A method, system, product, medium, and equipment for early warning of outlier events in lithium-ion batteries.
By constructing a comprehensive analysis model for multimodal data, and combining local outlier detection, residual analysis, and deep learning, the problems of accuracy and early warning in detecting outliers in lithium-ion batteries are solved, thereby improving the reliability and safety of the battery management system.
Patent Information
- Application Number
- CN202511019167.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-23
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2045-07-23
Smart Images

Figure CN120742119B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of battery fault prediction technology, and in particular to a method, system, product, medium, and device for early warning of outliers in lithium-ion batteries. Background Technology
[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.
[0003] In many fields such as new energy vehicles, energy storage systems, and consumer electronics, lithium-ion batteries serve as a core energy carrier, and their performance and safety are of paramount importance. However, during long-term use, lithium-ion batteries often exhibit outlier phenomena, where the parameters of individual cells or modules deviate significantly from the normal range, manifesting as capacity decay, increased internal resistance, lithium plating, thermal runaway, and other problems. These issues not only reduce the overall performance of the equipment but may also lead to safety accidents, posing a serious threat to the life and property safety of users.
[0004] Currently, while numerous studies have been conducted on methods for detecting and analyzing outlier phenomena in lithium-ion batteries, some shortcomings remain. Some methods focus on monitoring a single parameter, such as relying solely on changes in voltage or internal resistance to determine if a cell is outlier. However, outlier phenomena in batteries are often the result of multiple factors working together, and a single parameter cannot comprehensively and accurately reflect the true state of the cell, easily leading to misjudgments or omissions. For example, judging solely by voltage changes may overlook other potential problems such as increased internal resistance, causing some cells in the early stages of failure to be identified in a timely manner.
[0005] While existing technologies have attempted multi-parameter comprehensive analysis, most have failed to fully integrate multimodal data. For example, although some methods consider multiple parameters such as voltage and internal resistance simultaneously, they have failed to achieve synchronous acquisition and deep coupling analysis of different parameter data during data acquisition and analysis. This makes it difficult to accurately capture the characteristics of outlier cells when faced with complex battery state changes, reducing the accuracy and reliability of outlier detection.
[0006] Furthermore, current single detection methods have some limitations. For example, some experimental methods cause significant damage to the battery during the detection process, while methods that rely heavily on mathematical models depend excessively on model accuracy and training data, resulting in inconsistent detection efficiency. In addition, these methods often fail to provide early warnings of battery outlier phenomena, and cannot provide timely and accurate information to the battery management system to take appropriate measures to prevent safety accidents.
[0007] In summary, existing technologies for detecting and analyzing outlier phenomena in lithium-ion batteries suffer from limitations such as single detection dimensions, simplistic methods, and difficulty in providing early warnings. These issues restrict the effective identification and management of lithium-ion battery outlier phenomena, necessitating a new method that can comprehensively, accurately, and efficiently detect battery outliers to improve the safety and reliability of battery systems, ensure the stable operation of related equipment, and guarantee user safety. Summary of the Invention
[0008] To address the shortcomings of existing technologies, the present invention aims to provide a method, system, product, medium, and device for early warning of lithium-ion battery outliers. This method calculates the battery's lifespan degradation based on the capacity percentage during the constant voltage phase, and fully leverages the advantages of both practical experiments and model learning. Through comprehensive analysis of multimodal data, it achieves accurate identification and early warning of lithium-ion battery outlier phenomena.
[0009] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides a method for early warning of outlier activity in lithium-ion batteries, comprising the following steps: The electrical performance of lithium-ion batteries was tested to obtain the DC internal resistance, capacity percentage in the constant voltage stage, and capacity percentage in the constant current stage at different temperatures and with different capacities. A lifetime assessment model with three branches is constructed and trained. The three branches include a first branch for local outlier detection, a second branch for residual analysis, and a third branch for deep learning. The three branches in the life assessment model process the DC internal resistance, the capacity ratio of the constant voltage stage, and the capacity ratio of the constant current stage in parallel. The results obtained from the three branches are then fused and used to make a decision, resulting in the outlier detection result.
