-time test data. However, how to efficiently construct dynamic data sets, synchronize data in real time, perform advanced data processing, and make performance predictions using advanced machine learning algorithms remains a technical challenge for the industry.
Although the prior art also discloses a solid-state battery performance test technology, the core of the technology is to optimize the detection flow by dynamically adjusting the detection proportion and the data analysis of the depth, so as to improve the detection efficiency and the accuracy. However, such techniques preset standard subjectivity and timeliness: the system relies on preset standard curves and thresholds to make performance decisions, the setting of these standards may be subjective and as battery technology evolves and materials advance, the standards need to be updated regularly, otherwise the accuracy of the test results may be affected.
Disclosure of Invention
The invention aims to provide a battery performance testing method based on data analysis, so as to solve the problems in the background technology.
In order to solve the technical problems, the invention provides the following technical scheme:
A battery performance test method based on data analysis is characterized in that a dynamic data set is constructed, a database containing historical and real-time battery performance test data is established, and the battery performance data comprises battery cores and physical performance indexes under various working conditions;
The data are synchronized and archived in real time, the Internet of things is adopted, the data are directly captured from the battery performance testing equipment in real time, and the data are automatically uploaded to the cloud server; meanwhile, data archiving is carried out, and historical data is stored according to time sequence or batch classification;
cleaning and quality inspection of battery performance data are carried out, abnormal values of the battery performance data are automatically identified and removed, and missing battery performance data are filled;
Based on historical battery performance test data, using a support vector machine algorithm, selecting an optimal kernel function and parameter setting through cross validation, and training a support vector machine classification model capable of accurately distinguishing battery performance levels; in the characteristic selection stage, determining key parameters with the greatest influence on the battery performance by adopting a recursive characteristic elimination or correlation coefficient analysis-based method, wherein the key parameters comprise charge rate change, discharge platform stability and internal resistance growth rate, and the key parameters are used as support vector machine classification model input;
Automatic parameter adjustment and model self-optimization, on-line learning of a support vector machine classification model is configured, so that the support vector machine classification model can periodically review newly acquired test data without interrupting service, and parameters of the support vector machine classification model are automatically adjusted; searching for a better super-parameter combination by utilizing a gradient descent or particle swarm optimization algorithm so as to adapt to subtle changes of battery performance along with time and technical evolution;
and testing the battery performance by using a support vector machine classification model after automatic parameter adjustment and model self-optimization.
Further, capturing data from the battery performance test device in real time, and automatically uploading the data to the cloud server specifically comprises:
The method comprises the steps of carrying out Internet of things upgrading on battery performance testing equipment, integrating an intelligent sensor and a Micro Controller Unit (MCU), wherein the components are responsible for accurately acquiring performance data of a battery under various working conditions; and monitoring the running state of the equipment in real time; adopting MQTT (Message Queuing Telemetry Transport) low-power consumption wide area network (LPWAN) communication protocol to realize high-efficiency and safe data transmission between the test equipment and the cloud server; deploying edge computing nodes, which are positioned between the testing equipment and the cloud end, and executing preliminary data preprocessing and screening tasks;
In the data transmission process, all communication data are encrypted by adopting a TLS/SSL protocol, so that the safety and privacy protection of the data in the transmission process are ensured; the cloud server develops a RESTful API interface, receives data pushing from an edge computing node or directly from equipment, automatically analyzes the data pushing and stores the data pushing into a distributed database; the database adopts Amazon DynamoDB or Google Cloud Spanner cloud native database;
And (3) building a real-time data monitoring platform at the cloud, displaying the latest data and trend analysis of each test point, setting a threshold alarm mechanism, and immediately notifying related personnel through an email, a short message or an APP once the battery performance abnormality or the test equipment fault is detected, so as to quickly respond.
Further, the specific steps for data archiving include:
Adopting a time sequence database management system, and storing and inquiring and optimizing the data of the battery performance which changes along with time; during the data archiving process, a unique identifier is created for each lot of batteries and an efficient index structure is built in the database.
Further, cleaning and quality inspection of battery performance data are carried out, abnormal values of the battery performance data are automatically identified and removed, and missing battery performance data are filled, and the method comprises the following steps:
An automatic data preprocessing pipeline is constructed, the automatic data preprocessing pipeline is integrated before the data is stored after being uploaded to the cloud, and the data is cleaned and controlled in quality before entering a core database; automatically identifying abnormal values in the battery performance data by using a box graph analysis method, a Z-score method and an abnormality detection technology based on a clustering algorithm; filling missing values in the time sequence data by adopting a linear interpolation, spline interpolation or a time sequence prediction method based on an ARIMA model, so as to ensure data continuity;
for the missing value of the non-time series data, the K-nearest neighbor regression model or the random forest regression model can be trained through the complete data to predict the missing value;
the patterns or correlations in the data are analyzed, and the missing values are reasonably estimated by nearest neighbor matching or mean/median filling methods using the data characteristics of similar batches or material sets as references.
