A lithium battery life prediction method based on big data method
Through big data methods, the multi-source test data and real-time state of lithium batteries are collected and analyzed, and combined with the graph neural network model, the problem that traditional prediction methods cannot capture the actual state and environment changes of the battery are solved, and the accurate prediction and life of lithium batteries are achieved. extend.
Patent Information
- Application Number
- CN202410552611.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-07
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-05-07
AI Technical Summary
Traditional lithium battery life prediction methods cannot accurately capture the actual working state and environmental changes of the battery, and cannot consider the aging situation at different rates, resulting in rigid prediction results and the life of lithium batteries of electric vehicles cannot be accurately predicted.
The lithium battery life prediction method based on big data method is adopted to generate battery health trajectory and life prediction data through multi-source test data acquisition, real-time status monitoring, topological structure processing and graph neural network model transfer learning, considering operating conditions, usage mode and Environmental factors.
Accurate prediction of lithium battery life is achieved, charging and discharging strategies are optimized, energy loss is reduced, fault risk is reduced, battery life is extended, and overall system efficiency is improved.
Smart Images

Figure CN118425785B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of big data analysis, and in particular to a lithium battery life prediction method based on big data methods. Background Art
[0002] Lithium-ion (Li-ion) batteries are widely used in electric vehicles, renewable energy systems, and portable electronic products due to their high energy density, long cycle life, and relatively low self-discharge rate. With the development of the Internet and sensor technology, the scale and application scope of big data are constantly expanding. In the field of electric vehicles, various sensors, vehicle systems, and Internet-connected devices generate massive amounts of data. These data contain a variety of information such as the battery's charging and discharging process, temperature, voltage, and current. Accurately predicting the life of lithium-ion batteries is crucial to optimizing battery management systems, reducing waste, and improving overall efficiency. However, traditional lithium battery life prediction methods are usually based on mathematical models or experimental data, which cannot capture the actual working state and environmental changes of the battery, making lithium battery life prediction relatively rigid. At the same time, the high-speed charging and discharging of electric vehicles will accelerate the aging process of lithium batteries. Conventional life prediction models usually cannot consider aging at different rates. In addition, the usage patterns of different vehicles and users are different, so the usage methods and cycle times of lithium batteries will also be different, resulting in the inability to accurately predict the life of lithium batteries in electric vehicles. Summary of the invention
[0003] Based on this, the present invention provides a lithium battery life prediction method based on a big data method to solve at least one of the above technical problems.
[0004] To achieve the above purpose, a lithium battery life prediction method based on big data method includes the following steps:
[0005] Step S1: collecting multi-source test data of a lithium battery electric vehicle to obtain multi-source test characteristic data of the lithium battery; processing the battery transient response frequency according to the multi-source test characteristic data of the lithium battery to generate transient response frequency data;
[0006] Step S2: Monitor the battery status of the lithium battery pack to obtain battery status data; perform intelligent behavior detection based on the battery status data to generate battery behavior event data; perform optimal battery sampling rate processing on the transient response frequency data through the battery behavior event data, and perform adaptive real-time data acquisition to generate real-time lithium battery pack data;
[0007] Step S3: Process the lithium battery topology structure according to the real-time lithium battery pack data to generate battery assembly topology structure data; obtain real-time user operation behavior data and real-time environmental data; construct a time series matrix of the real-time lithium battery pack data, real-time user operation behavior data and real-time environmental data to obtain lithium battery telemetry matrix data;
[0008] Step S4: Use the preset graph neural network model to perform transfer learning on the battery connection topology data and build a single component behavior prediction model; use the single component behavior prediction model to predict the battery health trajectory of the lithium battery telemetry matrix data and perform battery attenuation curve processing to obtain battery attenuation curve data;
[0009] Step S5: Predict the battery life according to the battery attenuation curve data to generate battery predicted life data; perform closed-loop optimization processing on the battery life according to the battery predicted life data to generate battery closed-loop optimization data.
[0010] The present invention collects multi-source test data for lithium battery electric vehicles, which helps to obtain a comprehensive data set covering the performance of batteries under various working conditions. Such a data set can provide behavioral characteristics of the battery under different conditions, and comprehensively reflect the behavior of the battery under different working conditions, including multiple parameters such as charge and discharge cycles, temperature changes, current intensity, and voltage response. Subsequently, the obtained multi-source test characteristic data is used to process the battery transient response frequency. By analyzing the battery's ability to respond to rapid load changes, it is helpful to understand the dynamic performance of the battery and evaluate its ability to respond quickly to external changes. By monitoring the battery status of the lithium battery pack and obtaining battery status data, key parameters such as the battery's charging state, voltage, current, and temperature can be tracked in real time. Intelligent behavior detection is performed based on the battery status data, and advanced data analysis technology is used to identify and record abnormal behaviors of the battery in actual use, such as overcharge, overdischarge, or temperature abnormalities. The transient response frequency data is processed at the optimal battery sampling rate through the battery behavior event data, and the data acquisition frequency is optimized. The adaptive real-time data acquisition intelligently adjusts the sampling rate according to the battery's behavior and state, which not only reduces the burden of data storage and processing, but also improves the response speed and accuracy of the monitoring system. The topological structure of lithium batteries is processed according to the real-time lithium battery pack data to determine the connection relationship between battery cells and their configuration. Acquiring real-time user operation behavior data and real-time environmental data not only reflects the use status of the battery in actual operation, but also includes the impact of environmental factors such as temperature and humidity on battery performance. The construction of the time series matrix enables the data to be stored and analyzed in a unified format, which is convenient for efficient data processing and in-depth data analysis. The graph neural network model effectively learns the complex mutual influence and dependency between battery components by analyzing the connection structure between the battery cells, so that the single component behavior prediction model can accurately predict the behavior and state of each battery cell in future operation. The battery health trajectory prediction shows the changing trend of the health state of the battery during future use by analyzing the prediction data. Predicting the battery life according to the battery degradation curve data can generate accurate and reliable battery life prediction data, while considering various factors such as operating conditions, usage patterns and environmental factors, optimizing the battery charging and discharging strategy for real-time performance, reducing energy loss, and improving the overall system efficiency, and minimizing the risk of battery failure while maximizing the overall battery life. Therefore, the lithium battery life prediction method based on big data method of the present invention collects electric vehicle experimental data, intelligently formulates data sampling strategy, and uses battery behavior event detection to realize adaptive real-time collection of lithium battery status data in electric vehicles, thereby improving the efficiency of big data collection and identifying patterns and correlations that are difficult to distinguish from a single data source.The lithium battery topology is processed based on real-time lithium battery pack data, the connection relationship between battery cells is analyzed and determined, and the internal structure of the lithium battery is derived in real time. At the same time, real-time user operation behavior and environmental factors are monitored, and then the lithium battery performance attenuation analysis is performed according to different lithium battery usage modes to accurately predict the life of electric vehicle lithium batteries.
[0011] Preferably, step S1 comprises the following steps:
[0012] Step S11: Collect multi-source test data for lithium battery electric vehicles to generate lithium battery multi-source test data;
[0013] Step S12: performing data feature engineering processing according to the lithium battery multi-source test data to obtain lithium battery multi-source test feature data, wherein the lithium battery multi-source test feature data includes test battery pack state data, test vehicle driving condition data, and test driving scenario data;
[0014] Step S13: performing lithium battery working condition cluster analysis on the test battery pack status data using the test vehicle driving condition data and the test driving scenario data to obtain lithium battery working condition cluster data;
[0015] Step S14: extracting key operating characteristic data according to the lithium battery operating condition clustering data to obtain key operating characteristic data, wherein the key operating characteristic data includes charge and discharge rate characteristic data, battery load temperature gradient data, and battery load response data;
[0016] Step S15: Using the key operating characteristic data to process the battery transient response frequency, to obtain transient response frequency data.
[0017] The present invention can obtain a large amount of battery performance data under different working conditions, including charging, discharging, and performance under different temperatures, by collecting multi-source test data for lithium battery electric vehicles. Such data collection helps to capture the behavior of the battery under various working conditions in actual use, thereby providing more real and comprehensive basic data. The multi-source test data is converted and extracted to better understand the behavior of the battery, and the original data is converted into feature data that is easier to understand and apply, including information such as battery capacity, voltage change trend, and the relationship between temperature and performance. Cluster analysis is performed on the test battery pack state data using the test car driving condition data and the test driving scene data, so that the battery can be classified under different working conditions, so as to more clearly understand the performance characteristics of the battery under different conditions, and reveal the working mode and characteristics of the battery under different working conditions. The performance of the battery under different working conditions is deeply excavated, and the features that have an important impact on the battery performance are extracted. These feature data include charge and discharge rate characteristics, temperature gradients, and battery response characteristics, which can help understand the response mechanism of the battery under different working conditions. Through the battery transient response frequency processing, the dynamic response characteristics of the battery can be deeply studied. Convert key operating characteristic data into time and frequency domain information to gain a more comprehensive understanding of the battery's response process.