[0010] Furthermore, the first branch performs local outlier detection by constructing a matrix graph, the second branch performs residual analysis by constructing a fitting formula based on linear decay law and sinusoidal fluctuation, and the third branch captures the life decay law of lithium battery based on SOC DC internal resistance, constant voltage stage capacity ratio, and constant current stage capacity ratio to obtain outlier prediction results for lithium battery.
[0011] Furthermore, the third branch learns the relationship between the DC internal resistance of the SOC, the capacity ratio during the constant voltage stage, the capacity ratio during the constant current stage, and the lifespan degradation of the lithium battery through a multi-head attention mechanism, and uses a correction mechanism to correct the learned lifespan degradation pattern.
[0012] Furthermore, the specific steps for conducting electrical performance tests on lithium-ion batteries are as follows: Set different current, voltage, and temperature parameters; Electrical performance tests were conducted on the lithium-ion batteries to be tested by charging and discharging. By changing the current, voltage, and temperature parameters, repeated electrical performance tests were conducted, yielding different test results.
[0013] Furthermore, the training steps for the life assessment model are as follows: Multiple sets of lithium-ion battery electrical performance test data were obtained to form a dataset. The parameters in the dataset were preprocessed, and the data were labeled and divided into training set and test set. The life assessment model is trained using the training set and tested using the test set to obtain the trained life assessment model.
[0014] Furthermore, the specific steps for preprocessing the electrical performance parameters are as follows: Clean the electrical performance parameters to remove noise data and outliers; Electrical performance parameters with different dimensions and ranges are normalized to ensure they fall within the same numerical range.
[0015] A second aspect of the present invention provides a lithium-ion battery out-of-population early warning system, comprising: The data acquisition module is configured to perform electrical performance tests on lithium-ion batteries, including DC internal resistance, capacity percentage during constant voltage stage, and capacity percentage during constant current stage at different temperatures and different capacities. The model building module is configured to build a lifetime assessment model containing three branches and train the lifetime assessment model. The three branches include a first branch for local outlier detection, a second branch for residual analysis, and a third branch for deep learning. The outlier prediction module is configured to process the DC internal resistance, constant voltage stage capacity ratio, and constant current stage capacity ratio in parallel as three branches in the lifetime assessment model. The results obtained from the three branches are then fused and used to make a decision to obtain the outlier detection result.
[0016] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute steps in the lithium-ion battery out-of-population warning method as described in the first aspect of the present invention.
[0017] A fourth aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the lithium-ion battery out-of-population warning method as described in the first aspect of the present invention.
[0018] A fifth aspect of the present invention provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the lithium-ion battery out-of-population warning method as described in the first aspect of the present invention.
[0019] The above one or more technical solutions have the following beneficial effects: This application discloses a method, system, product, medium, and device for early warning of outlier phenomena in lithium-ion batteries. By comprehensively analyzing multimodal data using multiple methods, it achieves accurate identification and early warning of outlier phenomena in lithium-ion batteries. The method of this invention not only considers the key parameter of DC internal resistance (DCIR), but also incorporates multi-dimensional data such as the capacity percentage during the constant voltage stage and the capacity percentage during the constant current stage. The variation patterns of these parameters under different temperatures and charge states are comprehensively considered, making outlier detection more accurate.
[0020] This invention enables outlier detection using different outlier prediction methods and integrates all detection results for analysis. This method fully leverages the advantages of both experimental and mathematical model-based detection, offering greater adaptability. This adaptability not only improves the model's generalization ability but also ensures accurate assessment of battery health and lifespan degradation under various complex operating conditions. This invention achieves accurate identification and early warning of outlier phenomena in lithium-ion batteries. The method not only improves the accuracy of outlier detection and enhances the model's adaptability and learning ability but also provides comprehensive battery health assessment and early warning functions, significantly improving the intelligence and reliability of battery management systems. These technical effects make this invention valuable and promising for the management and maintenance of lithium-ion batteries.
