Abnormal lithium ion battery prediction method, system, equipment and medium
By screening the optimal feature combination in the lithium-ion battery capacity separation process and building an integrated learning model, the problems of long prediction period and low accuracy in the existing lithium-ion battery prediction methods are solved, and fast and accurate prediction of abnormal lithium-ion battery is achieved.
Patent Information
- Application Number
- CN202510874064.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-27
- Publication Date
- 2025-08-08
AI Technical Summary
The existing lithium-ion battery prediction methods have problems such as long prediction cycle, high equipment occupancy and low abnormal identification accuracy. Integrated learning technology has not yet formed a systematic application solution in the prediction of abnormal lithium battery, which makes it difficult to achieve an ideal level of prediction efficiency and accuracy.
By obtaining the charge and discharge data of the lithium-ion battery capacity division process, using gray correlation analysis to screen the optimal feature combination, and combining Pearson's correlation coefficient to screen representative features, input an integrated learning model based on the base classifier for prediction, and construct an abnormal lithium-ion battery prediction model.
It significantly shortens the prediction cycle, reduces the equipment occupancy rate, and improves the accuracy of abnormal identification, achieving the ability to quickly predict abnormal lithium-ion batteries before being left to stand.
Smart Images
Figure CN120446767A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of lithium-ion battery production prediction, and in particular to a method, system, device and medium for predicting abnormal lithium-ion batteries. Background Art
[0002] In the existing technology, the prediction of lithium-ion batteries mainly relies on traditional static prediction methods and data-driven methods. The traditional static prediction method inevitably requires the lithium-ion battery to be stationary for more than 72 hours, which has problems such as long prediction cycle and high equipment occupancy rate, and cannot meet the needs of large-scale production. The existing data-driven method has the obvious defects of poor model generalization ability, especially low anomaly recognition accuracy, and even the anomaly recognition accuracy is usually less than 65%. In addition, although ensemble learning technology has excellent performance in anomaly prediction in other fields, it has not yet formed a systematic application solution in the prediction of abnormal lithium batteries. As a result, the existing prediction methods are difficult to reach the ideal level in terms of efficiency and accuracy. Therefore, it is necessary to combine reasonable data and processing methods with ensemble learning technology to achieve high-quality abnormal lithium-ion battery prediction. Summary of the Invention
[0003] The present invention provides a method, system, device, and medium for predicting abnormal lithium-ion batteries. These methods address the challenges of traditional static lithium-ion battery prediction methods, which suffer from long prediction cycles, high equipment utilization, low anomaly identification accuracy, and an inability to meet the needs of large-scale production. Existing data-driven methods, on the other hand, suffer from poor model generalization and unsatisfactory anomaly identification accuracy. Furthermore, ensemble learning technology has yet to achieve a systematic application in the prediction of abnormal lithium batteries, resulting in technical issues that hinder the efficiency and accuracy of existing prediction methods from reaching ideal levels.
[0004] The present invention provides an abnormal lithium-ion battery prediction method, system, device and medium. The abnormal lithium-ion battery prediction method includes: obtaining charge and discharge data of multiple lithium-ion batteries in a capacity separation process; integrating the charge and discharge data according to the numbering order of the lithium-ion batteries and performing feature extraction to obtain original features; based on the original features and pre-obtained abnormal battery features, using Grey Relational Analysis (GRA) to screen out an optimal feature combination from the original features; calculating the correlation coefficient between every two optimal features in the optimal feature combination to screen out representative features; inputting the representative features into an abnormal lithium-ion battery prediction model to obtain an abnormal lithium-ion battery prediction result; wherein the abnormal lithium-ion battery prediction model is an ensemble learning model constructed based on a base classifier.
[0005] In one embodiment of the present invention, based on the original features and the pre-obtained abnormal battery features, the step of using grey correlation analysis to screen out the optimal feature combination from the original features includes: respectively calculating the grey correlation between each original feature and the abnormal battery feature; the optimal feature takes the original feature whose grey correlation value is higher than the preset correlation threshold as the optimal feature, and combines all the optimal features into the optimal feature combination.
[0006] In one embodiment of the present invention, the grey correlation degree between each original feature and the abnormal battery feature is calculated according to the following formula:
[0007] Among them, r i is the grey correlation between the ith original feature and the abnormal battery feature, x0(k) is the kth eigenvalue of the abnormal battery feature, x i (k) is the kth eigenvalue of the i-th original feature, ρ is the resolution coefficient, which ranges from [0,1], and n is the number of eigenvalues of each original feature.
[0008] In one embodiment of the present invention, the step of screening and obtaining a representative feature according to the correlation coefficient between each two optimal features in the optimal feature combination includes: calculating the Pearson correlation coefficient (Pearson Correlation Coefficient) between the two optimal features in the optimal feature combination: when the Pearson correlation coefficient between the two optimal features is less than a preset coefficient threshold, both optimal features are retained; when the Pearson correlation coefficient between the two optimal features is greater than or equal to the preset coefficient threshold, one of the optimal features is randomly selected to be retained; and the retained optimal feature is used as the representative feature.
[0009] In one embodiment of the present invention, the Pearson correlation coefficient is calculated according to the following formula:
[0010] in, is the Pearson correlation coefficient between the optimal feature X and the optimal feature Y, cov(X,Y) is the covariance between the optimal feature X and the optimal feature Y, σ X is the standard deviation of the optimal feature X, σ Y is the standard deviation of the optimal feature Y.
[0011] In one embodiment of the present invention, the abnormal lithium-ion battery prediction model is obtained through pre-training, including the following steps: extracting feature data based on pre-acquired charge and discharge data of the lithium-ion battery in the capacity division stage; dividing the feature data according to a preset division ratio to obtain a validation set and multiple sub-training sets; training multiple different learning models used as base classifiers using the multiple sub-training sets to obtain target learning models corresponding to the learning models; sorting and optimizing all target learning models using an evolutionary algorithm combined with the validation set to obtain an optimal combination sequence, and constructing an ensemble learning model based on the optimal combination sequence to obtain the abnormal lithium-ion battery prediction model.
[0012] In one embodiment of the present invention, all target learning models are sorted and optimized by combining the verification set with an evolutionary algorithm to obtain an optimal combination order, and the steps of constructing an integrated learning model according to the optimal combination order to obtain the abnormal lithium-ion battery prediction model include: randomly generating multiple individuals and using the generated individuals as the initial population; wherein each individual represents a combination order of multiple target learning models; using the screening ratio of the integrated model corresponding to each individual for abnormal batteries in the verification set as the fitness value corresponding to each individual; screening individuals whose fitness values exceed a preset fitness threshold from the initial population as the screening population; and randomly adding one to each individual in the screening population. A target learning model is generated, and the fitness value of each individual is calculated, and individuals exceeding the preset fitness threshold are screened out as offspring populations; new individuals are obtained from the individuals of the offspring population through a crossover operation, and the new individuals are added to the offspring population to update the offspring population into a new offspring population; individuals whose fitness value exceeds the preset fitness threshold of each individual in the offspring population are screened, and the offspring population is continuously updated until a stopping condition is reached, and the combination sequence of multiple target learning models corresponding to the individual with the highest fitness value is used as the optimal combination sequence; an integrated learning model is constructed according to the optimal combination sequence and its corresponding multiple target learning models to obtain an abnormal lithium-ion battery prediction model.
[0013] The present invention also provides an abnormal lithium-ion battery prediction system, which includes: a data acquisition unit for acquiring charge and discharge data of multiple lithium-ion batteries in a capacity separation process; a feature extraction unit for integrating the charge and discharge data according to the numbering order of the lithium-ion batteries and performing feature extraction to obtain original features; a correlation analysis unit for screening the original features using grey correlation analysis based on the original features and pre-obtained abnormal battery features to obtain an optimal feature combination; a Pearson coefficient screening unit for calculating the correlation coefficient between every two optimal features in the optimal feature combination to screen out representative features; and an abnormality prediction unit for inputting the representative features into an abnormal lithium-ion battery prediction model to obtain an abnormal lithium-ion battery prediction result; wherein the abnormal lithium-ion battery prediction model is an ensemble learning model constructed based on a base classifier.
[0014] The present invention also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device implements the abnormal lithium-ion battery prediction method as described above.
[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor of a computer, the computer is caused to execute the abnormal lithium-ion battery prediction method as described above.
[0016] Beneficial effects of the present invention: The present invention proposes an abnormal lithium-ion battery prediction method, system, device and medium, which extracts the original features of the charge and discharge data of the lithium-ion battery capacity classification stage, and performs grey correlation analysis with the abnormal battery features obtained in advance, and preliminarily screens the original features to obtain the optimal feature combination, thereby improving the accuracy of the feature data. Re-screening is performed based on the Pearson correlation coefficient between the features of the optimal feature combination after the initial screening to obtain representative features, and the features with low correlation are screened out as the input of the abnormal lithium-ion battery prediction model, which not only reduces the dimension of the data set and improves the training efficiency of the model, but also avoids the overfitting problem caused by too many features. At the same time, by retaining features with high correlation with the target variable, it ensures that the model can capture key information, thereby improving the performance of the model. Finally, the representative features after re-screening are input into the trained evolutionary ensemble learning model based on the Cascading method: the abnormal lithium-ion battery prediction model, which can quickly predict abnormal lithium-ion batteries. The advantage of this method is that it can predict abnormal lithium-ion batteries before the lithium-ion batteries are stationary. At the same time, compared with the traditional stationary prediction method, the present invention significantly shortens the prediction cycle and reduces the equipment occupancy rate. Through the optimized ensemble learning model, the accuracy of abnormality identification is improved, solving the problems of low prediction efficiency and insufficient accuracy in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be derived from these drawings without inventive effort.
[0018] In the attached figure:
[0019] Figure 1 A schematic flow chart of a method for predicting abnormal lithium-ion batteries according to an embodiment of the present invention;
[0020] Figure 2 1 is a schematic structural diagram of a Cascading method provided in one embodiment of the present invention;
[0021] Figure 3 This is a structural block diagram of an abnormal lithium-ion battery prediction system provided in one embodiment of the present invention;
[0022] Figure 4 FIG. 1 is a structural diagram of an electronic device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0023] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.
[0024] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.
[0025] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.
[0026] See Figure 1 , Figure 1 A flow chart of a method for predicting abnormal lithium-ion batteries according to an embodiment of the present invention is shown in FIG. Figure 1 As shown, the following steps are included:
[0027] S100, acquiring charge and discharge data of a plurality of lithium-ion batteries in a capacity division process.
[0028] The capacity grading process for lithium-ion batteries is a necessary step before static testing. This process follows a charge and discharge pattern: stagnation, constant current and constant voltage charging, stagnation, multiple constant current charges, stagnation, constant current and constant voltage charging, and stagnation. During this process, abnormal lithium-ion batteries will exhibit slight differences from normal ones. Traditional static testing relies on the actual performance of the lithium-ion battery during the static test to determine if there are quality issues. Therefore, data from the capacity grading stage can be used to distinguish between these batteries. The voltage data of the lithium-ion batteries during the capacity grading process is recorded to provide raw charge and discharge data for subsequent predictions.
[0029] S200 , integrating the charge and discharge data according to the serial number sequence of the lithium-ion batteries and performing feature extraction to obtain original features.
[0030] S300 , based on the original features and the abnormal battery features obtained in advance, using grey relational analysis to screen out the optimal feature combination from the features.
[0031] To assess the correlation between the extracted features and battery anomalies, this paper applies grey relational analysis. This method, based on grey system theory, measures correlation based on the similarity between factors. Grey relational analysis provides a quantitative measure of a system's evolution and is therefore well-suited for dynamic process analysis.
[0032] Specifically, in one embodiment of the present invention, based on the original features and the pre-obtained abnormal battery features, the step of using grey relational analysis to screen out the optimal feature combination from the original features includes:
[0033] S310 calculates the grey correlation between each original feature and the abnormal battery feature.
[0034] In order to reduce the dimension of input data and improve data quality, grey correlation analysis was used to screen out original features whose grey correlation with abnormal battery features met the requirements. In order to more accurately analyze the sequences between the comparison sequence and the reference sequence, the original data set was normalized before the correlation analysis to avoid the influence of different dimensions on the analysis results.
[0035] Furthermore, the grey correlation degree between each original feature and the abnormal battery feature is calculated according to the following formula:
[0036] Among them, r i is the grey correlation between the ith original feature and the abnormal battery feature, x0(k) is the kth eigenvalue of the abnormal battery feature, x i (k) is the kth eigenvalue of the i-th original feature, ρ is the resolution coefficient, which ranges from [0,1] and is usually 0.5, and n is the number of eigenvalues of each original feature. is the minimum difference between the two levels, that is, the global minimum difference, the double maximum in the denominator It is the maximum difference between the two levels (global maximum difference).
[0037] S320: The original features whose grey correlation values are higher than a preset correlation threshold are taken as optimal features, and all the optimal features are combined into the optimal feature combination.
[0038] After grey correlation analysis, the grey correlation between each original feature and the preset abnormal battery feature is determined, and the original feature whose grey correlation is higher than the preset correlation threshold is taken as the optimal feature, and all the optimal features are combined to obtain the optimal feature combination. At this time, the dimension of the optimal feature combination after screening is significantly less than that of the original feature before processing.
[0039] S400, calculating the correlation coefficient between every two optimal features in the optimal feature combination, and screening to obtain representative features.
[0040] Among the optimal feature combinations obtained after screening, some optimal features still have obvious correlation with other optimal features. This means that there is more identical or related information in the data represented by the two optimal features. Therefore, for multiple optimal features with high correlation, one of the optimal features can be selected as the representative input model. This can further reduce the dimension of the input data and improve the efficiency of model training and prediction while ensuring less information loss.
[0041] Specifically, in a specific embodiment of the present application, the step of screening and obtaining a representative feature according to the correlation coefficient between every two optimal features in the optimal feature combination includes:
[0042] S410, calculating the Pearson correlation coefficient between the two optimal features in the optimal feature combination:
[0043] When evaluating the correlation between various features, a correlation heat map is usually used as an analysis method, and the methods commonly used to draw a correlation heat map are: Pearson correlation coefficient, Spearman's Rank Correlation Coefficient, Kendall's Tau Coefficient, etc., where each method can achieve better results for data types with different characteristics. The data used in this application conforms to the normal distribution law, so the Pearson correlation coefficient is selected as the correlation calculation method for evaluating the optimal feature combination. The value range is between -1 and 1, where -1 represents a complete negative correlation, 1 represents a complete positive correlation, and 0 represents no linear correlation. For data with other characteristics, for example: Spearman's Rank Correlation Coefficient does not require the data to obey the normal distribution, is not sensitive to outliers, and is applicable to situations where there is a monotonic relationship but not a linear relationship between variables. Kendall's Rank Correlation Coefficient has relatively loose requirements on data and is applicable to situations where the sample size is small. Those skilled in the art can select other correlation calculation methods according to the data type. Furthermore, in one embodiment of the present invention, the Pearson correlation coefficient is calculated according to the following formula:
[0044] in, is the Pearson correlation coefficient between the optimal feature X and the optimal feature Y, cov(X,Y) is the covariance between the optimal feature X and the optimal feature Y, σ X is the standard deviation of the optimal feature X, σ Y is the standard deviation of the optimal feature Y, and the Pearson correlation coefficient between the two optimal features in the optimal feature combination is calculated using this formula.
[0045] When the Pearson correlation coefficient between two optimal features is less than a preset coefficient threshold, both optimal features are retained.
[0046] When the Pearson correlation coefficient between two optimal features is greater than or equal to the preset coefficient threshold, one of the optimal features is randomly selected and retained.
[0047] According to the Pearson correlation coefficient between the two optimal features, when it is greater than the preset coefficient threshold, it indicates that the correlation between the two optimal features is strong, so one of the optimal features is randomly selected and retained. If the Pearson correlation coefficient between the two optimal features is less than or equal to the preset coefficient threshold, it indicates that the correlation between the two optimal features is weak, and there is less common content in the information expressed by the two optimal features, so both optimal features are retained.
[0048] S420: The best feature is retained as a representative feature.
[0049] S500: Inputting the representative features into an abnormal lithium-ion battery prediction model to obtain an abnormal lithium-ion battery prediction result; wherein the abnormal lithium-ion battery prediction model is an ensemble learning model constructed based on a base classifier.
[0050] The obtained representative features are input into the abnormal lithium-ion battery prediction model, and the model outputs the abnormal lithium-ion battery prediction results. The construction of the ensemble learning model is generally based on various strategies such as bagging, boosting, and cascading. Considering the size of the training set and the proportion of abnormal data, traditional bagging, boosting and other strategies are not applicable. For this reason, this article will use an ensemble learning framework based on the cascading method. The cascading method, also known as cascade integration or sequence integration, is a strategy that arranges multiple models in order and processes the data set in sequence. In this method, each model is assigned a specific task or focuses on a certain aspect of the data, and its output will be directly used as the input of the next model. The core idea of cascade integration is to gradually filter and refine the data, take advantage of the complementarity of different models in processing data, have strong universality, and simple training, thereby improving the accuracy and generalization ability of the final model.
[0051] like Figure 2As shown in the figure, it is a schematic diagram of the Cascading method structure, and the representative characteristics of multiple lithium-ion batteries D train Input to the first base classifier M1, after screening by the base classifier M1, the sample D1 whose abnormal probability exceeds the preset abnormal probability threshold is eliminated. P(y=1)>0.98 means that the abnormal probability of sample D1 is greater than 0.98, that is, it is judged to be an abnormal lithium-ion battery, and its corresponding label is 1, that is, The optimal value of the preset abnormal probability threshold is: 0.98, and then the data set of sample D1 will be eliminated Continue to input the subsequent base classifiers M2 and M3. Here, we only take the case where the base classifier M1 screens out one abnormal sample as an example. A base classifier screens out multiple abnormal samples, and the number of abnormal samples screened out by each base classifier may also be different. Figure 2 The possibility of parallel screening is shown in Figure 1. If necessary, only base classifier M2 or base classifier M3 can be used for screening. When the screening results of base classifier M2 and base classifier M3 for the same sample are different, the screening is performed according to the abnormal battery union method. For example, when base classifier M2 or base classifier M3 determines that battery No. 2 is an abnormal battery, then battery No. 2 is determined to be an abnormal battery and the label is 1. Finally, the representative features of multiple lithium-ion batteries corresponding to the input abnormal lithium-ion battery prediction model are passed through all base classifiers to eliminate abnormal samples to obtain the normal lithium-ion battery sample D. answer Finally, all lithium-ion batteries with label 1 are predicted to be abnormal lithium-ion batteries.
[0052] Furthermore, in one embodiment of the present application, the abnormal lithium-ion battery prediction model is pre-trained and the steps include:
[0053] First, characteristic data is extracted based on the previously acquired charge and discharge data of the lithium-ion battery in the capacity division stage;
[0054] Secondly, the feature data is divided into a validation set and multiple sub-training sets according to a preset division ratio;
[0055] Thirdly, multiple sub-training sets are used to train different learning models using multiple base classifiers to obtain target learning models corresponding to the learning models;
[0056] The base classifier can be trained according to the corresponding sub-training set respectively, or can be trained through multiple sub-training sets. The abnormal screening ratio can be set to a smaller value so that the learning model can complete training faster.
[0057] Finally, all target learning models are sorted and optimized by combining the evolutionary algorithm with the verification set to obtain the optimal combination order, and an integrated learning model is constructed according to the optimal combination order to obtain the abnormal lithium-ion battery prediction model.
[0058] Multiple trained learning models are sorted using an evolutionary algorithm, enabling each base classifier to achieve high prediction accuracy for the final ensemble model even with a low screening ratio. Compared to traditional single prediction models, this approach offers a higher accuracy while ensuring a 100% screening ratio. For example, assuming 10 out of 100 lithium-ion batteries contain abnormalities, a single prediction model might need to screen 15 or even more to identify them, while the ensemble learning model only needs to screen 11 to identify all abnormal batteries. This is because a single model needs to set very conservative parameters and a high abnormal screening ratio to predict all abnormal batteries, while ensemble learning allows each base classifier to have a smaller abnormal screening ratio to guarantee the final result. Consequently, the training efficiency and final prediction results of a single prediction model are inferior to those of the ensemble model.
[0059] Specifically, in a specific embodiment of the present invention, all target learning models are sorted and optimized by combining the validation set with an evolutionary algorithm to obtain an optimal combination sequence, and an integrated learning model is constructed according to the optimal combination sequence to obtain the abnormal lithium-ion battery prediction model. The steps include:
[0060] Randomly generate multiple individuals and use the generated individuals as the initial population; each individual represents a combination sequence of multiple target learning models;
[0061] The screening ratio of abnormal batteries in the validation set by the ensemble model corresponding to each individual is taken as the fitness value corresponding to each individual;
[0062] Selecting individuals whose fitness values exceed a preset fitness threshold from the initial population as a screening population;
[0063] Randomly adding a target learning model to each individual in the screening population, calculating the fitness value of each individual, and screening individuals that exceed the preset fitness threshold as the offspring population;
[0064] Obtain new individuals from the individuals of the offspring population through a crossover operation, and add the new individuals to the offspring population to update it into a new offspring population;
[0065] Screening the individuals in the offspring population whose fitness value exceeds the preset fitness threshold, and continuing to update the offspring population until a stopping condition is reached, and taking the combination order of multiple target learning models corresponding to the individual with the highest fitness value as the optimal combination order;
[0066] An integrated learning model is constructed according to the optimal combination sequence and its corresponding multiple target learning models to obtain an abnormal lithium-ion battery prediction model.
[0067] Compared to traditional evolutionary algorithms, the evolutionary algorithm used in this invention has been modified to improve training efficiency, retaining only the crossover and mutation processes of the population. However, those skilled in the art can select other optimization algorithms to obtain the optimal combination order of base classifiers as needed. To further determine the irreplaceability of the base classifiers in the optimal combination order, ablation experiments can be performed to verify whether the prediction effect of the ensemble learning model is affected by gradually removing or modifying certain base classifiers, ultimately verifying the optimal combination order.
[0068] like Figure 3 As shown, the abnormal lithium-ion battery prediction system 200 includes: a data acquisition unit 210, a feature extraction unit 220, a correlation analysis unit 230, a Pearson coefficient screening unit 240, and an abnormality prediction unit 250, wherein the data acquisition unit 210 is used to obtain charge and discharge data of multiple lithium-ion batteries in the capacity separation process; the feature extraction unit 220 is used to integrate the charge and discharge data according to the numbering order of the lithium-ion batteries and perform feature extraction to obtain original features; the correlation analysis unit 230 is used to screen the original features based on the original features and the abnormal battery features obtained in advance using grey correlation analysis to obtain the optimal feature combination; the Pearson coefficient screening unit 240 is used to calculate the correlation coefficient between each two optimal features in the optimal feature combination to screen out representative features; the abnormality prediction unit 250 is used to input the representative features into the abnormal lithium-ion battery prediction model to obtain abnormal lithium-ion battery prediction results; wherein the abnormal lithium-ion battery prediction model is an ensemble learning model constructed based on a base classifier.
[0069] The above modules may be embedded in or independent of the processor in the computer device in hardware format, or may be stored in the memory of the computer device in software format, so that the processor can call the corresponding operations of the above modules.
[0070] It should be noted that, in order to highlight the innovative part of the present invention, this embodiment does not introduce modules that are not closely related to solving the technical problem proposed by the present invention, but this does not mean that there are no other modules in this embodiment.
[0071] See Figure 4 The present invention also discloses an electronic device 3, including a memory 32, a processor 31 and a bus, and may also include a computer program stored in the memory 32 and executable on the processor 31, such as a Pearson correlation coefficient calculation program.
[0072] Among them, the memory 32 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 32 can be an internal storage unit of the electronic device 3, such as a mobile hard disk of the electronic device 3. In other embodiments, the memory 32 can also be an external storage device of the electronic device 3, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. equipped on the electronic device 3. Furthermore, the memory 32 can also include both an internal storage unit of the electronic device 3 and an external storage device. The memory 32 can not only be used to store application software installed on the electronic device 3 and various types of data, such as the gray correlation between each original feature and the abnormal battery feature, but can also be used to temporarily store data that has been output or is about to be output.
[0073] In some embodiments, the processor 31 may be comprised of an integrated circuit, such as a single packaged integrated circuit or a combination of multiple packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 31 is the control core (Control Unit) of the electronic device 3. It utilizes various interfaces and circuits to connect the various components of the entire electronic device 3. It executes or runs programs or modules (such as evolutionary algorithm calculation programs) stored in the memory 32 and accesses data stored in the memory 32 to perform various functions and process data of the electronic device 3.
[0074] The processor 31 executes the operating system and various installed application programs of the electronic device 3. The processor 31 executes the application programs to implement the steps in the above-mentioned abnormal lithium-ion battery prediction method.
[0075] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory 32 and executed by the processor 31 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device 3.
[0076] The above-mentioned integrated unit implemented in the form of a software functional module can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The above-mentioned software functional module is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, computer device, or network device, etc.) or a processor to perform part of the functions of the battery detection method described in various embodiments of the present application.
[0077] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical principles disclosed herein are intended to be covered by the claims of the present invention.
Claims
1. A method for predicting abnormal lithium-ion batteries, characterized in that: The method comprises: Acquire charge and discharge data of multiple lithium-ion batteries in a capacity division process; Integrating the charge and discharge data according to the serial number sequence of the lithium-ion batteries and performing feature extraction to obtain original features; Based on the original features and the abnormal battery features obtained in advance, the optimal feature combination is screened from the original features using grey relational analysis; Calculating the correlation coefficient between every two optimal features in the optimal feature combination, and screening to obtain representative features; The representative features are input into an abnormal lithium-ion battery prediction model to obtain an abnormal lithium-ion battery prediction result; wherein the abnormal lithium-ion battery prediction model is an integrated learning model constructed based on a base classifier.
2. The abnormal lithium-ion battery prediction method according to claim 1, characterized in that: The step of screening the optimal feature combination from the original features using grey relational analysis based on the original features and the pre-obtained abnormal battery features includes: Calculating the grey correlation degree between each original feature and the abnormal battery feature respectively; The optimal feature takes the original feature whose grey correlation value is higher than a preset correlation threshold as the optimal feature, and combines all the optimal features into the optimal feature combination.
3. The abnormal lithium-ion battery prediction method according to claim 2, characterized in that: The grey correlation degree between each original feature and the abnormal battery feature is calculated according to the following formula: Among them, r i is the grey correlation between the ith original feature and the abnormal battery feature, x0(k) is the kth eigenvalue of the abnormal battery feature, x i (k) is the kth eigenvalue of the i-th original feature, ρ is the resolution coefficient, which ranges from [0,1], and n is the number of eigenvalues of each original feature.
4. The abnormal lithium-ion battery prediction method according to claim 1, characterized in that: The step of screening and obtaining representative features according to the correlation coefficient between every two optimal features in the optimal feature combination includes: Calculate the Pearson correlation coefficient between the two best features in the optimal feature combination: When the Pearson correlation coefficient between two optimal features is less than the preset coefficient threshold, both optimal features are retained; When the Pearson correlation coefficient between the two optimal features is greater than or equal to the preset coefficient threshold, one of the optimal features is randomly selected and retained; The best feature is retained as the representative feature.
5. The abnormal lithium-ion battery prediction method according to claim 4, characterized in that: The Pearson correlation coefficient is calculated according to the following formula: in, is the Pearson correlation coefficient between the optimal feature X and the optimal feature Y, cov(X,Y) is the covariance between the optimal feature X and the optimal feature Y, σ X is the standard deviation of the optimal feature X, σ Y is the standard deviation of the optimal feature Y.
6. The abnormal lithium-ion battery prediction method according to claim 1, characterized in that: The abnormal lithium-ion battery prediction model is obtained through pre-training steps including: Extracting characteristic data based on previously acquired charge and discharge data of the lithium-ion battery in the capacity division stage; Dividing the feature data into a validation set and multiple sub-training sets according to a preset division ratio; Using multiple sub-training sets to train multiple different learning models as base classifiers, a target learning model corresponding to the learning model is obtained; All target learning models are sorted and optimized by combining the evolutionary algorithm with the verification set to obtain the optimal combination sequence, and an integrated learning model is constructed according to the optimal combination sequence to obtain the abnormal lithium-ion battery prediction model.
7. The abnormal lithium-ion battery prediction method according to claim 6, characterized in that: The steps of sorting and optimizing all target learning models by combining the evolutionary algorithm with the verification set to obtain an optimal combination sequence, and constructing an integrated learning model according to the optimal combination sequence to obtain the abnormal lithium-ion battery prediction model include: Randomly generate multiple individuals and use the generated individuals as the initial population; each individual represents a combination sequence of multiple target learning models; The screening ratio of abnormal batteries in the validation set by the ensemble model corresponding to each individual is taken as the fitness value corresponding to each individual; Selecting individuals whose fitness values exceed a preset fitness threshold from the initial population as a screening population; Randomly adding a target learning model to each individual in the screening population, calculating the fitness value of each individual, and screening individuals that exceed the preset fitness threshold as the offspring population; Obtain new individuals from the individuals of the offspring population through a crossover operation, and add the new individuals to the offspring population to update it into a new offspring population; Screening the individuals in the offspring population whose fitness value exceeds the preset fitness threshold, and continuing to update the offspring population until a stopping condition is reached, and taking the combination order of multiple target learning models corresponding to the individual with the highest fitness value as the optimal combination order; An integrated learning model is constructed according to the optimal combination sequence and its corresponding multiple target learning models to obtain an abnormal lithium-ion battery prediction model.
8. An abnormal lithium-ion battery prediction system, characterized in that: The system comprises: A data acquisition unit, used to acquire charge and discharge data of multiple lithium-ion batteries in the capacity division process; A feature extraction unit, configured to integrate the charge and discharge data according to the serial number sequence of the lithium-ion batteries and perform feature extraction to obtain original features; A correlation analysis unit, configured to screen the original features and obtain an optimal feature combination by using grey correlation analysis based on the original features and the abnormal battery features obtained in advance; A Pearson coefficient screening unit is used to calculate the correlation coefficient between every two optimal features in the optimal feature combination and screen out representative features; an abnormality prediction unit, configured to input the representative features into an abnormal lithium-ion battery prediction model, An abnormal lithium-ion battery prediction result is obtained; wherein, the abnormal lithium-ion battery prediction model is an integrated learning model constructed based on a base classifier.
9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the abnormal lithium-ion battery prediction method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the abnormal lithium-ion battery prediction method according to any one of claims 1 to 7.