[0018] Preferably, step S15 comprises the following steps:
[0019] Step S151: performing spectrum conversion processing on the charge and discharge rate characteristic data and the battery load response data to obtain charge and discharge rate spectrum data and battery load response spectrum data respectively;
[0020] Step S152: performing frequency component analysis on the battery load response spectrum data to generate load response frequency component data;
[0021] Step S153: performing frequency domain step calculation using the charge and discharge rate spectrum data to generate frequency domain step data; performing linear ramp processing according to the frequency domain step data to obtain charge and discharge linear ramp data;
[0022] Step S154: performing transient behavior frequency division on the load response frequency component data by using the charge and discharge linear ramp data, and obtaining quasi-low frequency behavior range data, quasi-medium frequency behavior range data, and quasi-high frequency behavior range data respectively;
[0023] Step S155: performing overshoot and undershoot calculations on the charge and discharge rate spectrum data based on the pseudo low-frequency behavior range data, and adjusting the low-frequency range data to generate low-frequency response range data;
[0024] Step S156: performing sinusoidal wave amplitude response calculation on the charge and discharge rate spectrum data based on the pseudo intermediate frequency behavior range data, and adjusting the intermediate frequency range data to generate intermediate frequency response range data;
[0025] Step S157: Perform nonlinear thermal effect processing using the battery load temperature gradient data to generate nonlinear thermal effect data; perform thermal runaway point risk calculation based on the nonlinear thermal effect data to obtain thermal runaway point risk data; perform high-frequency response range correction on the quasi-high-frequency behavior range data using the thermal runaway point risk data to generate high-frequency response range data.
[0026] The present invention converts the charge and discharge rate characteristic data and the battery load response data into a representation in the spectrum domain by performing spectrum conversion processing on the charge and discharge rate characteristic data and the battery load response data, so as to better understand the frequency distribution of the charge and discharge process and the battery load response, and help identify the main frequency components in the charge and discharge process. Performing frequency component analysis on the battery load response spectrum data can further identify the frequency components in the load response and reveal the dynamic response of the battery at different frequencies. Using the charge and discharge rate spectrum data to perform frequency domain step calculations helps to evaluate the response speed of the battery at different frequencies, so as to better understand the transient behavior of the battery. By performing transient behavior frequency division on the load response frequency component data through the charge and discharge linear ramp data, the load response data can be divided into response data in different frequency ranges, and the behavior characteristics of the battery at different frequencies can be distinguished. Based on the quasi-low-frequency behavior range data, the overshoot and undershoot calculation of the charge and discharge rate spectrum data can evaluate the overshoot and undershoot in the charge and discharge process within the low-frequency range, which helps to optimize the control strategy of the charge and discharge process. By calculating the sine wave amplitude response of the charge and discharge rate spectrum data based on the pseudo-intermediate frequency behavior range data and adjusting the intermediate frequency range data, the battery's response to the sine wave amplitude in the intermediate frequency range can be evaluated and adjusted accordingly, which helps to understand the dynamic response characteristics of the battery in the intermediate frequency range. Using the battery load temperature gradient data for nonlinear thermal effect processing can evaluate the thermal runaway risk of the battery under different temperature gradients and make corresponding corrections. This helps to improve the safety and stability of the battery system.
[0027] Preferably, step S2 comprises the following steps:
[0028] Step S21: monitoring the battery status of the lithium battery pack through the automobile BMS battery management system to obtain battery status data, wherein the battery status data includes battery pack voltage data, battery pack current data, battery pack temperature data and battery remaining capacity data;
[0029] Step S22: defining battery status events according to the lithium battery working condition clustering data, and generating battery status event data;
[0030] Step S23: constructing a battery finite state machine model based on the battery status event data;
[0031] Step S24: using the battery finite state machine model to perform behavior state transition detection on the battery state data to generate battery behavior event data;
[0032] Step S25: Adaptively adjust the instantaneous response frequency data according to the battery behavior event data, and perform Nyquist sampling calculation to generate optimal battery sampling rate data;
[0033] Step S26: Based on the optimal battery sampling rate data, adaptive real-time data collection is performed on the lithium battery pack to generate real-time lithium battery pack data.
[0034] The present invention monitors the battery state of a lithium battery pack through an automobile BMS battery management system, and obtains comprehensive battery state data including voltage, current, temperature and remaining capacity. According to the clustering data of the lithium battery working conditions, a series of battery state events are defined to form battery state event data. The definition of these events is based on the understanding of the behavioral characteristics of the battery under different working conditions, which helps to identify and record various state changes that occur in the battery pack. A battery finite state machine model is constructed through the battery state event data, which can clearly describe the conversion law of the battery between different states. This model helps to understand the operating mechanism and state evolution process of the battery. The battery finite state machine model is used to detect the behavior state conversion of the battery state data, which reflects the switching of the battery between different states and helps to analyze the operation trajectory and abnormal behavior of the battery. Optimization based on the actual behavior and state changes of the battery helps to improve the efficiency and accuracy of data acquisition, and ensure that the real-time monitoring and analysis of the battery pack state can be more accurate and timely. Adaptive real-time data acquisition is performed on the lithium battery pack based on the optimal battery sampling rate data. These data have high real-time and accuracy and can reflect the state and performance changes of the battery pack in a timely manner.
[0035] Preferably, step S25 comprises the following steps:
[0036] Step S251: extracting battery behavior features according to the battery behavior event data to generate battery behavior feature data;
[0037] Step S252: performing behavior trend analysis on the battery behavior characteristic data to generate behavior trend data;
[0038] Step S253: performing adaptive frequency threshold calculation on the transient response frequency data to generate adaptive frequency threshold data, wherein the adaptive frequency threshold data includes an adaptive low frequency threshold, an adaptive middle frequency threshold and an adaptive high frequency threshold;
[0039] Step S254: using the behavior trend data based on the adaptive filter to identify the battery behavior frequency band of the adaptive frequency threshold data, and performing Nyquist sampling rate calculation to generate the Nyquist sampling rate;
[0040] Step S255: Perform sampling rate smoothing processing according to the Nyquist sampling rate to generate optimal battery sampling rate data.
[0041] The present invention extracts battery behavior features through battery behavior event data, and can extract key features from complex behavior data, such as peak current and temperature rise rate during charging and discharging. Performing behavior trend analysis on battery behavior feature data can reveal the behavior change trend of the battery in different time periods, such as the change trend of current and temperature during charging and discharging, which helps to discover the regularity and abnormality of the battery operation process. By performing adaptive frequency threshold calculation on transient response frequency data, thresholds of three frequency bands, low frequency, medium frequency and high frequency, are included. These threshold data can be dynamically adjusted according to the real-time behavior and state changes of the battery, ensuring that the identification and analysis of battery behavior have higher accuracy and adaptability. Based on the adaptive filter, the adaptive frequency threshold data is used to identify the battery behavior frequency band, and the frequency band can be divided according to the frequency characteristics of the battery behavior, and then the optimal sampling rate is determined to ensure the efficiency and accuracy of data acquisition. The sampling rate is smoothed according to the Nyquist sampling rate, and such processing can make the sampling rate more stable and continuous, ensuring that the monitoring and analysis of battery behavior are of high quality and high efficiency.
[0042] Preferably, step S3 comprises the following steps:
[0043] Step S31: collecting user operation behavior according to real-time lithium battery pack data to generate user operation behavior data; obtaining real-time environmental data through weather API;
[0044] Step S32: performing data association processing on the real-time lithium battery pack data, the user operation behavior data and the real-time environment data to generate original lithium battery telemetry data;
[0045] Step S33: performing local outlier factor detection on the original lithium battery telemetry data, and performing outlier elimination processing to generate standard lithium battery telemetry data;
[0046] Step S34: normalizing the standard lithium battery telemetry data and extracting relevant feature data to generate lithium battery telemetry feature data, wherein the lithium battery telemetry feature data includes lithium battery pack feature data, user operation behavior feature data and real-time environment feature data;
[0047] Step S35: performing lithium battery topology processing based on the lithium battery pack characteristic data to generate battery assembly topology data;
[0048] Step S36: construct a time series matrix for the lithium battery telemetry feature data, and annotate the matrix rows and columns with event labels to obtain lithium battery telemetry matrix data.
[0049] The present invention collects user operation behavior based on real-time lithium battery pack data, and obtains real-time environmental data in combination with weather API, which can comprehensively collect various information including battery pack status, user operation behavior and environmental conditions. Data association processing of real-time lithium battery pack data, user operation behavior data and real-time environmental data helps to integrate data from different sources to form a complete data set. The original lithium battery telemetry data is subjected to local outlier factor detection and outlier removal processing, which helps to remove outlier interference in the data and improve the quality and accuracy of the data. The standard lithium battery telemetry data is normalized. These feature data include the characteristics of various information such as battery pack status, user operation behavior and environmental conditions, which helps to fully understand the operation of the battery system from different angles. The lithium battery topology structure is processed through the lithium battery pack feature data, which helps to understand the correlation between battery components and the structural characteristics of the battery system. The time series matrix of the lithium battery telemetry feature data is constructed, the time series data is converted into a structured matrix form, and the data is annotated with event labels. Processing the lithium battery topology through lithium battery pack characteristic data can clarify the relationship and connection between battery components, reveal the dependency and mechanism of action between battery components, and better understand the overall operation of the battery system by deeply understanding the topological relationship between battery components. Converting complex time series data into a structured matrix form and annotating the data with event labels makes the data easier to understand and process. Such processing helps convert time series data into an input format that can be used for algorithms such as machine learning and deep learning.