[0021] This invention utilizes a multi-head attention mechanism to learn the relationship between SOC DC internal resistance, capacity percentage during constant voltage and constant current phases, and lithium battery lifespan degradation. The multi-head attention mechanism automatically focuses on important features and relationships in the data, improving the model's ability to learn complex data patterns. Furthermore, a correction mechanism is used to correct the learned lifespan degradation patterns, further enhancing the model's accuracy and reliability. This combination of deep learning and correction mechanisms enables the model to more accurately capture battery lifespan degradation patterns, thus achieving more reliable outlier warnings.
[0022] By leveraging the comprehensive analysis of multimodal data and the predictive capabilities of deep learning models, this invention can identify battery cells that may exhibit outlier behavior in advance. This early warning function enables the battery management system to take preventative maintenance measures before problems worsen, thus avoiding safety incidents.
[0023] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0025] Figure 1 This is a flowchart of the lithium-ion battery out-of-group early warning method in Embodiment 1 of the present invention. Detailed Implementation
[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0027] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0028] Example 1: Embodiment 1 of this invention provides a lithium-ion battery out-of-group early warning method. In order to explore the reasons that cause individual cells to become out of the battery pack due to capacity decay, increased internal resistance, lithium plating and other factors in actual use of lithium-ion batteries, an out-of-group analysis method based on multi-field coupling is proposed to predict whether the cells are abnormal in advance, so as to warn of potential safety accidents in the future.
[0029] like Figure 1 As shown, the specific steps include: S1: Conduct electrical performance tests on lithium-ion batteries to obtain DC internal resistance, capacity percentage during constant voltage stage, and capacity percentage during constant current stage at different temperatures and capacities.
[0030] S101: Set different current, voltage, and temperature parameters.
[0031] S102: Electrical performance test of the lithium-ion battery under test by charging and discharging.
[0032] S103: Repeat the electrical performance test experiment by changing the current, voltage and temperature parameters to obtain different test results.
[0033] In one specific implementation, the prepared battery is subjected to electrical performance testing to obtain electrical performance test results information, including the DC internal resistance (DCIR) at different SOCs at different temperatures and the capacity percentage during the constant voltage stage.
[0034] This embodiment tests the DC internal resistance of lithium-ion batteries at different states of charge (SOC) and measures the capacity decay (CV) ratio during cycle life testing. Specifically, four lithium iron phosphate batteries are tested for DC internal resistance (DCIR) at different temperatures and SOCs. Experiment 1: Set the temperature to 25℃.
[0035] 1. Charge at 1C constant current to 3.65V, then charge at 0.05C constant voltage.
[0036] 2. Let stand for 30 minutes.
[0037] 3. 1C constant current discharge for 60 seconds.
[0038] 4. Adjust the capacity to 95% SOC 5. Let stand for 30 minutes.
[0039] 6. 1C constant current discharge for 60 seconds.
[0040] Repeat steps 4 through 6 to obtain DCIR data for 90%, 80%, 50%, 20%, 10%, and 5% SOC, as shown in Table 1. Table 1. Results of battery state of health (SOH) / battery charge (SOC) at 25℃.
[0041]
[0042] Experiment 2: Set the temperature to 45℃.
[0043] 1. Charge at 1C constant current to 3.65V, then charge at 0.05C constant voltage: 2. Let stand for 30 minutes; 3. 1C constant current discharge for 60 seconds; 4. Adjust the capacity to 95% SOC 5. Let stand for 30 minutes; 6. 1C constant current discharge for 60 seconds; Repeat steps 4 through 6 to obtain DCIR data for 90%, 80%, 50%, 20%, 10%, and 5% SOC, as shown in Table 2.
[0044] Table 2. Results of battery state of health (SOH) / battery charge (SOC) at 45℃.
[0045]
[0046] Experiment 3: Set the temperature to -10℃.
[0047] 1. Charge at 1C constant current to 3.65V, then charge at 0.05C constant voltage: 2. Let stand for 30 minutes; 3. 1C constant current discharge for 60 seconds; 4. Adjust the capacity to 95% SOC 5. Let stand for 30 minutes; 6. 1C constant current discharge for 60 seconds; Repeat steps 4 through 6 to obtain DCIR data for 90%, 80%, 50%, 20%, 10%, and 5% SOC, as shown in Table 3.
[0048] Table 3. Results of battery state of health (SOH) / battery charge (SOC) under -10℃ conditions.