Further, using a support vector machine algorithm, selecting an optimal kernel function and parameter setting through cross-validation, training a support vector machine classification model capable of accurately distinguishing battery performance levels includes: determining an optimal kernel function and corresponding parameters by combining grid search with cross-validation by using a plurality of kernel functions; the performance of the model is comprehensively evaluated by adopting various evaluation indexes, including the accuracy, the recall rate, the F1 fraction and the area under an AUC-ROC curve, so that the model can be accurately classified, and the prediction effect of each category can be balanced.
Further, by adopting a method of recursive feature elimination or correlation coefficient analysis, determining the key parameters with the greatest influence on the battery performance comprises:
In the feature selection stage, firstly, starting an RFE process from a complete battery performance feature set, and focusing on accuracy, recall and F1 score evaluation indexes; calculating importance scores of each feature by using a support vector machine model of preliminary training; removing one or more features with the lowest scores from the current feature set, retraining the support vector machine model, and reevaluating model performance;
setting a threshold value to determine when to stop feature elimination, and verifying the validity of the feature subset at each step through cross verification to ensure the robustness of the feature selection process;
Before the correlation coefficient analysis, carrying out standardization processing on the battery performance data, eliminating dimension influence, and evaluating the correlation between each pair of characteristics by using a Pierson correlation coefficient or a Speermann correlation coefficient;
identifying features associated with battery performance while taking into account multiple collinearity between the features; eliminating redundant characteristics which have low correlation with a target variable and are highly correlated with other characteristics, and reserving key parameters which have the greatest influence on the battery performance and are mutually independent, wherein the key parameters comprise charge rate change, discharge platform stability and internal resistance increase rate;
Combining the features screened based on the correlation coefficient analysis with the result of the RFE process, preferentially considering the features confirmed to be important in both methods, and finally determining the feature set input into the support vector machine classification model.
Further, the battery performance test specifically includes:
After the new battery performance test data is uploaded to a cloud server in real time through the Internet of things equipment, the steps of format unification, missing value inspection and filling, abnormal value identification and processing pretreatment are carried out through a data preprocessing pipeline;
the preprocessed data is then sent to the latest optimized support vector machine classification model;
the support vector machine model evaluates the performance of each battery batch based on the selected key features and classifies it into different performance levels.
The beneficial effects are that: the technical scheme provided by the application realizes remarkable technical leaps in the field of battery performance test:
The test precision and efficiency are obviously improved: by constructing a dynamic data set containing historical and real-time data and combining advanced internet of things technology and edge calculation, the scheme realizes the instant collection and efficient transmission of battery performance data, and the test period is obviously shortened. The intelligent application of the Support Vector Machine (SVM) classification model and the automatic parameter adjustment and self-optimization mechanism ensure the high precision of the test result, adapt to the change of the battery performance along with the technical development through continuous learning and improve the test efficiency and the generalization capability of the model.
Enhancement decision and response speed: the real-time data monitoring platform and the alarm system are established, and the cloud instant data analysis is combined, so that abnormal battery performance or equipment faults can be rapidly identified and countermeasures can be triggered, the safety in production and use is effectively improved, the downtime caused by faults is reduced, and the operation efficiency of enterprises is enhanced.
Optimizing resource management and reducing cost: the intelligent data archiving and efficient data storage strategy greatly optimizes the data management flow and reduces the storage and operation cost. The data preprocessing and the model self-optimization reduce manual intervention, reduce labor cost, improve data quality and analysis efficiency, and provide precious data resources for long-term performance tracking and research and development.
The flexibility and the adaptability of the system are improved: through the intelligent strategy of online learning and feature selection, the technical scheme can flexibly adapt to continuous innovation of battery technology, ensures that the test model always keeps the analysis capability of the forefront, is crucial to rapidly changing battery materials and designs, and is beneficial to battery manufacturers to rapidly respond to market and technical changes.