[0050] Preferably, step S35 includes the following steps:
[0051] Step S351: Acquire lithium battery diagram structure data; construct a battery structure physical model according to the lithium battery diagram structure data;
[0052] Step S352: identifying battery component nodes according to the lithium battery graph structure data, and generating battery component node data;
[0053] Step S353: Calculate the node-related behavior coefficient for the battery assembly node data to generate the node-related behavior coefficient;
[0054] Step S354: Analyze the component connection relationship according to the node-related behavior coefficient, draw a connection relationship diagram for the lithium battery pack, and generate battery connection relationship diagram data;
[0055] Step S355: performing node connection power analysis on the battery connection relationship diagram data through the lithium battery telemetry matrix data based on the battery structure physical model to generate node connection power data;
[0056] Step S356: Use the node connection power data to perform topological structure correction on the battery connection relationship diagram data to generate battery assembly topological structure data.
[0057] The present invention can form a physical structure simulation of a battery system by acquiring lithium battery graph structure data and constructing a battery structure physical model based on it. This model can clearly describe the layout relationship and connection mode between battery components. Battery component node identification based on lithium battery graph structure data is helpful for node-level analysis and modeling of the battery system. By identifying the node information of each battery component, the composition and connection relationship of the battery system can be accurately represented. The node-related behavior coefficients of the battery component node data are calculated to quantitatively describe the related behavior relationship between the battery components. The calculation of these coefficients helps to understand the degree of influence and interaction mode between different components. The component connection relationship analysis is performed according to the node-related behavior coefficients, and a battery connection relationship diagram is drawn to clearly show the connection mode and relationship between the battery components, which helps to understand the overall structure and working principle of the battery system. Based on the battery structure physical model, the node connection dynamic analysis of the battery connection relationship diagram data is performed through the lithium battery telemetry matrix data, which can reveal the dynamic relationship and interaction effect between the battery components. Such analysis helps to understand the dynamic behavior and response characteristics of the battery system. The topological structure of the battery connection relationship diagram data is corrected using the node connection dynamic data, and the connection relationship between the battery components can be adjusted and optimized according to the power analysis results.
[0058] Preferably, step S4 comprises the following steps:
[0059] Step S41: using a preset graph neural network model to perform transfer learning on the battery connection topology data to generate an initial single component behavior prediction model;
[0060] Step S42: training the initial single component behavior prediction model using preset battery component behavior sample data, and performing model evaluation and optimization to generate a single component behavior prediction model;
[0061] Step S43: Perform frequent pattern mining on the lithium battery telemetry matrix data to generate component mining pattern data; use the component mining pattern data to perform pattern propagation and model optimization on the single component behavior prediction model to generate an optimized single component behavior prediction model;
[0062] Step S44: transmitting the lithium battery telemetry matrix data to the optimized single component behavior prediction model to perform single component behavior prediction and generate single component behavior prediction data;
[0063] Step S45: performing battery health trajectory prediction according to the single component behavior prediction data to generate battery health trajectory data;
[0064] Step S46: Process the battery decay curve based on the battery health trajectory data to obtain battery decay curve data.
[0065] The present invention utilizes a preset graph neural network model to perform transfer learning on battery connection topology data, and can apply the existing graph neural network model to new battery component topology data. Such transfer learning helps to share existing knowledge and experience between different battery systems, thereby accelerating the model training and optimization process. Frequent pattern mining processing of lithium battery telemetry matrix data can discover frequently occurring patterns and laws from a large amount of data, which helps to understand the behavioral relationships and characteristics between battery components, further improve the model's predictive ability and adaptability, and make it closer to actual scenarios. The lithium battery telemetry matrix data is transmitted to the optimized single component behavior prediction model for single component behavior prediction, and the optimized prediction model can be used to predict the actual data, thereby obtaining the behavior prediction results of a single battery component. Battery health trajectory prediction is performed based on single component behavior prediction data, and the battery health status change trend can be inferred based on the battery component behavior prediction results.
[0066] Preferably, step S5 comprises the following steps:
[0067] Step S51: performing degradation state analysis according to battery attenuation curve data to generate battery degradation state data;
[0068] Step S52: defining the action space according to the battery degradation state data to generate battery degradation action space data;
[0069] Step S53: Based on the battery degradation action space data, the battery attenuation curve data is predicted by a preset deep deterministic policy gradient to generate battery predicted life data;
[0070] Step S54: Obtaining battery theoretical life data through the vehicle BMS battery management system;
[0071] Step S55: performing difference calculation on the battery theoretical life data through the battery predicted life data, and performing dynamic difference tracking to generate cycle life difference data;
[0072] Step S56: Perform three-dimensional attribution analysis based on the cycle life difference data, and perform battery life closed-loop optimization processing to generate battery life closed-loop optimization data.
[0073] The present invention performs degradation state analysis based on battery decay curve data, which can deeply understand the degradation conditions and trends of the battery, thereby forming battery degradation state data. The action space is defined based on the battery degradation state data to determine the operating space of the battery's behavior range under different degradation states. Based on the battery degradation action space data, the battery decay curve data is used to predict the battery life through a preset deep deterministic policy gradient, and the battery life can be accurately predicted using advanced algorithms and models. The battery theoretical life data is calculated by difference calculation based on the battery predicted life data, and dynamic difference tracking is performed, so that the difference between the actual life and the theoretical life can be monitored in time, and the difference can be tracked and analyzed. Three-dimensional attribution analysis is performed based on the cycle life difference data, so that the influencing factors of the life difference can be deeply analyzed, and targeted life optimization measures can be taken to extend the battery life and improve the stability and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0074] Figure 1 A schematic diagram of the steps of a lithium battery life prediction method based on a big data method of the present invention;
[0075] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;
[0076] Figure 3 for Figure 1 Detailed implementation steps of step S5 are shown in the flowchart.
[0077] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0078] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.
[0079] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.
[0080] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.
[0081] To achieve this, please refer to Figures 1 to 3 The present invention provides a lithium battery life prediction method based on a big data method, comprising the following steps:
[0082] Step S1: collecting multi-source test data of a lithium battery electric vehicle to obtain multi-source test characteristic data of the lithium battery; processing the battery transient response frequency according to the multi-source test characteristic data of the lithium battery to generate transient response frequency data;
[0083] Step S2: Monitor the battery status of the lithium battery pack to obtain battery status data; perform intelligent behavior detection based on the battery status data to generate battery behavior event data; perform optimal battery sampling rate processing on the transient response frequency data through the battery behavior event data, and perform adaptive real-time data acquisition to generate real-time lithium battery pack data;
[0084] Step S3: Process the lithium battery topology structure according to the real-time lithium battery pack data to generate battery assembly topology structure data; obtain real-time user operation behavior data and real-time environmental data; construct a time series matrix of the real-time lithium battery pack data, real-time user operation behavior data and real-time environmental data to obtain lithium battery telemetry matrix data;
[0085] Step S4: Use the preset graph neural network model to perform transfer learning on the battery connection topology data and build a single component behavior prediction model; use the single component behavior prediction model to predict the battery health trajectory of the lithium battery telemetry matrix data and perform battery attenuation curve processing to obtain battery attenuation curve data;
[0086] Step S5: Predict the battery life according to the battery attenuation curve data to generate battery predicted life data; perform closed-loop optimization processing on the battery life according to the battery predicted life data to generate battery closed-loop optimization data.
[0087] The present invention collects multi-source test data for lithium battery electric vehicles, which helps to obtain a comprehensive data set covering the performance of batteries under various working conditions. Such a data set can provide behavioral characteristics of the battery under different conditions, and comprehensively reflect the behavior of the battery under different working conditions, including multiple parameters such as charge and discharge cycles, temperature changes, current intensity, and voltage response. Subsequently, the obtained multi-source test characteristic data is used to process the battery transient response frequency. By analyzing the battery's ability to respond to rapid load changes, it is helpful to understand the dynamic performance of the battery and evaluate its ability to respond quickly to external changes. By monitoring the battery status of the lithium battery pack and obtaining battery status data, key parameters such as the battery's charging state, voltage, current, and temperature can be tracked in real time. Intelligent behavior detection is performed based on the battery status data, and advanced data analysis technology is used to identify and record abnormal behaviors of the battery in actual use, such as overcharge, overdischarge, or temperature abnormalities. The transient response frequency data is processed at the optimal battery sampling rate through the battery behavior event data, and the data acquisition frequency is optimized. The adaptive real-time data acquisition intelligently adjusts the sampling rate according to the battery's behavior and state, which not only reduces the burden of data storage and processing, but also improves the response speed and accuracy of the monitoring system. The topological structure of lithium batteries is processed according to the real-time lithium battery pack data to determine the connection relationship between battery cells and their configuration. Acquiring real-time user operation behavior data and real-time environmental data not only reflects the use status of the battery in actual operation, but also includes the impact of environmental factors such as temperature and humidity on battery performance. The construction of the time series matrix enables the data to be stored and analyzed in a unified format, which is convenient for efficient data processing and in-depth data analysis. The graph neural network model effectively learns the complex mutual influence and dependency between battery components by analyzing the connection structure between the battery cells, so that the single component behavior prediction model can accurately predict the behavior and state of each battery cell in future operation. The battery health trajectory prediction shows the changing trend of the health state of the battery during future use by analyzing the prediction data. Predicting the battery life according to the battery degradation curve data can generate accurate and reliable battery life prediction data, while considering various factors such as operating conditions, usage patterns and environmental factors, optimizing the battery charging and discharging strategy for real-time performance, reducing energy loss, and improving the overall system efficiency, and minimizing the risk of battery failure while maximizing the overall battery life. Therefore, the lithium battery life prediction method based on big data method of the present invention collects electric vehicle experimental data, intelligently formulates data sampling strategy, and uses battery behavior event detection to realize adaptive real-time collection of lithium battery status data in electric vehicles, thereby improving the efficiency of big data collection and identifying patterns and correlations that are difficult to distinguish from a single data source.The lithium battery topology is processed based on real-time lithium battery pack data, the connection relationship between battery cells is analyzed and determined, and the internal structure of the lithium battery is derived in real time. At the same time, real-time user operation behavior and environmental factors are monitored, and then the lithium battery performance attenuation analysis is performed according to the different usage modes of the lithium battery to accurately predict the life of the lithium battery of electric vehicles.