[0049]
[0050] The measurement results in this embodiment yielded DC internal resistance values at different temperatures (25℃, 45℃, and -10℃) with varying SOCs. The battery health status percentage corresponding to the DCIR data obtained from existing cycle life tests was used as a reference. Tables 1, 2, and 3 show that the battery's DC internal resistance under actual operating conditions affects its healthy charging and lifespan.
[0051] S2: Construct a lifetime assessment model containing three branches and train the lifetime assessment model.
[0052] In one specific implementation, this embodiment designs a hybrid model H-MFR (Hybrid Matrix-Fitting-Relation Model) for lithium-ion battery outlier feature recognition, which integrates matrix graphs, fitting formulas, and deep learning. The three branches include a first branch for local outlier detection, a second branch for residual analysis, and a third branch for deep learning. The first branch performs local outlier detection by constructing a matrix graph; the second branch performs residual analysis by constructing a fitting formula based on linear decay laws and sinusoidal fluctuations; and the third branch captures the lifespan decay law of the lithium battery based on the SOC DC internal resistance, the capacity percentage during the constant voltage stage, and the capacity percentage during the constant current stage, thus obtaining the outlier prediction result for the lithium battery.
[0053] S201: The first approach involves creating a matrix diagram based on the three temperature measurement results obtained in S1. The constant voltage ratio determines the internal impedance value of the battery. The higher the battery impedance, the greater the temperature rise and the higher the risk factor. If the battery data in the battery pack is not shown in the data matrix diagram or the CV capacity ratio suddenly increases, it can be inferred that the battery is out of order.
[0054] Specifically, in the first branch, the DCIR values of each SOC point at 25 ℃, 45 ℃, and -10 ℃ are used as input to construct a 3×N matrix M∈R^{3×N} (N is the number of SOC points). A 2-D CNN (3×3 convolutional kernel) is used to extract local spatial correlations, and channel attention is used to emphasize the temperature dimension differences. Finally, a 128-dimensional vector is obtained through global average pooling, and the outlier probability is output after passing through a fully connected layer.
[0055] S202: The second branch performs linear fitting on the obtained results and determines whether there are outliers in the battery based on the fitting formula.
[0056] This embodiment constructs a fitting formula based on the linear decay law and sinusoidal fluctuation: .
[0057] The linear part includes: Used to describe the overall downward trend, the fluctuation part: Used to describe periodic fluctuations. 'a' represents the initial y-intercept, 'b' represents the linear decay slope (a negative value indicates a decrease), 'c' represents the fluctuation amplitude, 'd' represents the fluctuation frequency, 'e' represents the phase shift, and 'y' is the prediction result.
[0058] Specifically, the second branch structure is a spatial feature extractor + bidirectional temporal modeler (CNN + Bi-LSTM) structure, which combines the local spatial feature extraction capability of the convolutional neural network (CNN) with the temporal modeling capability of the bidirectional long short-term memory network (Bi-LSTM). Outlier probability is calculated through residual analysis, with the inputs being the number of iterations, the measured SOH value, and the predicted value fitted using the above fitting formula. The residual δ is designed as the difference between the predicted and actual values. The resulting residual sequence δ(x) is input into the 1-D CNN + Bi-LSTM to capture the nonlinearity and temporal drift of the residuals with each iteration; the residual distribution parameters are output, and the negative log-likelihood loss is calculated. The outlier probability is output using the 3σ rule. After training, the fitting formula in this embodiment can be written as: .
[0059] S203: The third branch learns the relationship between the SOC DC internal resistance, the capacity ratio in the constant voltage stage, the capacity ratio in the constant current stage, and the life decay of the lithium battery through a multi-head attention mechanism, and uses a correction mechanism to correct the learned life decay pattern.
[0060] The third branch of the life assessment model comprises a backbone network, a deep learning network, and a classifier. The number of iterations is set, and the inputs are the SOC DC internal resistance, the capacity percentage during the constant voltage stage, and the capacity percentage during the constant current stage at different temperatures. The backbone network extracts different features from the input data and maps features of different dimensions to a unified dimension. The deep learning network, based on a multi-head attention mechanism, learns the relationships between features extracted by the backbone network. The learning results are then classified by a classifier to obtain outlier predictions.