In summary, the technical progress of the application is to construct a highly integrated, automatic and intelligent battery performance test system, which solves the limitations of the traditional test method in terms of efficiency, precision, response speed and cost control by deeply fusing big data, internet of things, cloud computing and AI technology.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The application discloses a battery performance test method based on data analysis, as shown in figure 1, comprising the following steps:
Constructing a dynamic data set, and establishing a database containing historical and real-time battery performance test data, wherein the battery performance data comprises electric cores and physical performance indexes under various working conditions, and the battery performance database is updated regularly to ensure timeliness and comprehensiveness of model training data; the battery performance database comprises basic information of production batch numbers, material compositions, manufacturing dates and technical specifications of the batteries, and detailed test data of battery cell performance (such as charge and discharge cycle times, capacity retention rate and internal resistance change) and physical performance (such as thermal stability, mechanical strength and safety performance test results); each battery performance test data is associated with battery performance test environment parameters (such as temperature and humidity), so that traceability and comparability of test results are ensured;
The data are synchronized and archived in real time, the data are directly captured from the battery performance testing equipment in real time by adopting the Internet of things, and the data are automatically uploaded to a cloud server, so that the freshness of the data is ensured;
capturing data from battery performance test equipment in real time, and automatically uploading the data to a cloud server, wherein the specific implementation details are as follows:
internet of things architecture configuration and implementation
Intelligent transformation of equipment end: the battery performance test equipment is subjected to Internet of things upgrading, an intelligent sensor and a microcontroller unit (MCU) are integrated, and the components are responsible for accurately acquiring performance data of the battery under various working conditions, including but not limited to current, voltage, temperature and humidity, monitoring the running state of the equipment in real time and ensuring the continuity and accuracy of data acquisition;
Standardized communication protocol: adopting MQTT (Message Queuing Telemetry Transport) low-power consumption wide area network (LPWAN) communication protocol to realize high-efficiency and safe data transmission between the test equipment and the cloud server; the MQTT protocol is particularly suitable for real-time data transmission requirements in the application of the Internet of things due to the characteristics of light weight and low bandwidth consumption, supports a publish/subscribe mode, and is convenient for the management and data distribution of large-scale equipment;
edge calculation optimization: deploying edge computing nodes, which are positioned between the testing equipment and the cloud end, and executing preliminary data preprocessing and screening tasks; the edge calculation can effectively reduce the data transmission delay, lighten the processing load of the cloud server, for example, perform data denoising and outlier filtering operation, and ensure that the data uploaded to the cloud is fresh and has high value;
Secure encryption and identity authentication: in the data transmission process, all communication data are encrypted by adopting a TLS/SSL protocol, so that the safety and privacy protection of the data in the transmission process are ensured; meanwhile, a unique identity is allocated to each test device, and a strict access control and identity authentication mechanism is implemented to prevent unauthorized access and data tampering;
Cloud platform interface and data storage: the cloud server develops a RESTful API interface, receives data pushing from an edge computing node or directly from equipment, automatically analyzes the data pushing and stores the data pushing into a distributed database; the database adopts Amazon DynamoDB or Google Cloud Spanner cloud native database, so as to ensure high availability, durability and elastic expansion capability of the data;
real-time monitoring and alarm system: setting up a real-time data monitoring platform at a cloud end, displaying the latest data and trend analysis of each test point, setting a threshold alarm mechanism, and immediately notifying related personnel through an email, a short message or an APP once abnormal battery performance or test equipment failure is detected, so as to quickly respond to the processing;
meanwhile, data archiving is carried out, historical data is stored according to time sequence or batch classification, and long-term tracking and analysis are facilitated;
Specific implementation details for data archiving include the following:
Time series data archiving strategy: adopting a time sequence database management system (such as InfluxDB or TimescaleDB) to store and query and optimize the data of the battery performance which changes along with time; the database can efficiently process high-frequency data writing and complex time sequence inquiry, supports rapid retrieval of historical data according to time ranges and batch numbers, and provides convenience for long-term performance trend analysis;
Batch sort indexing mechanism: in the process of data archiving, creating a unique identifier for each batch of batteries, and establishing an efficient index structure in a database; the index not only comprises a batch number, but also comprises key metadata of production date, material type and technical specification, so that battery performance records of specific batches or material groups can be rapidly positioned, and batch-to-batch performance comparison and trend research are facilitated;
data compression and storage optimization: considering the mass characteristics of battery performance data, adopting an intelligent data compression algorithm to reduce the occupation of storage space on the premise of not affecting the data integrity and analysis precision; meanwhile, a data layering storage strategy is implemented, frequently accessed recent data is stored in a high-speed storage medium (such as SSD), and older historical data is migrated to a low-cost cold storage solution, so that storage cost and access efficiency are balanced;
Data lifecycle management: making a data retention policy and automatically managing the life cycle of data; for example, the set rule automatically cleans historical data exceeding a specified age, or dynamically adjusts a storage hierarchy according to the data access frequency, so that important and common data is ensured to be always easy to access, and storage resources are reasonably released;
Data visualization and report generation: developing customized data visualization tools or integrating existing BI (business intelligence) platforms, such as Tableau or Power BI, so that researchers and engineers can intuitively view and analyze long-term archived data; support generation of periodic performance reports including, but not limited to, batch performance comparison reports, time series trend analysis reports, helping decision makers to quickly grasp battery performance change laws and potential problems;
Data backup and restore planning: ensuring high availability and disaster recovery capability of a data archiving system, and copying data to a remote or cloud storage service through a periodic full-scale backup and incremental backup strategy; the AWS S3 or Google Cloud Storage service is adopted, so that high reliability and quick recovery capability are provided, history data can be quickly recovered even in the face of unexpected situations, and the continuity of long-term tracking and analysis work is not affected;