[0088] In the embodiment of the present invention, reference Figure 1 The above is a schematic diagram of the steps of a lithium battery life prediction method based on a big data method of the present invention. In this embodiment, the lithium battery life prediction method based on a big data method includes the following steps:
[0089] Step S1: collecting multi-source test data of a lithium battery electric vehicle to obtain multi-source test characteristic data of the lithium battery; processing the battery transient response frequency according to the multi-source test characteristic data of the lithium battery to generate transient response frequency data;
[0090] In an embodiment of the present invention, lithium battery test data is collected from multiple sources, including official data provided by manufacturers and reports published by scientific research institutions. Format the data and correct errors. Extract features such as maximum capacity and average discharge rate from basic test data. Extract driving mode and scenario features based on actual driving conditions. Select representative features through PCA dimensionality reduction. Apply the K-means clustering algorithm to determine the appropriate number of clusters. After standardizing the data, perform K-means to classify the data into clusters of different working conditions. Analyze the characteristics and center points of each cluster and define key features that affect battery performance. Calculate the impact of charge and discharge rate and temperature change rate on performance. Analyze the speed and stability of the battery's response under sudden load changes. Use Fourier transform to extract frequency components, especially the main frequency response.
[0091] Step S2: Monitor the battery status of the lithium battery pack to obtain battery status data; perform intelligent behavior detection based on the battery status data to generate battery behavior event data; perform optimal battery sampling rate processing on the transient response frequency data through the battery behavior event data, and perform adaptive real-time data acquisition to generate real-time lithium battery pack data;
[0092] In an embodiment of the present invention, monitoring parameters are set in the automobile BMS to obtain the voltage, current, temperature and remaining capacity data of the battery pack in real time. The data is analyzed using a clustering algorithm to classify the battery status into normal, overloaded, high temperature and other categories. A battery finite state machine model is established to clarify the transition rules between states. Real-time data is input into the model to detect state changes based on the rules. The transient response frequency is adjusted, and the optimal sampling rate is calculated using the Nyquist sampling theorem to ensure that state changes are captured. The sampling frequency is adjusted according to the calculation results to achieve adaptive real-time data acquisition and generate real-time lithium battery pack data.
[0093] Step S3: Process the lithium battery topology structure according to the real-time lithium battery pack data to generate battery assembly topology structure data; obtain real-time user operation behavior data and real-time environmental data; construct a time series matrix of the real-time lithium battery pack data, real-time user operation behavior data and real-time environmental data to obtain lithium battery telemetry matrix data;
[0094] In an embodiment of the present invention, real-time lithium battery pack data is obtained from the automobile BMS, while user operation behavior and environmental data are monitored. These data are correlated and processed to form a time series matrix, which includes the lithium battery status, user behavior and environmental factors. An outlier detection algorithm is used to identify and remove outliers to ensure data accuracy. Feature extraction is performed on the standardized data, such as battery pack voltage, user acceleration, ambient temperature, etc. Topological structure processing identifies the connection and arrangement of battery components. A time series matrix is constructed and event types are marked to obtain a matrix describing the lithium battery telemetry data.
[0095] Step S4: Use the preset graph neural network model to perform transfer learning on the battery connection topology data and build a single component behavior prediction model; use the single component behavior prediction model to predict the battery health trajectory of the lithium battery telemetry matrix data and perform battery attenuation curve processing to obtain battery attenuation curve data;
[0096] In an embodiment of the present invention, a pre-trained graph neural network model is used to perform transfer learning on the battery connection topology data to establish a single component behavior prediction model. Through pattern propagation, the mined frequent pattern information is integrated into the model to optimize its prediction ability. The lithium battery telemetry matrix data is input for prediction to obtain the behavior prediction data of the single component. Based on this data, combined with the working characteristics and attenuation laws of the battery components, the health status trajectory of the battery is predicted. Generate battery health trajectory data to describe the health change trend of the battery in the future. Finally, draw the battery attenuation curve to reflect its performance attenuation during use, such as capacity attenuation, internal resistance increase, etc.
[0097] Step S5: Predict the battery life according to the battery attenuation curve data to generate battery predicted life data; perform closed-loop optimization processing on the battery life according to the battery predicted life data to generate battery closed-loop optimization data.
[0098] In an embodiment of the present invention, the attenuation curve data is analyzed, and the capacity attenuation rate at each time point is calculated to determine the degradation state of the battery. A gradually increasing rate indicates accelerated degradation, and a stable rate indicates normal degradation. Generate a corresponding digital label according to the degradation state, such as 1 for normal, 2 for accelerated, and 3 for severe degradation. Operations defined based on state data, such as adjusting the number of charges or current. The optimal operating strategy is learned using a deep deterministic policy gradient method and applied to battery attenuation curve data to predict life. Compare the predicted life with the theoretical life data, calculate the cycle life difference, and dynamically track life changes. Analyze the impact of each factor on the difference, and establish a statistical model to determine the main influencing factors. Strategies include improving charging management, optimizing driving behavior, etc. By optimizing strategies, the difference is reduced, the battery life is extended, and the performance stability is improved.
[0099] Preferably, step S1 comprises the following steps:
[0100] Step S11: Collect multi-source test data for lithium battery electric vehicles to generate lithium battery multi-source test data;
[0101] Step S12: performing data feature engineering processing according to the lithium battery multi-source test data to obtain lithium battery multi-source test feature data, wherein the lithium battery multi-source test feature data includes test battery pack state data, test vehicle driving condition data, and test driving scenario data;
[0102] Step S13: performing lithium battery working condition cluster analysis on the test battery pack status data using the test vehicle driving condition data and the test driving scenario data to obtain lithium battery working condition cluster data;
[0103] Step S14: extracting key operating characteristic data according to the lithium battery operating condition clustering data to obtain key operating characteristic data, wherein the key operating characteristic data includes charge and discharge rate characteristic data, battery load temperature gradient data, and battery load response data;
[0104] Step S15: Using the key operating characteristic data to process the battery transient response frequency, to obtain transient response frequency data.
[0105] In an embodiment of the present invention, the manufacturer of the electric vehicle is contacted to obtain the officially provided lithium battery test data, which generally includes the battery charging and discharging cycle, battery capacity, voltage change, etc. Consult relevant scientific research institutions or industry reports to obtain publicly released lithium battery test data, such as the battery performance test results in the annual report released by the International Electric Vehicle Association. The collected data is uniformly formatted, and all timestamps are unified to UTC standard time. Remove invalid data (such as records with too many null values) and correct incorrect data input (such as time format errors). Extract features from the basic test data of the battery, such as the maximum capacity of the battery, the average discharge rate, and the battery aging index (calculated by comparing the rate of change of the initial capacity with the current capacity). Extract driving mode features according to the actual use conditions of the vehicle (such as driving speed, acceleration, and parking frequency). Analyze the impact of different driving scenarios (such as low-speed driving in urban areas, highway driving, mountain driving, etc.) on battery performance and extract scene features. Use principal component analysis (PCA) to reduce the dimension, reduce the number of features, and extract the most representative features. Apply the K-means clustering algorithm to classify the test data. Select a suitable K value (such as determining the number of clusters by the elbow method). Extract relevant data from the results of feature engineering processing, mainly battery driving condition data and driving scenario data. Use standardized data to perform the K-means algorithm to classify data points into different clusters under different working conditions. Each cluster represents a specific working environment and usage mode, such as high temperature and high load use, low temperature and low speed driving, etc. Analyze the characteristics of each cluster, such as at what temperature the batteries in the cluster usually work and how the battery load changes. Identify the center points of each cluster, which represent the typical working state of the battery under this working condition. Based on the clustering results, define the key features that affect battery performance and life, such as charge and discharge rate, battery load temperature gradient, and battery load response. Calculate the average charge and discharge rate of the battery in each cluster, and analyze the impact of rate changes on battery performance. Measure and calculate the temperature change rate of the battery under different loads, which can be obtained through the data of the temperature sensor. Analyze the voltage and current response of the battery under sudden load changes (such as rapid acceleration) to evaluate the battery's response speed and stability. The transient response frequency refers to the rate of change and stability of the output power of the battery when it undergoes an instantaneous load change. This usually involves the analysis of rapid changes in the battery's voltage and current. Record the voltage and current response data of the battery under a specific load change (such as 100% to 50% load). Use Fourier transform to analyze this data and extract the frequency components, especially the main frequency response.