[0061] Specifically, multi-head attention (8 heads) learns cross-feature coupling, and its formula is: .
[0062] In this context, Query, Key, and Value represent Q, K, and V in the attention mechanism, respectively. f() represents the mapping function. SOC-DCIRall represents the DC internal resistance sequence at different SOC points at different temperatures. CV% and CC% represent the capacity percentages during the constant voltage and constant current phases, respectively. A correction mechanism is used to suppress attention noise. Finally, the deep learning network outputs a 256-dimensional relation vector, which, after passing through a classifier, provides the outlier probability.
[0063] The correction mechanism includes estimating noise in the multi-head attention mechanism, using learnable gating for noise filtering, and finally, after residual and normalization operations, outputting a stable 256-dimensional relation vector. The correction mechanism adopts the noise removal method of existing attention mechanisms, which will not be elaborated here.
[0064] S204: The three branches output the outlier probability or feature vector of the same battery sample, and finally give a comprehensive outlier score through the adaptive weight fusion module (AWF).
[0065] S2041: Set the outlier threshold θ and calculate the confidence level of each branch.
[0066] S2042: Softmax normalization yields the confidence weights for different branches.
[0067] S2043: Calculate the overall outlier score: .
[0068] Where Sout is the overall outlier score, w1, w2, and w3 are the confidence weights of the first, second, and third branches, respectively, and p1, p2, and p3 are the outputs of the first, second, and third branches, respectively. An outlier alarm is triggered when Sout ≥ θ.
[0069] S205: Train the H-MFR model using historical data or existing public datasets.
[0070] S2051: Obtain multiple sets of lithium-ion battery electrical performance test data to form a dataset, preprocess the parameters in the dataset, and divide the labeled data into training and testing sets. The specific steps for preprocessing the electrical performance parameters are as follows: Clean the electrical performance parameters to remove noise data and outliers; Electrical performance parameters with different dimensions and ranges are normalized to ensure they fall within the same numerical range.
[0071] S2052: Train the life assessment model using the training set and test the training effect using the test set to obtain the trained life assessment model.
[0072] In this embodiment, the three branches are trained using their respective losses (MSE, NLL, BCE), while the other branches are frozen during training. Based on the different characteristics of the three branches, the cross-entropy between the final Sout and the true label is used as the total loss. Only the top-level parameters of the adaptive weight fusion module and the third branch are updated to prevent catastrophic forgetting. Generalization ability is improved by using unlabeled vehicle-side data and applying consistency regularization to the consistency losses output by the first and second branches.
[0073] S3: The three branches in the life assessment model process the DC internal resistance, the capacity ratio of the constant voltage stage, and the capacity ratio of the constant current stage in parallel. The results obtained from the three branches are fused and decided to obtain the outlier detection result.
[0074] Example 2: Embodiment 2 of the present invention provides a lithium-ion battery out-of-population early warning system, comprising: The data acquisition module is configured to perform electrical performance tests on lithium-ion batteries, including DC internal resistance, capacity percentage during constant voltage stage, and capacity percentage during constant current stage at different temperatures and different capacities. The model selection module is configured to match the appropriate lifetime assessment model based on the electrical performance test conditions. The outlier prediction module is configured to use a selected lifetime assessment model to capture the lifetime degradation law of lithium batteries based on the SOC DC internal resistance, constant voltage stage capacity ratio, and constant current stage capacity ratio, and obtain the outlier prediction results of lithium batteries. Specifically, a multi-head attention mechanism is used to learn the relationship between SOC DC internal resistance, constant voltage stage capacity ratio, constant current stage capacity ratio and lithium battery lifetime degradation, and a correction mechanism is used to correct the learned lifetime degradation law.
[0075] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps of the lithium-ion battery out-of-population warning method as described in Embodiment 1 of the present invention.
[0076] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps of the lithium-ion battery out-of-population warning method as described in Embodiment 1 of the present invention.
[0077] Example 5: Embodiment 5 of the present invention provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in the lithium-ion battery out-of-population warning method as described in Embodiment 1 of the present invention.
[0078] The steps and methods involved in Examples 2, 3, 4 and 5 above correspond to those in Example 1. For specific implementation methods, please refer to the relevant description section of Example 1.
[0079] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)). The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for early warning of outlier activity in lithium-ion batteries, characterized in that, Includes the following steps: The electrical performance of lithium-ion batteries was tested to obtain the DC internal resistance, capacity percentage in the constant voltage stage, and capacity percentage in the constant current stage at different temperatures and with different capacities. A lifetime assessment model with three branches is constructed and trained. The three branches include a first branch for local outlier detection, a second branch for residual analysis, and a third branch for deep learning. The three branches in the life assessment model process DC internal resistance, constant voltage stage capacity ratio, and constant current stage capacity ratio in parallel. The results obtained from the three branches are then fused and used to make a decision to obtain the outlier detection result. The first branch performs local outlier detection by constructing a matrix graph. The second branch performs residual analysis by constructing a fitting formula based on linear decay law and sinusoidal fluctuation. The third branch captures the life decay law of lithium battery based on SOC DC internal resistance, capacity ratio in constant voltage stage, and capacity ratio in constant current stage, and obtains the outlier prediction results of lithium battery.
2. The lithium-ion battery outlier warning method as described in claim 1, characterized in that, The third branch learns the relationship between the DC internal resistance of the SOC, the capacity ratio during the constant voltage stage, the capacity ratio during the constant current stage, and the lifespan degradation of the lithium battery through a multi-head attention mechanism, and uses a correction mechanism to correct the learned lifespan degradation pattern.
3. The lithium-ion battery outlier warning method as described in claim 1, characterized in that, The specific steps for testing the electrical performance of lithium-ion batteries are as follows: Set different current, voltage, and temperature parameters; Electrical performance tests were conducted on the lithium-ion batteries to be tested by charging and discharging. By changing the current, voltage, and temperature parameters, repeated electrical performance tests were conducted, yielding different test results.
4. The lithium-ion battery outlier warning method as described in claim 1, characterized in that, The training steps for the life assessment model are as follows: Multiple sets of lithium-ion battery electrical performance test data were obtained to form a dataset. The electrical performance parameters in the dataset were preprocessed, and the data were labeled and divided into training set and test set. The life assessment model is trained using the training set and tested using the test set to obtain the trained life assessment model.
5. The lithium-ion battery outlier warning method as described in claim 4, characterized in that, The specific steps for preprocessing electrical performance parameters are as follows: Clean the electrical performance parameters to remove noise data and outliers; Electrical performance parameters with different dimensions and ranges are normalized to ensure they fall within the same numerical range.
6. A lithium-ion battery out-of-population early warning system, characterized in that, include: The data acquisition module is configured to perform electrical performance tests on lithium-ion batteries, including DC internal resistance, capacity percentage during constant voltage stage, and capacity percentage during constant current stage at different temperatures and different capacities. The model building module is configured to build a lifetime assessment model containing three branches and train the lifetime assessment model. The three branches include a first branch for local outlier detection, a second branch for residual analysis, and a third branch for deep learning. The outlier prediction module is configured to process the DC internal resistance, constant voltage stage capacity ratio, and constant current stage capacity ratio in parallel as three branches in the lifetime assessment model. The results obtained from the three branches are fused and decided to obtain the outlier detection result. The first branch performs local outlier detection by constructing a matrix graph. The second branch performs residual analysis by constructing a fitting formula based on linear decay law and sinusoidal fluctuation. The third branch captures the life decay law of lithium battery based on SOC DC internal resistance, capacity ratio in constant voltage stage, and capacity ratio in constant current stage, and obtains the outlier prediction results of lithium battery.
7. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the lithium-ion battery out-of-group warning method as described in any one of claims 1-5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-5, the lithium-ion battery out-of-population warning method.
9. A computer device, characterized in that, include: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the lithium-ion battery out-of-population warning method as described in any one of claims 1-5.
Citation Information
Patent Citations
New method for predicating residual life of lithium-ion battery
CN106291372A
Lithium ion battery health state diagnosis method based on deep learning model
CN113009349A