Through the detailed implementation strategy, the data archiving flow not only ensures the safe storage and efficient organization of the historical data, but also promotes deep insight into the battery performance and long-term trend analysis, performs battery performance data cleaning and quality inspection, automatically identifies and eliminates abnormal values of the battery performance data, fills up missing battery performance data, and ensures the accuracy and the integrity of warehouse-in data;
Specific implementation details for performing battery performance data cleaning and quality inspection include the following steps:
1. Constructing a data preprocessing pipeline: an automatic data preprocessing pipeline is constructed, the pipeline is integrated before the data is stored after being uploaded to the cloud, and the data is cleaned and controlled in quality before entering a core database; the pipeline comprises, but is not limited to, data analysis, uniform format, outlier detection and processing and missing value filling key links;
2. outlier identification and processing:
Statistical methods are combined with machine learning: automatically identifying abnormal values in the battery performance data by using a box graph analysis method, a Z-score method and an abnormality detection technology based on a clustering algorithm; for the found abnormal value, different processing strategies are adopted according to the distribution characteristics and business logic, such as direct elimination, marked as abnormal but reserved, or smooth processing according to the adjacent normal value;
Context-aware adjustment: considering that the performance of the battery is influenced by testing conditions (such as temperature and humidity), the abnormal value identification process can combine with testing environment parameters to carry out context-sensitive adjustment so as to avoid misjudgment;
3. Missing value filling strategy:
Interpolation based on time series: filling missing values in the time sequence data by adopting a linear interpolation, spline interpolation or a time sequence prediction method based on an ARIMA model, so as to ensure data continuity;
model prediction filling: for the missing value of the non-time series data, the K-nearest neighbor regression model or the random forest regression model can be trained through the complete data to predict the missing value;
filling modes: analyzing patterns or correlations in the data, reasonably estimating the missing values by using data characteristics of similar batches or material groups as references through a nearest neighbor matching or mean/median filling method;
4. data quality monitoring and feedback loop:
Real-time data quality monitoring: embedding a real-time monitoring module in the data cleaning flow, continuously tracking data quality indexes such as missing value proportion and abnormal value frequency, and triggering early warning once exceeding a preset threshold;
Closed loop feedback and continuous optimization: establishing a data quality feedback mechanism, recording and feeding back the problems identified in the cleaning process and the measures taken to a front-end data acquisition and equipment maintenance team, and promoting the quality improvement of a data acquisition source; meanwhile, continuously adjusting a cleaning algorithm and parameters according to feedback information to form a continuously optimized closed loop;
Through the detailed implementation details, the data cleaning and quality inspection process is tightly integrated with the whole technical scheme, so that the high quality and reliability of battery performance test data are ensured, and a solid foundation is laid for subsequent model training and performance evaluation;
Based on historical battery performance test data, using a support vector machine algorithm, selecting an optimal kernel function and parameter setting through cross validation, and training a support vector machine classification model capable of accurately distinguishing battery performance levels;
Selecting the optimal kernel function and parameter setting through cross validation by using a support vector machine algorithm, and training a support vector machine classification model capable of accurately distinguishing the battery performance level comprises the following steps: using a plurality of kernel functions (such as linear kernel, RBF kernel and polynomial kernel), determining the optimal kernel function and corresponding parameters (such as C and gamma) by combining grid search (GRID SEARCH) with cross-validation (k-fold cross-validation is adopted, and k value can be selected according to data quantity and calculation resources, such as 5-fold or 10-fold); however, when implementing automatic parameter adjustment, it is possible to consider using an optimization strategy built in, for example SMO (Sequential Minimal Optimization), libSVM, and fine-tuning the loss function (such as the hinge loss function) if necessary, so as to adapt to the specificity of the battery performance data;
The performance of the model is comprehensively evaluated by adopting various evaluation indexes, including accuracy, recall rate, F1 fraction and area under an AUC-ROC curve, so that the model can be accurately classified and the prediction effect of each category can be balanced;
In the face of different kernel functions and parameter combinations, a plurality of models with similar performances exist, and prediction results of the models can be integrated through voting and bagging, boosting integrated learning methods so as to improve the stability and accuracy of overall prediction;
In the characteristic selection stage, determining key parameters with the greatest influence on the battery performance, such as charge rate change, discharge platform stability and internal resistance growth rate, by adopting a recursive characteristic elimination (RFE) or correlation coefficient analysis-based method, and inputting the key parameters as a support vector machine classification model;
Feature selection stage to ensure that the selected features are able to maximally characterize battery performance and improve accuracy of model predictions, we resort to the following implementation details to perform Recursive Feature Elimination (RFE) and correlation coefficient analysis-based methods:
recursive Feature Elimination (RFE)
Initializing and defining a target: firstly, starting an RFE process from a complete battery performance characteristic set, wherein the explicit aim is to optimize the performance of a support vector machine classification model, and particularly pay attention to accuracy, recall and F1 score evaluation indexes;
Feature scoring and ordering: calculating importance scores of each feature by using a support vector machine model of preliminary training; this is based on feature weights, i.e. the extent to which they contribute to model decision boundaries; then, arranging the features in descending order of scores;
Gradual feature elimination: removing one or more features with the lowest scores from the current feature set, retraining the support vector machine model, and reevaluating model performance; this process is repeated, and the model is re-evaluated after each feature removal until a predetermined feature quantity threshold is reached or the model performance is no longer significantly improved;
threshold and iteration strategy: setting a threshold value to determine when to stop feature elimination, such as when the performance of the model is reduced by more than a certain proportion after the next feature is eliminated; in addition, the robustness of the feature selection process can be ensured by cross-verifying the validity of the feature subset at each step;
Based on correlation coefficient analysis