[0106] Preferably, step S15 comprises the following steps:
[0107] Step S151: performing spectrum conversion processing on the charge and discharge rate characteristic data and the battery load response data to obtain charge and discharge rate spectrum data and battery load response spectrum data respectively;
[0108] Step S152: performing frequency component analysis on the battery load response spectrum data to generate load response frequency component data;
[0109] Step S153: performing frequency domain step calculation using the charge and discharge rate spectrum data to generate frequency domain step data; performing linear ramp processing according to the frequency domain step data to obtain charge and discharge linear ramp data;
[0110] Step S154: performing transient behavior frequency division on the load response frequency component data by using the charge and discharge linear ramp data, and obtaining quasi-low frequency behavior range data, quasi-medium frequency behavior range data, and quasi-high frequency behavior range data respectively;
[0111] Step S155: performing overshoot and undershoot calculations on the charge and discharge rate spectrum data based on the pseudo low-frequency behavior range data, and adjusting the low-frequency range data to generate low-frequency response range data;
[0112] Step S156: performing sinusoidal wave amplitude response calculation on the charge and discharge rate spectrum data based on the pseudo intermediate frequency behavior range data, and adjusting the intermediate frequency range data to generate intermediate frequency response range data;
[0113] Step S157: Perform nonlinear thermal effect processing using the battery load temperature gradient data to generate nonlinear thermal effect data; perform thermal runaway point risk calculation based on the nonlinear thermal effect data to obtain thermal runaway point risk data; perform high-frequency response range correction on the quasi-high-frequency behavior range data using the thermal runaway point risk data to generate high-frequency response range data.
[0114] In an embodiment of the present invention, Fourier transform (FFT) is used to perform spectrum conversion on the charge and discharge rate data and the battery load response data, and the time series data is converted into frequency domain data to reveal the frequency components in the data. FFT is applied to the charge and discharge rate data to generate charge and discharge rate spectrum data, which will show the intensity of the charge and discharge behavior at different frequencies. FFT is applied to the battery load response data to generate battery load response spectrum data, which will help identify the response frequency of the battery to instantaneous load changes. Detailed frequency component analysis is performed on the battery load response spectrum data using spectrum analysis tools (such as MATLAB or Python's SciPy library). The main frequency components and their corresponding amplitudes are identified, and these components indicate the main frequency region of the battery response. Step calculation is applied to the charge and discharge rate spectrum data. Step calculation analyzes the change trend of the spectrum data by gradually increasing the frequency points, and generates frequency domain step data, including the frequency points and corresponding amplitudes of each step. Linear ramp processing is applied to the frequency domain step data, and a linear ramp is formed by connecting the amplitudes of each frequency point to simulate the smooth transition of the charge and discharge behavior. According to the charge and discharge linear ramp data, the load response frequency component data is divided into three parts: quasi-low frequency behavior range, quasi-medium frequency behavior range and quasi-high frequency behavior range. This division is based on the physical meaning of frequency and its impact on battery performance. Overshoot and undershoot calculations are performed on the quasi-low frequency behavior range data. Overshoot refers to the phenomenon that the battery output power exceeds the predetermined value after the initial load is applied, and undershoot refers to the situation where the output power is lower than the predetermined value. According to the results of overshoot and undershoot, the low-frequency part of the charge and discharge rate spectrum data is adjusted to reflect the performance of the battery in actual use. The sine wave amplitude response calculation is performed on the quasi-medium frequency behavior range data to analyze the response strength and stability of the battery under medium frequency load changes. The sine wave response refers to the power output performance of the battery when it receives a load change shaped like a sine wave. According to the results of the sine wave amplitude response, the medium frequency part of the charge and discharge rate spectrum data is adjusted to ensure that the data reflects the performance of the battery under common usage frequencies. The battery load temperature gradient data is used to process nonlinear thermal effects to analyze the rapid changes in battery temperature under high load and high frequency operation and its impact on battery performance. Nonlinear thermal effect refers to the nonlinear increase in battery temperature when it is subjected to high-frequency load changes, which leads to a decrease in battery performance or an increase in safety risks. The thermal runaway point risk calculation is performed based on the nonlinear thermal effect data to determine the dangerous temperature point reached by the battery under extreme conditions. The thermal runaway point refers to the runaway reaction inside the battery, causing the temperature to rise sharply to a level that causes a safety accident. The thermal runaway point risk data is used to correct the quasi-high-frequency behavior range data to ensure that the high-frequency data reflects the safety limits and performance boundaries of the battery during high-speed operation, and finally obtain the main frequency response data.
[0115] Preferably, step S2 comprises the following steps:
[0116] Step S21: monitoring the battery status of the lithium battery pack through the automobile BMS battery management system to obtain battery status data, wherein the battery status data includes battery pack voltage data, battery pack current data, battery pack temperature data and battery remaining capacity data;
[0117] Step S22: defining battery status events according to the lithium battery working condition clustering data, and generating battery status event data;
[0118] Step S23: constructing a battery finite state machine model based on the battery status event data;
[0119] Step S24: using the battery finite state machine model to perform behavior state transition detection on the battery state data to generate battery behavior event data;
[0120] Step S25: Adaptively adjust the instantaneous response frequency data according to the battery behavior event data, and perform Nyquist sampling calculation to generate optimal battery sampling rate data;
[0121] Step S26: Based on the optimal battery sampling rate data, adaptive real-time data collection is performed on the lithium battery pack to generate real-time lithium battery pack data.
[0122] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S2 in the embodiment are shown in the flowchart. In this embodiment, step S2 includes:
[0123] Step S21: monitoring the battery status of the lithium battery pack through the automobile BMS battery management system to obtain battery status data, wherein the battery status data includes battery pack voltage data, battery pack current data, battery pack temperature data and battery remaining capacity data;
[0124] In an embodiment of the present invention, corresponding monitoring parameters are set in the automobile BMS battery management system to monitor the status of the lithium battery pack in real time. For example, the system can continuously read the voltage, current, temperature and remaining capacity data of the battery pack, and sample at a certain frequency, usually in seconds. These data can be directly obtained by sensors or calculated within the battery management system. For example, the voltage can be calculated by the sum of the voltages of the battery cells, the current can be calculated by the difference between the input and output of the battery pack, the temperature can be obtained by a temperature sensor, and the remaining capacity can be calculated by the difference between the total capacity of the battery pack and the used capacity.
[0125] Step S22: defining battery status events according to the lithium battery working condition clustering data, and generating battery status event data;
[0126] In an embodiment of the present invention, a clustering algorithm is used to analyze and classify the acquired battery status data, and the data is divided into different groups or categories, each category representing a specific battery state. For example, the battery status can be divided into different states such as normal operation, overload, over-discharge, high temperature, and low temperature according to data features such as the voltage, current, temperature, and remaining capacity of the battery pack. Then, corresponding events are defined for each state. For example, when the battery pack is in an overload state, it is defined as an overload event; when the temperature of the battery pack is too high, it is defined as a high temperature event, etc. The battery status event data finally generated contains event information under various states.
[0127] Step S23: constructing a battery finite state machine model based on the battery status event data;
[0128] In the embodiment of the present invention, a battery finite state machine model is established based on the battery state event data. In this model, each state event is defined as a state of the state machine, and the transition between states represents the transition process of the battery between different states. For example, the normal working state can be used as the initial state, and then the state transition can be triggered according to different events, such as the overload event triggering the state transition to the overload state, the overtemperature event triggering the state transition to the high temperature state, etc. In this way, a finite state machine model that describes the battery state change is constructed.
[0129] Step S24: using the battery finite state machine model to perform behavior state transition detection on the battery state data to generate battery behavior event data;
[0130] In the embodiment of the present invention, the real-time monitored battery status data is input into the model, and the battery behavior state transition is detected by the state transition rules in the state machine model. For example, if the current battery state is a normal working state, but the battery temperature is detected to rise suddenly and exceeds the preset threshold, then according to the rules of the state machine model, the state transition to the high temperature state will be triggered, and the corresponding battery behavior event data will be generated, indicating that the battery is in a high temperature state.
[0131] Step S25: Adaptively adjust the instantaneous response frequency data according to the battery behavior event data, and perform Nyquist sampling calculation to generate optimal battery sampling rate data;
[0132] In the embodiment of the present invention, the transient response frequency is adjusted. For example, if a change in the battery state is detected and the battery state needs to be monitored more frequently, the sampling rate can be increased to increase the monitoring frequency; conversely, if the battery state is stable, the sampling rate can be reduced to save energy. Then, according to the Nyquist sampling theorem, the optimal battery sampling rate data is calculated to ensure that the sampling frequency can accurately capture the information of the battery state change.
[0133] Step S26: Based on the optimal battery sampling rate data, adaptive real-time data collection is performed on the lithium battery pack to generate real-time lithium battery pack data.
[0134] In the embodiment of the present invention, the sampling frequency of the real-time data acquisition system is adjusted according to the optimal battery sampling rate data to realize adaptive real-time data acquisition of the lithium battery pack. For example, if the calculated optimal sampling rate is 10 times per second, the data acquisition system is set to collect battery status data 10 times per second. In this way, the data of the lithium battery pack can be obtained in real time, and monitoring and control can be performed according to the real-time data.
[0135] Preferably, step S25 comprises the following steps:
[0136] Step S251: extracting battery behavior features according to the battery behavior event data to generate battery behavior feature data;
[0137] Step S252: performing behavior trend analysis on the battery behavior characteristic data to generate behavior trend data;
[0138] Step S253: performing adaptive frequency threshold calculation on the transient response frequency data to generate adaptive frequency threshold data, wherein the adaptive frequency threshold data includes an adaptive low frequency threshold, an adaptive middle frequency threshold and an adaptive high frequency threshold;
[0139] Step S254: using the behavior trend data based on the adaptive filter to identify the battery behavior frequency band of the adaptive frequency threshold data, and performing Nyquist sampling rate calculation to generate the Nyquist sampling rate;
[0140] Step S255: Perform sampling rate smoothing processing according to the Nyquist sampling rate to generate optimal battery sampling rate data.
[0141] In an embodiment of the present invention, by analyzing the battery behavior event data, key information describing the battery behavior characteristics is extracted. For example, for an overload event, the duration of the current exceeding the rated value can be extracted as a feature; for a high temperature event, the duration and frequency of the temperature exceeding the threshold can be extracted as a feature. These characteristic data can be used to describe the working state and behavior mode of the battery. The extracted battery behavior characteristic data is subjected to trend analysis to identify the changing trend of the battery behavior. For example, the duration of the current overload event can be subjected to trend analysis using a time series analysis method to determine whether there is a trend of gradual increase or decrease. The adaptive frequency threshold of the transient response frequency is calculated based on the battery behavior characteristic data and the behavior trend data. For example, the thresholds of the low frequency, medium frequency and high frequency can be dynamically adjusted based on the frequency distribution in the battery behavior characteristic data and the changing trend in the behavior trend data. The generated adaptive frequency threshold data is processed using an adaptive filter to identify the frequency band of the battery behavior. For example, the threshold ranges of different frequency bands can be determined based on the changing trend in the battery behavior trend data, and the transient response frequency data can be processed in segments to identify the battery behavior in different frequency bands. Then, according to the Nyquist sampling theorem, the Nyquist sampling rate of each frequency band is calculated to ensure that the sampling frequency can accurately capture the changes in battery behavior. The sampling rate is smoothed to eliminate the sudden changes and fluctuations in the sampling rate. For example, methods such as sliding average or exponentially weighted moving average can be used to average the continuous sampling rate data to generate the best battery sampling rate data. In this way, the stability and accuracy of the sampling rate can be guaranteed, thereby achieving effective monitoring and control of battery behavior.
[0142] Preferably, step S3 comprises the following steps:
[0143] Step S31: collecting user operation behavior according to real-time lithium battery pack data to generate user operation behavior data; obtaining real-time environmental data through weather API;
[0144] Step S32: performing data association processing on the real-time lithium battery pack data, the user operation behavior data and the real-time environment data to generate original lithium battery telemetry data;
[0145] Step S33: performing local outlier factor detection on the original lithium battery telemetry data, and performing outlier elimination processing to generate standard lithium battery telemetry data;
[0146] Step S34: normalizing the standard lithium battery telemetry data and extracting relevant feature data to generate lithium battery telemetry feature data, wherein the lithium battery telemetry feature data includes lithium battery pack feature data, user operation behavior feature data and real-time environment feature data;
[0147] Step S35: performing lithium battery topology processing based on the lithium battery pack characteristic data to generate battery assembly topology data;
[0148] Step S36: construct a time series matrix for the lithium battery telemetry feature data, and annotate the matrix rows and columns with event labels to obtain lithium battery telemetry matrix data.
[0149] In an embodiment of the present invention, real-time lithium battery pack data, such as voltage, current, temperature, etc., is obtained from the automobile BMS battery management system. At the same time, the user's operating behavior is monitored, such as starting, stopping the vehicle, accelerating, decelerating, etc. These operating behaviors can be collected by the vehicle's control system or on-board sensors. For example, when the user starts the vehicle, the start time and the current state of the vehicle are recorded; when the user accelerates, the amplitude and duration of the acceleration are recorded. Then, real-time environmental data, such as temperature, humidity, air pressure, etc., are obtained by calling the weather API. The real-time lithium battery pack data, user operation behavior data, and real-time environmental data are associated and processed, and they are merged into a data set. For example, these data can be matched in chronological order according to the timestamp to form a complete data sequence. In this way, the original lithium battery telemetry data is obtained, which contains the status information of the lithium battery pack, the user's operating behavior, and the influencing factors of the environment. The original lithium battery telemetry data is analyzed using an outlier detection algorithm to identify the existing outliers. For example, an isolation forest algorithm or an outlier detection algorithm can be used to detect outliers. Then, the detected outliers are eliminated to ensure the accuracy and reliability of the data. The standard lithium battery telemetry data is normalized to unify the scales between different data. Then, feature extraction is performed on the normalized data. For example, the average voltage, current fluctuation degree, temperature change rate and other features of the battery pack can be calculated; the acceleration, braking degree and other features can be extracted for user operation behavior; and for environmental data, the mean, variance and other statistical features of temperature, humidity and air pressure can be extracted. In this way, the lithium battery telemetry feature data containing various features are obtained. According to the lithium battery pack feature data, the battery components are topologically processed. This includes identifying the connection relationship and dependency relationship between battery components, and determining the arrangement and layout of battery components. For example, the battery components can be divided into different modules according to the characteristics of the battery voltage, capacity and so on, and the connection mode and correlation between each module can be determined. In this way, data describing the topological structure of the battery components is obtained. The lithium battery telemetry feature data is constructed according to the time series to form a time series matrix. Each row represents the feature data of a time point, and each column represents a feature. Then, according to the event label, the rows and columns of the matrix are labeled to mark the event type and feature attributes corresponding to each time point. This results in matrix data describing the lithium battery telemetry data.
[0150] Preferably, step S35 includes the following steps:
[0151] Step S351: Acquire lithium battery diagram structure data; construct a battery structure physical model according to the lithium battery diagram structure data;
[0152] Step S352: identifying battery component nodes according to the lithium battery graph structure data, and generating battery component node data;
[0153] Step S353: Calculate the node-related behavior coefficient for the battery assembly node data to generate the node-related behavior coefficient;
[0154] Step S354: Analyze the component connection relationship according to the node-related behavior coefficient, draw a connection relationship diagram for the lithium battery pack, and generate battery connection relationship diagram data;
[0155] Step S355: performing node connection power analysis on the battery connection relationship diagram data through the lithium battery telemetry matrix data based on the battery structure physical model to generate node connection power data;
[0156] Step S356: Use the node connection power data to perform topological structure correction on the battery connection relationship diagram data to generate battery assembly topological structure data.
[0157] In an embodiment of the present invention, lithium battery graph structure data is obtained by means of design drawings or CAD files of lithium battery components, including information such as the layout of battery components, connection methods, and connection relationships between battery cells. Then, based on these structural data, a physical model of the battery structure is constructed. For example, the battery components can be abstracted as nodes, and the connection relationships between battery cells can be abstracted as edges, thereby constructing a graphical physical model. Based on the lithium battery graph structure data, the nodes of the battery components are identified, that is, the position and relevance of each battery component in the entire battery structure are determined. Based on the battery component node data, the correlation behavior coefficients between the nodes are calculated to describe the mutual influence relationship between the nodes. For example, correlation coefficients such as the distance between nodes, connection strength, and current transmission efficiency can be calculated. These coefficients can reflect the physical connection between nodes and the behavioral characteristics between battery components. Based on the node correlation behavior coefficients, the connection relationship of the battery components is analyzed to determine the connection strength and relevance between the components. Then, based on the analysis results, a connection relationship diagram between the battery components is drawn. Using the battery structure physical model, combined with the lithium battery telemetry matrix data, the battery connection relationship diagram data is subjected to node connection dynamic analysis. This includes modeling and analyzing the dynamic characteristics such as current transmission and energy conversion between nodes. For example, the dynamic parameters between nodes, such as voltage drop, power loss, etc., can be calculated based on the current transmission and the connection relationship between battery components. In this way, data describing the node connection dynamics are obtained. According to the node connection dynamics data, the topology of the battery connection relationship diagram data is corrected and optimized. For example, the connection relationship and connection strength between nodes can be adjusted, and the layout and connection method of battery components can be optimized to improve the performance and stability of the battery pack. The corrected topology data reflects a more reasonable connection relationship between battery components.
[0158] Preferably, step S4 comprises the following steps:
[0159] Step S41: using a preset graph neural network model to perform transfer learning on the battery connection topology data to generate an initial single component behavior prediction model;
[0160] Step S42: training the initial single component behavior prediction model using preset battery component behavior sample data, and performing model evaluation and optimization to generate a single component behavior prediction model;
[0161] Step S43: Perform frequent pattern mining on the lithium battery telemetry matrix data to generate component mining pattern data; use the component mining pattern data to perform pattern propagation and model optimization on the single component behavior prediction model to generate an optimized single component behavior prediction model;
[0162] Step S44: transmitting the lithium battery telemetry matrix data to the optimized single component behavior prediction model to perform single component behavior prediction and generate single component behavior prediction data;
[0163] Step S45: performing battery health trajectory prediction according to the single component behavior prediction data to generate battery health trajectory data;
[0164] Step S46: Process the battery decay curve based on the battery health trajectory data to obtain battery decay curve data.
[0165] In an embodiment of the present invention, a preset graph neural network model is prepared, which has been trained and optimized in other fields or similar tasks and has a certain generalization ability. Then, the battery connection topology data is input into the model for transfer learning. During the transfer learning process, the model will be adjusted and optimized according to the characteristics and patterns of the battery connection topology data to adapt to the task requirements of battery component behavior prediction. The initial single component behavior prediction model is trained using the preset battery component behavior sample data. During the training process, the model will optimize parameters according to the characteristics and labels of the sample data to improve the prediction accuracy and generalization ability of the model. Then, the model is evaluated by a model evaluation method (such as cross validation, ROC curve, etc.), and problems with the model are found and tuned. Frequent pattern mining is performed on the lithium battery telemetry matrix data to identify frequent patterns and association rules in the data. Then, the single component behavior prediction model is propagated using these mined pattern data, that is, the mined pattern information is integrated into the model to optimize the prediction ability of the model. For example, the parameters or structure of the model can be adjusted according to the results of frequent pattern mining to make it closer to the actual situation of the data and improve the prediction accuracy and stability. The lithium battery telemetry matrix data is input into the optimized single component behavior prediction model for prediction. The model predicts the behavior of a single component, such as current, temperature, etc., based on the input data characteristics. The prediction result is the single component behavior prediction data, which describes the predicted behavior state of each component at different time points. Using the single component behavior prediction data, combined with the working characteristics of the battery component and the battery attenuation law, the battery health status trajectory is predicted. For example, the health status of the battery, such as capacity decay, internal resistance increase, etc., can be inferred based on the predicted behavior state of the battery component. Based on the prediction results, the battery health trajectory data is generated to describe the health change trend of the battery in the future. Based on the battery health trajectory data, the battery attenuation curve is drawn. The attenuation curve reflects the performance attenuation of the battery during use, such as capacity decay, internal resistance increase, etc.
[0166] Preferably, step S45 includes the following steps:
[0167] Step S451: performing behavior trend analysis on the single component behavior prediction data to generate single component behavior trend data;
[0168] Step S452: Score the performance behavior trend according to the single component behavior trend data to generate performance behavior trend score data;
[0169] Step S453: performing abnormal performance behavior evaluation on the single-group behavior trend score data using a preset abnormal behavior threshold to generate performance abnormal behavior data;
[0170] Step S454: performing battery failure risk causal reasoning processing on the performance abnormal behavior data to generate battery failure risk data;
[0171] Step S455: Based on the battery failure risk data, the performance abnormal behavior data is subjected to time series evolution processing, and Monte Carlo simulation is performed to generate battery health trajectory data.
[0172] In an embodiment of the present invention, the collected single component behavior prediction data is trend analyzed to observe its changing trend and periodicity. For example, the data can be smoothed by using methods such as moving average method or exponential smoothing method, and then the long-term trend and periodic change of the data can be identified by methods such as trend line fitting or sliding window analysis. Using the single component behavior trend data, the behavior trend of the battery component is scored according to a preset scoring rule or indicator system. For example, different weights and scoring criteria can be given considering factors such as the stability of the behavior, the persistence of the trend, and the amplitude of the fluctuation. According to the scoring rules, the performance behavior trend score of each battery component is calculated. These scoring data reflect the degree of goodness of the single component behavior trend. The single component behavior trend scoring data is evaluated for abnormal performance behavior according to a preset abnormal behavior threshold. The behavior score is compared with the threshold, and the part exceeding the threshold is identified as abnormal behavior. Using the performance abnormal behavior data, the causal relationship between abnormal behavior and battery failure is analyzed by causal reasoning method. For example, a Bayesian network, decision tree, etc. can be used to establish an association model between abnormal behavior and failure, and the cause and risk factors of the failure can be inferred. Based on the battery failure risk data and time series evolution results, Monte Carlo simulation is performed to simulate different failure scenarios and development paths. Through multiple simulations and random sampling, the battery health change trajectory and risk trend in the future are described.
[0173] Preferably, step S5 comprises the following steps:
[0174] Step S51: performing degradation state analysis according to battery attenuation curve data to generate battery degradation state data;
[0175] Step S52: defining the action space according to the battery degradation state data to generate battery degradation action space data;
[0176] Step S53: Based on the battery degradation action space data, the battery attenuation curve data is predicted by a preset deep deterministic policy gradient to generate battery predicted life data;
[0177] Step S54: Obtaining battery theoretical life data through the vehicle BMS battery management system;
[0178] Step S55: performing difference calculation on the battery theoretical life data through the battery predicted life data, and performing dynamic difference tracking to generate cycle life difference data;
[0179] Step S56: Perform three-dimensional attribution analysis based on the cycle life difference data, and perform battery life closed-loop optimization processing to generate battery life closed-loop optimization data.
[0180] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S5 in the flowchart, in this example, step S5 includes:
[0181] Step S51: performing degradation state analysis according to battery attenuation curve data to generate battery degradation state data;
[0182] In an embodiment of the present invention, the collected attenuation curve data is analyzed. For example, the capacity decay rate at each time point is calculated, and then the degradation state of the battery is judged according to the changing trend of the decay rate. If the decay rate gradually increases, it means that the battery is in an accelerated degradation stage; if the decay rate remains stable, it means that the battery is in a normal degradation stage. Based on the results of the degradation state analysis, the battery degradation state data is generated. For example, the degradation state of the battery can be represented by numbers or labels, such as 1 for normal state, 2 for accelerated degradation state, 3 for severe degradation state, etc.
[0183] Step S52: defining the action space according to the battery degradation state data to generate battery degradation action space data;
[0184] In the embodiment of the present invention, different actions or operations taken by the battery are defined based on the battery degradation state data. For example, the action may be defined as increasing the number of charging times, adjusting the charging current, etc. According to the defined action space, various operations that can be taken in different degradation states are listed, and the parameters or characteristics corresponding to each operation are recorded.
[0185] Step S53: Based on the battery degradation action space data, the battery attenuation curve data is predicted by a preset deep deterministic policy gradient to generate battery predicted life data;
[0186] In an embodiment of the present invention, a deep deterministic policy gradient method is used to learn the optimal operation strategy of the battery based on the degradation action space data of the battery. For example, by training a neural network model, it can select the optimal operation action according to the current state and degradation degree of the battery. The learned optimal operation strategy is applied to the battery attenuation curve data to predict the battery life. By simulating different operation scenarios and degradation states, the life prediction results of the battery under different operation strategies can be obtained.
[0187] Step S54: Obtaining battery theoretical life data through the vehicle BMS battery management system;
[0188] In the embodiment of the present invention, the automobile BMS battery management system is used to obtain the theoretical life data of the battery from the automobile battery. The data includes information such as the battery's design life, production date, and life recommended by the manufacturer.
[0189] Step S55: performing difference calculation on the battery theoretical life data through the battery predicted life data, and performing dynamic difference tracking to generate cycle life difference data;
[0190] In an embodiment of the present invention, the battery predicted life data is compared with the battery theoretical life data to calculate the difference between the two, which can be a simple numerical subtraction. For example, the predicted life of the battery at a certain point in time is obtained to be 950 charging cycles through the prediction model. The difference between the predicted life and the theoretical life is calculated: 1000-950=50 charging cycles. As time goes by, the battery predicted life data is continuously updated and compared with the battery theoretical life data, and the cycle life difference data is continuously updated. In this way, dynamic tracking and monitoring of battery life changes can be achieved.
[0191] Step S56: Perform three-dimensional attribution analysis based on the cycle life difference data, and perform battery life closed-loop optimization processing to generate battery life closed-loop optimization data.
[0192] In an embodiment of the present invention, the change in battery life can be analyzed from multiple dimensions, such as the use environment, charging method, driving habits, etc. These dimensions are used as the basis for attribution analysis to determine the main factors that cause the change in cycle life difference. These data are associated with the cycle life difference data to analyze the relationship between each factor and the cycle life difference. Based on the collected data, a multivariate regression model or other suitable statistical model is established to analyze the relationship between the cycle life difference and each attribution dimension. In the model, the cycle life difference is used as the dependent variable, and each attribution dimension is used as an independent variable. The degree of influence of each factor on the cycle life difference is determined by regression analysis. Using the established model, each attribution dimension is analyzed to find out the factor that has the greatest impact on the change in cycle life difference. Using the established model, each attribution dimension is analyzed to find out the factor that has the greatest impact on the change in cycle life difference. These strategies include improving charging management strategies, optimizing driving behavior guidance, and improving environmental adaptability. By implementing optimization strategies, it is expected to reduce the cycle life difference, extend the service life of the battery, and improve the stability of battery performance.
[0193] The beneficial effect of the present application is that multi-source test data collection is carried out for lithium battery electric vehicles, covering parameters such as charge and discharge cycles, temperature changes, and current intensity to obtain comprehensive performance data. Subsequently, these data are used to process the battery transient response frequency and analyze its ability to respond to rapid load changes to evaluate its dynamic performance. By real-time monitoring of the battery status, including parameters such as charging status, voltage, current, and temperature, and using intelligent behavior detection technology to identify abnormal behaviors, such as overcharging, overdischarging, or temperature abnormalities. Optimize the data acquisition frequency, intelligently adjust the sampling rate according to the battery behavior and status, and improve the response speed and accuracy of the monitoring system. At the same time, the connection relationship between battery cells is determined by the topological structure processing of the lithium battery pack, and real-time user operation behavior and environmental data are obtained to reflect the actual use and the impact of environmental factors on battery performance. Construct a time series matrix to uniformly store and analyze data, use the graph neural network model to analyze the connection structure between cells, and predict the behavior and state of the battery cell in the future. The battery health trajectory prediction shows the changing trend of the health status during future use, predicts the battery life through the degradation curve data, and generates accurate life prediction data. Optimize charging and discharging strategies, reduce energy loss, lower the risk of failure, extend battery life, and improve overall system efficiency.
[0194] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.
[0195] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.
Claims
1. A lithium battery life prediction method based on big data method, characterized in that: The following steps are involved: Step S1: Collect multi-source test data of lithium battery electric vehicles to obtain multi-source test feature data of lithium batteries; Performing battery transient response frequency processing according to lithium battery multi-source test characteristic data to generate transient response frequency data; step S1 includes the following steps: Step S11: Collect multi-source test data for lithium battery electric vehicles to generate lithium battery multi-source test data; Step S12: performing data feature engineering processing according to the lithium battery multi-source test data to obtain lithium battery multi-source test feature data, wherein the lithium battery multi-source test feature data includes test battery pack state data, test vehicle driving condition data, and test driving scenario data; Step S13: performing lithium battery working condition cluster analysis on the test battery pack status data using the test vehicle driving condition data and the test driving scenario data to obtain lithium battery working condition cluster data; Step S14: extracting key operating characteristic data according to the lithium battery operating condition clustering data to obtain key operating characteristic data, wherein the key operating characteristic data includes charge and discharge rate characteristic data, battery load temperature gradient data, and battery load response data; Step S15: Processing the instantaneous response frequency of the battery using the key working characteristic data to obtain instantaneous response frequency data; wherein the instantaneous response frequency data includes low frequency response range data, medium frequency response range data and high frequency response range data, and step S15 includes the following steps: Step S151: performing spectrum conversion processing on the charge and discharge rate characteristic data and the battery load response data to obtain charge and discharge rate spectrum data and battery load response spectrum data respectively; Step S152: performing frequency component analysis on the battery load response spectrum data to generate load response frequency component data; Step S153: performing frequency domain step calculation using the charge and discharge rate spectrum data to generate frequency domain step data; performing linear ramp processing according to the frequency domain step data to obtain charge and discharge linear ramp data; Step S154: performing transient behavior frequency division on the load response frequency component data by using the charge and discharge linear ramp data, and obtaining quasi-low frequency behavior range data, quasi-medium frequency behavior range data, and quasi-high frequency behavior range data respectively; Step S155: performing overshoot and undershoot calculations on the charge and discharge rate spectrum data based on the pseudo low-frequency behavior range data, and adjusting the low-frequency range data to generate low-frequency response range data; Step S156: performing sinusoidal wave amplitude response calculation on the charge and discharge rate spectrum data based on the pseudo intermediate frequency behavior range data, and adjusting the intermediate frequency range data to generate intermediate frequency response range data; Step S157: performing nonlinear thermal effect processing using the battery load temperature gradient data to generate nonlinear thermal effect data; performing thermal runaway point risk calculation based on the nonlinear thermal effect data to obtain thermal runaway point risk data; performing high-frequency response range correction on the quasi-high-frequency behavior range data using the thermal runaway point risk data to generate high-frequency response range data; Step S2: monitor the battery status of the lithium battery pack to obtain battery status data; perform intelligent behavior detection based on the battery status data to generate battery behavior event data; perform optimal battery sampling rate processing on the transient response frequency data through the battery behavior event data, and perform adaptive real-time data acquisition to generate real-time lithium battery pack data; Step S2 includes the following steps: Step S21: monitoring the battery status of the lithium battery pack through the automobile BMS battery management system to obtain battery status data, wherein the battery status data includes battery pack voltage data, battery pack current data, battery pack temperature data and battery remaining capacity data; Step S22: defining battery status events according to the lithium battery working condition clustering data, and generating battery status event data; Step S23: constructing a battery finite state machine model based on the battery status event data; Step S24: using the battery finite state machine model to perform behavior state transition detection on the battery state data to generate battery behavior event data; Step S25: Adaptively adjust the instantaneous response frequency data according to the battery behavior event data, and perform Nyquist sampling calculation to generate optimal battery sampling rate data; Step S25 includes the following steps: Step S251: extracting battery behavior features according to the battery behavior event data to generate battery behavior feature data; Step S252: performing behavior trend analysis on the battery behavior characteristic data to generate behavior trend data; Step S253: performing adaptive frequency threshold calculation on the transient response frequency data to generate adaptive frequency threshold data, wherein the adaptive frequency threshold data includes an adaptive low frequency threshold, an adaptive middle frequency threshold and an adaptive high frequency threshold; Step S254: using the behavior trend data based on the adaptive filter to identify the battery behavior frequency band of the adaptive frequency threshold data, and performing Nyquist sampling rate calculation to generate the Nyquist sampling rate; Step S255: performing sampling rate smoothing processing according to the Nyquist sampling rate to generate optimal battery sampling rate data; Step S26: performing adaptive real-time data collection on the lithium battery pack based on the optimal battery sampling rate data to generate real-time lithium battery pack data; Step S3: Process the lithium battery topology structure according to the real-time lithium battery group data to generate battery assembly topology structure data; obtain real-time user operation behavior data and real-time environmental data; construct a time series matrix of the real-time lithium battery group data, real-time user operation behavior data and real-time environmental data to obtain lithium battery telemetry matrix data; Step S3 includes the following steps: Step S31: collecting user operation behavior according to real-time lithium battery pack data to generate user operation behavior data; obtaining real-time environmental data through weather API; Step S32: performing data association processing on the real-time lithium battery pack data, the user operation behavior data and the real-time environment data to generate original lithium battery telemetry data; Step S33: performing local outlier factor detection on the original lithium battery telemetry data, and performing outlier elimination processing to generate standard lithium battery telemetry data; Step S34: normalizing the standard lithium battery telemetry data and extracting relevant feature data to generate lithium battery telemetry feature data, wherein the lithium battery telemetry feature data includes lithium battery pack feature data, user operation behavior feature data and real-time environment feature data; Step S35: Processing the lithium battery topology structure through the lithium battery pack characteristic data to generate battery assembly topology structure data; Step S35 includes the following steps: Step S351: Acquire lithium battery diagram structure data; construct a battery structure physical model according to the lithium battery diagram structure data; Step S352: identifying battery component nodes according to the lithium battery graph structure data, and generating battery component node data; Step S353: Calculate the node-related behavior coefficient for the battery assembly node data to generate the node-related behavior coefficient; Step S354: Analyze the component connection relationship according to the node-related behavior coefficient, draw a connection relationship diagram for the lithium battery pack, and generate battery connection relationship diagram data; Step S355: performing node connection power analysis on the battery connection relationship diagram data through the lithium battery telemetry matrix data based on the battery structure physical model to generate node connection power data; Step S356: using the node connection power data to perform topological structure correction on the battery connection relationship diagram data to generate battery assembly topological structure data; Step S36: constructing a time series matrix for the lithium battery telemetry feature data, and marking the matrix rows and columns with event labels to obtain lithium battery telemetry matrix data; Step S4: Use the preset graph neural network model to perform transfer learning on the battery connection topology data and build a single component behavior prediction model; use the single component behavior prediction model to predict the battery health trajectory of the lithium battery telemetry matrix data and perform battery attenuation curve processing to obtain battery attenuation curve data; Step S5: Predict the battery life according to the battery attenuation curve data to generate battery predicted life data; perform closed-loop optimization processing on the battery life according to the battery predicted life data to generate battery closed-loop optimization data.
2. The lithium battery life prediction method based on big data method according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: using a preset graph neural network model to perform transfer learning on the battery connection topology data to generate an initial single component behavior prediction model; Step S42: training the initial single component behavior prediction model using preset battery component behavior sample data, and performing model evaluation and optimization to generate a single component behavior prediction model; Step S43: Perform frequent pattern mining on the lithium battery telemetry matrix data to generate component mining pattern data; use the component mining pattern data to perform pattern propagation and model optimization on the single component behavior prediction model to generate an optimized single component behavior prediction model; Step S44: transmitting the lithium battery telemetry matrix data to the optimized single component behavior prediction model to perform single component behavior prediction and generate single component behavior prediction data; Step S45: performing battery health trajectory prediction according to the single component behavior prediction data to generate battery health trajectory data; Step S46: Process the battery decay curve based on the battery health trajectory data to obtain battery decay curve data.
3. The lithium battery life prediction method based on big data method according to claim 2 is characterized in that: Step S45 includes the following steps: Step S451: performing behavior trend analysis on the single component behavior prediction data to generate single component behavior trend data; Step S452: Score the performance behavior trend according to the single component behavior trend data to generate performance behavior trend score data; Step S453: performing abnormal performance behavior evaluation on the single-group behavior trend score data using a preset abnormal behavior threshold to generate performance abnormal behavior data; Step S454: performing battery failure risk causal reasoning processing on the performance abnormal behavior data to generate battery failure risk data; Step S455: Based on the battery failure risk data, the performance abnormal behavior data is subjected to time series evolution processing, and Monte Carlo simulation is performed to generate battery health trajectory data.
4. The lithium battery life prediction method based on big data method according to claim 3 is characterized in that: Step S5 includes the following steps: Step S51: performing degradation state analysis according to battery attenuation curve data to generate battery degradation state data; Step S52: defining the action space according to the battery degradation state data to generate battery degradation action space data; Step S53: Based on the battery degradation action space data, the battery attenuation curve data is predicted by a preset deep deterministic policy gradient to generate battery predicted life data; Step S54: obtaining battery theoretical life data through the vehicle BMS battery management system; Step S55: performing difference calculation on the battery theoretical life data through the battery predicted life data, and performing dynamic difference tracking to generate cycle life difference data; Step S56: Perform three-dimensional attribution analysis based on the cycle life difference data, and perform battery life closed-loop optimization processing to generate battery life closed-loop optimization data.
Citation Information
Patent Citations
Uninterrupted power supply system and method with mixture of fuel cell and super capacitor
CN108683247A
Rechargeable battery health state prediction method
CN116047314A