Data preprocessing: before the correlation coefficient analysis is carried out, the battery performance data is subjected to standardized processing, the dimension influence is eliminated, and the fairness and significance of the correlation among different features are ensured;
calculating a correlation coefficient matrix: evaluating the correlation between each pair of features using Pearson (Pearson) correlation coefficients or Spearman (Spearman) level correlation coefficients;
Feature selection: identifying features associated with battery performance (e.g., class labels) while taking into account multiple collinearity between the features; removing redundant characteristics which have low correlation with a target variable and are highly correlated with other characteristics, and reserving key parameters which have the greatest influence on the battery performance and are mutually independent, such as charge rate change, discharge platform stability and internal resistance increase rate;
Binding RFE results: combining the features screened based on the correlation coefficient analysis with the result of the RFE process, preferentially considering the features which are confirmed to be important in both methods, and finally determining a feature set input into a classification model of the support vector machine;
Comprehensive implementation
In the whole feature selection process, a mode of combining iteration and cross verification is adopted, so that the selected feature set can reflect key indexes of battery performance and can keep stable performance on different data subsets; by the comprehensive method, a simple and efficient feature set can be obtained, the generalization capability and the practicability of the support vector machine classification model are further improved, and the accuracy and the reliability of the support vector machine classification model in battery performance test are ensured;
Automatic parameter adjustment and model self-optimization, on-line learning of a support vector machine classification model is configured, so that the support vector machine classification model can periodically review newly acquired test data without interrupting service, and parameters of the support vector machine classification model are automatically adjusted; in the process, a gradient descent or particle swarm optimization algorithm is utilized to find a better super-parameter combination so as to adapt to subtle changes of battery performance along with time and technical evolution; meanwhile, the support vector machine classification model performs self-verification regularly, and the accuracy and the reliability of the support vector machine classification model are ensured through comparison with a manually marked test result;
The specific steps of automatic parameter adjustment and model self-optimization are as follows:
Online learning mechanism configuration
1. Real-time data access and buffering
Data access layer: configuring a data access module to be integrated at an API interface of a cloud server and responsible for receiving latest battery performance test data pushed by an edge computing node in real time; data is temporarily stored in a cache to reduce the impact on the online learning module directly, ensuring service stability;
2. Timing triggers and data batching
A scheduler: implementing an intelligent scheduler, and automatically triggering review and parameter adjustment processes of the model according to preset time intervals (such as daily, weekly or on demand); the dispatcher is simultaneously responsible for sorting new data in the buffer area in batches to form a small batch data set suitable for model training so as to control the consumption of computing resources and response time of training;
3. parameter optimization algorithm implementation
Gradient descent optimization: aiming at parameters (such as C and gamma) of a support vector machine, adopting random gradient descent (SGD) or Batch Gradient Descent (BGD), and gradually adjusting super parameters according to gradient information on a new data set when the model is reviewed each time so as to optimize the performance of the model;
Particle Swarm Optimization (PSO): as a global optimization algorithm, PSO simulates the foraging behavior of a bird group, and searches for an optimal solution in a solution space through a plurality of groups of particles (representing different parameter combinations); searching a super-parameter space in parallel by using a PSO algorithm, and searching for parameter configuration capable of maximizing model accuracy;
4. Model verification and self-assessment
Cross-validation and a/B test: after each parameter adjustment, a cross verification technology (such as k-fold cross verification) is applied to verify the new model version, so that the improvement of the model performance is ensured; meanwhile, an A/B test is implemented, so that a part of actual test data is processed by the new model and the old model in parallel, and the prediction result is compared, so that improved effectiveness is ensured;
5. Dynamic adjustment and rollback mechanism
Performance monitoring and dynamic adjustment: deploying a performance monitoring module, and tracking the prediction accuracy, response time and resource consumption of the model in real time; if the performance of the new model version is reduced or the resource consumption is overlarge, automatically rolling back to the last stable version by the system, and recording an abnormal log for further analysis;
self-adaptive learning rate adjustment: dynamically adjusting the learning rate according to the model convergence condition and the change of data distribution; reducing the learning rate when the model approaches an optimal solution for fine tuning; when encountering abrupt change of data distribution, temporarily increasing learning rate and accelerating adaptation to new conditions;
6. safety and stability assurance
Model version control: the Git or similar version control system is adopted to record the version adjusted by each model, so that any time point can be quickly traced back to the previous stable version;
resource isolation and load balancing: the online learning process runs in an isolated computing environment and is separated from the main service, and the online adjustment process is ensured not to influence the real-time data processing and service stability through the load balancing technology of the cloud platform;
By implementing the specific details, the support vector machine classification model can be continuously and self-optimized to adapt to dynamic changes of battery performance test data, and meanwhile, the continuity of service and the reliability of the model are ensured;
Testing the battery performance by using a support vector machine classification model after automatic parameter adjustment and model self-optimization;
The test for battery performance specifically includes:
real-time data access and processing:
After new battery performance test data is uploaded to a cloud server in real time through Internet of things equipment, format unification, missing value inspection and filling, outlier identification and processing pretreatment steps are carried out through a data pretreatment assembly line, and data quality is ensured to meet model input requirements;
The preprocessed data is then sent to the latest optimized support vector machine classification model; because the model has online learning capability, the model can process the data in real time without offline retraining, thereby ensuring the timeliness of the test result;
Performance testing and grading:
The support vector machine model evaluates the performance of each batch of batteries based on selected key characteristics (such as charge rate change, discharge platform stability and internal resistance increase rate) and divides the batteries into different performance levels; the model can identify which batteries have excellent performance and which need to be concerned or eliminated, and provides basis for subsequent quality control and product improvement.
The technical scheme provided by the application realizes remarkable technical leap in the field of battery performance test, and compared with the traditional method, the technical scheme has the technical progress and advantages of multiple dimensions, and the technical effects are particularly shown in the following points:
The test precision and efficiency are obviously improved: by constructing a dynamic data set containing historical and real-time data and combining advanced internet of things technology and edge calculation, the scheme realizes the instant collection and efficient transmission of battery performance data, and the test period is obviously shortened. The intelligent application of the Support Vector Machine (SVM) classification model and the automatic parameter adjustment and self-optimization mechanism ensure the high precision of the test result, adapt to the change of the battery performance along with the technical development through continuous learning and improve the test efficiency and the generalization capability of the model.
Enhancement decision and response speed: the real-time data monitoring platform and the alarm system are established, and the cloud instant data analysis is combined, so that abnormal battery performance or equipment faults can be rapidly identified and countermeasures can be triggered, the safety in production and use is effectively improved, the downtime caused by faults is reduced, and the operation efficiency of enterprises is enhanced.
Optimizing resource management and reducing cost: the intelligent data archiving and efficient data storage strategy greatly optimizes the data management flow and reduces the storage and operation cost. The data preprocessing and the model self-optimization reduce manual intervention, reduce labor cost, improve data quality and analysis efficiency, and provide precious data resources for long-term performance tracking and research and development.
The flexibility and the adaptability of the system are improved: through the intelligent strategy of online learning and feature selection, the technical scheme can flexibly adapt to continuous innovation of battery technology, ensures that the test model always keeps the analysis capability of the forefront, is crucial to rapidly changing battery materials and designs, and is beneficial to battery manufacturers to rapidly respond to market and technical changes.
Promote technical innovation and competitive upgrade: the technical scheme not only improves the scientificity and efficiency of battery performance test, but also accelerates the research and development process of battery technology and improves the product quality and market competitiveness by providing a powerful data analysis platform. The method provides powerful technical support for the electric automobile industry and related research institutions, and promotes the technical innovation and sustainable development of the whole new energy industry chain.
In summary, the technical progress of the application is to construct a highly integrated, automatic and intelligent battery performance testing system, which solves the limitations of the traditional testing method in terms of efficiency, precision, response speed and cost control by deeply fusing big data, internet of things, cloud computing and AI technology, brings a technical innovation to the battery industry, and promotes the industry to develop towards more efficient, intelligent and sustainable directions.
It should be noted that, in the present specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment is mainly described in a different point from other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, with reference to the description of the method embodiments in part. The apparatus and system embodiments described above are merely illustrative, in which elements illustrated as separate elements may or may not be physically separate, and elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
Claims (1)
1. The battery performance test method based on data analysis is characterized by comprising the following steps:
Constructing a dynamic data set, and establishing a database containing historical and real-time battery performance test data, wherein the battery performance data comprises electric cores and physical performance indexes under various working conditions;
The data are synchronized and archived in real time, the Internet of things is adopted, the data are directly captured from the battery performance testing equipment in real time, and the data are automatically uploaded to the cloud server; meanwhile, data archiving is carried out, and historical data is stored according to time sequence or batch classification;
cleaning and quality inspection of battery performance data are carried out, abnormal values of the battery performance data are automatically identified and removed, and missing battery performance data are filled;
Based on historical battery performance test data, using a support vector machine algorithm, selecting an optimal kernel function and parameter setting through cross validation, and training a support vector machine classification model capable of accurately distinguishing battery performance levels; in the characteristic selection stage, determining key parameters with the greatest influence on the battery performance by adopting a recursive characteristic elimination or correlation coefficient analysis-based method, wherein the key parameters comprise charge rate change, discharge platform stability and internal resistance growth rate, and the key parameters are used as support vector machine classification model input;
Automatic parameter adjustment and model self-optimization, on-line learning of a support vector machine classification model is configured, so that the support vector machine classification model can periodically review newly acquired test data without interrupting service, and parameters of the support vector machine classification model are automatically adjusted; searching for a better super-parameter combination by utilizing a gradient descent or particle swarm optimization algorithm so as to adapt to subtle changes of battery performance along with time and technical evolution;
Testing the battery performance by using a support vector machine classification model after automatic parameter adjustment and model self-optimization;
Capturing data from battery performance testing equipment in real time, and automatically uploading the data to a cloud server specifically comprises:
The battery performance test equipment is subjected to Internet of things upgrading, an intelligent sensor and a microcontroller unit are integrated, and the components are responsible for accurately acquiring performance data of the battery under various working conditions; and monitoring the running state of the equipment in real time; adopting an MQTT low-power-consumption wide area network communication protocol to realize efficient and safe data transmission between the test equipment and the cloud server; deploying edge computing nodes, which are positioned between the testing equipment and the cloud end, and executing preliminary data preprocessing and screening tasks;
In the data transmission process, all communication data are encrypted by adopting a TLS/SSL protocol, so that the safety and privacy protection of the data in the transmission process are ensured; the cloud server develops a RESTful API interface, receives data pushing from an edge computing node or directly from equipment, automatically analyzes the data pushing and stores the data pushing into a distributed database; the database adopts Amazon DynamoDB or Google Cloud Spanner cloud native database;
setting up a real-time data monitoring platform at a cloud end, displaying the latest data and trend analysis of each test point, setting a threshold alarm mechanism, and immediately notifying related personnel through an email, a short message or an APP once abnormal battery performance or test equipment failure is detected, so as to quickly respond to the processing;
The data archiving method specifically comprises the following steps:
Adopting a time sequence database management system, and storing and inquiring and optimizing the data of the battery performance which changes along with time; in the process of data archiving, creating a unique identifier for each batch of batteries, and establishing an efficient index structure in a database;
cleaning and quality inspection of battery performance data are carried out, abnormal values of the battery performance data are automatically identified and removed, and missing battery performance data are filled, and the method comprises the following steps:
An automatic data preprocessing pipeline is constructed, the automatic data preprocessing pipeline is integrated before the data is stored after being uploaded to the cloud, and the data is cleaned and controlled in quality before entering a core database; automatically identifying abnormal values in the battery performance data by using a box graph analysis method, a Z-score method and an abnormality detection technology based on a clustering algorithm; filling missing values in the time sequence data by adopting a linear interpolation, spline interpolation or a time sequence prediction method based on an ARIMA model, so as to ensure data continuity;
for the missing value of the non-time series data, the K-nearest neighbor regression model or the random forest regression model can be trained through the complete data to predict the missing value;
Analyzing patterns or correlations in the data, reasonably estimating the missing values by using data characteristics of similar batches or material groups as references through a nearest neighbor matching or mean/median filling method;
Selecting the optimal kernel function and parameter setting through cross validation by using a support vector machine algorithm, and training a support vector machine classification model capable of accurately distinguishing the battery performance level comprises the following steps: determining an optimal kernel function and corresponding parameters by combining grid search with cross-validation by using a plurality of kernel functions; the performance of the model is comprehensively evaluated by adopting various evaluation indexes, including accuracy, recall rate, F1 fraction and area under an AUC-ROC curve, so that the model can be accurately classified and the prediction effect of each category can be balanced;
The method for determining key parameters with the greatest influence on the battery performance by adopting recursive feature elimination or based on correlation coefficient analysis comprises the following steps:
In the feature selection stage, firstly, starting an RFE process from a complete battery performance feature set, and focusing on accuracy, recall and F1 score evaluation indexes; calculating importance scores of each feature by using a support vector machine model of preliminary training; removing one or more features with the lowest scores from the current feature set, retraining the support vector machine model, and reevaluating model performance;
setting a threshold value to determine when to stop feature elimination, and verifying the validity of the feature subset at each step through cross verification to ensure the robustness of the feature selection process;
Before the correlation coefficient analysis, carrying out standardization processing on the battery performance data, eliminating dimension influence, and evaluating the correlation between each pair of characteristics by using a Pierson correlation coefficient or a Speermann correlation coefficient;
identifying features associated with battery performance while taking into account multiple collinearity between the features; eliminating redundant characteristics which have low correlation with a target variable and are highly correlated with other characteristics, and reserving key parameters which have the greatest influence on the battery performance and are mutually independent, wherein the key parameters comprise charge rate change, discharge platform stability and internal resistance increase rate;
Combining the characteristics screened based on the correlation coefficient analysis with the result of the RFE process, and finally determining a characteristic set input into a support vector machine classification model;
The test for battery performance specifically includes:
After the new battery performance test data is uploaded to a cloud server in real time through the Internet of things equipment, the steps of format unification, missing value inspection and filling, abnormal value identification and processing pretreatment are carried out through a data preprocessing pipeline;
the preprocessed data is then sent to the latest optimized support vector machine classification model;
the support vector machine model evaluates the performance of each battery batch based on the selected key features and classifies it into different performance levels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410712934.1A CN118275903B (en) | 2024-06-04 | 2024-06-04 | Battery performance test method based on data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410712934.1A CN118275903B (en) | 2024-06-04 | 2024-06-04 | Battery performance test method based on data analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118275903A CN118275903A (en) | 2024-07-02 |
| CN118275903B true CN118275903B (en) | 2024-09-03 |
Family
ID=91634484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410712934.1A Active CN118275903B (en) | 2024-06-04 | 2024-06-04 | Battery performance test method based on data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118275903B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118710257A (en) * | 2024-08-27 | 2024-09-27 | 启东市航新实用技术研究所 | Substation storage battery pack operation and maintenance management method and system based on artificial intelligence |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104182630A (en) * | 2014-08-20 | 2014-12-03 | 国家电网公司 | Residual battery capacity detection method based on simplified least square support vector machine |
| CN108805217A (en) * | 2018-06-20 | 2018-11-13 | 山东大学 | A kind of health state of lithium ion battery method of estimation and system based on support vector machines |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11074522B2 (en) * | 2016-06-15 | 2021-07-27 | Roger N. Anderson | Electric grid analytics learning machine |
| CN109598449B (en) * | 2018-12-14 | 2023-09-12 | 蓝谷智慧(北京)能源科技有限公司 | Lithium ion battery pack performance evaluation method based on charging data |
| CN114976306A (en) * | 2022-03-29 | 2022-08-30 | 上海师范大学 | Intelligent echelon utilization method of power lithium battery based on vehicle-connected cloud |
| CN117872143A (en) * | 2023-12-11 | 2024-04-12 | 量道（厦门）新能源科技有限公司 | Battery abnormality detection method, battery abnormality detection device, electronic equipment and storage medium |
-
2024
- 2024-06-04 CN CN202410712934.1A patent/CN118275903B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104182630A (en) * | 2014-08-20 | 2014-12-03 | 国家电网公司 | Residual battery capacity detection method based on simplified least square support vector machine |
| CN108805217A (en) * | 2018-06-20 | 2018-11-13 | 山东大学 | A kind of health state of lithium ion battery method of estimation and system based on support vector machines |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118275903A (en) | 2024-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN118275903B (en) | Battery performance test method based on data analysis | |
| Bahga et al. | Analyzing massive machine maintenance data in a computing cloud | |
| CN112766550B (en) | Random forest-based power failure sensitive user prediction method, system, storage medium and computer equipment | |
| CN110659173A (en) | Operation and maintenance system and method | |
| US10444746B2 (en) | Method for managing subsystems of a process plant using a distributed control system | |
| CN113454552A (en) | Sensor metering data integration | |
| CN102955977A (en) | Energy efficiency service method and energy efficiency service platform adopting same on basis of cloud technology | |
| JP7401677B2 (en) | Model update system, model update method and related equipment | |
| WO2024066346A1 (en) | Alarm processing method and apparatus, and storage medium and electronic apparatus | |
| CN116992346A (en) | Enterprise production data processing system based on artificial intelligence big data analysis | |
| CN118152378B (en) | Construction method and system of intelligent data center | |
| CN115941696A (en) | Heterogeneous Big Data Distributed Cluster Storage Optimization Method | |
| CN117317408A (en) | Battery and heat optimization management method based on big data and artificial intelligence | |
| CN116070143A (en) | Power distribution network multi-source heterogeneous data fusion method and system based on artificial intelligence | |
| CN118017502A (en) | Digital twinning-based power distribution calculation power prediction method, system and medium | |
| CN114548493A (en) | Method and system for predicting current overload of electric energy meter | |
| CN113610225A (en) | Quality evaluation model training method and device, electronic equipment and storage medium | |
| Awad et al. | Proposed Methodology for Battery Aging and Drainage Mitigation | |
| CN118233325B (en) | IT equipment integrated operation and maintenance management evaluation method and system based on Internet of things | |
| CN117745239B (en) | Self-service template customization on-line filling system and method based on flow engine | |
| CN117971337A (en) | Hybrid cloud automatic configuration method based on LSTM model | |
| CN118367609B (en) | Intelligent data optimization processing system of new energy station | |
| CN117171551B (en) | Large-scale industrial equipment data analysis and intelligent management method | |
| Li et al. | Node Resource Balance Scheduling Algorithm of Power Internet of Things Based on Big Data | |
| Heistracher et al. | MOXAI–Manufacturing Optimization through Model-Agnostic Explainable AI and Data-Driven Process Tuning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |