Battery cell self-discharge prediction method and device, electronic equipment and storage medium
By using a cell self-discharge prediction model, machine learning combined with process data is used to quickly and efficiently predict the cell self-discharge level, solving the problems of high cost and long testing time in existing technologies, and achieving more reliable prediction and process optimization.
Patent Information
- Application Number
- CN202411128855.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-11-05
AI Technical Summary
Existing technologies for testing the self-discharge of battery cells are costly, time-consuming, and resource-intensive, making it difficult to efficiently detect the self-discharge level of battery cells.
By training a cell self-discharge prediction model, and using process data and historical self-discharge data from the cell production line, a machine learning model is used for prediction, including feature data screening and model training, to quickly predict the cell's self-discharge level.
It reduced time, manpower, and material costs, improved the reliability and generalization ability of prediction results, and optimized the process level of battery cell production lines.
Smart Images

Figure CN118897989A_ABST
Abstract
Description
Technical Field
[0001] The present invention mainly relates to the technical field of battery cell abnormality detection, and in particular to a battery cell self-discharge prediction method, device, electronic equipment and storage medium. Background Art
[0002] The self-discharge phenomenon of battery cells (such as lithium iron phosphate energy storage cells) affects important aspects such as battery cell assembly and battery SOC estimation correction in actual production. Therefore, it is necessary to detect the self-discharge level of battery cells during the production process. In related technologies, manufacturers test battery cells with self-discharge test equipment before the battery cells are put on the market. Although this method can detect abnormal characteristics in the battery cells, the cost of the test equipment is high, the test process requires human, material and financial resources, and the test time is long. Summary of the invention
[0003] The purpose of the present invention is to provide a method, device, electronic device and storage medium for predicting battery cell self-discharge, which can quickly and efficiently predict the self-discharge level of the battery cell through a trained battery cell self-discharge prediction model. The feature data used by the battery cell self-discharge prediction model are all derived from the process data of each production process, which reduces the time, manpower and material costs and improves the reliability of the prediction results.
[0004] In a first aspect, the present application provides a method for predicting self-discharge of a battery cell, comprising: collecting process data and historical self-discharge data of a formation stage from a battery cell production line, using the process data to construct a data set to be trained, wherein the formation stage includes multiple processes, and the data set to be trained corresponds to multiple formation features set for the formation stage, and each formation feature is associated with a category of data to be collected in the corresponding process; performing correlation screening on the multiple formation features to obtain a relevant feature subset, and then performing mutual information screening on the relevant feature subset to obtain an optimal feature subset; selecting the best input data in the process data according to the optimal feature subset, and using the best input data and the historical self-discharge data to train a machine learning model to obtain a battery cell self-discharge prediction model; and inputting the current process data of the battery cell production line into the battery cell self-discharge prediction model, so that the battery cell self-discharge prediction model outputs a prediction result.
[0005] In some embodiments, the formation stage includes a liquid injection process and a high-temperature immersion process, and the formation characteristics include any one or a combination of the following: Hi-pot resistance; cell weighing before the first liquid injection; cell weighing after the first liquid injection; first liquid injection volume; liquid retention volume; weighing before the second liquid injection; weighing after the second liquid injection; high-temperature immersion time.
[0006] In some embodiments, the formation stage includes a formation process, a capacity division process and an OCV process, and the formation characteristics include any one or a combination of the following: formation liquid loss; formation charging capacity; capacity division charging capacity; capacity division discharge capacity; capacity division discharge average voltage; capacity division discharge average temperature; capacity division first effect; OCV open circuit voltage; OCV internal resistance; OCV test temperature.
[0007] In some embodiments, the correlation screening of the multiple formation characteristics to obtain a correlated feature subset includes: using the Pearson correlation coefficient to calculate the linear correlation between the multiple formation characteristics, and grouping the multiple formation characteristics according to a preset correlation threshold to determine at least one high-correlation feature group and at least one low-correlation feature group; retaining each low-correlation feature group, retaining one formation feature in each high-correlation feature group, and deleting the remaining formation features to obtain a correlated feature subset.
[0008] In some embodiments, the mutual information screening of the relevant feature subsets to obtain the optimal feature subset includes: using the mutual information method to calculate the importance between each formation feature in the relevant feature subset and the target value; sorting the relevant feature subsets in order of importance from large to small, retaining the formation features with corresponding importance in the top 80%, deleting the remaining formation features, and obtaining the optimal feature subset.
[0009] In some embodiments, the method of using the optimal input data and the historical self-discharge data to train the machine learning model to obtain a battery cell self-discharge prediction model includes: determining a sample data set and a test data set based on the optimal input data and the historical self-discharge data, and performing multiple rounds of iterative training on the machine learning model using the sample data set, wherein the sample data set includes multiple samples, each sample includes a sample input obtained from the optimal input data and a true label obtained from the historical self-discharge data, and the true label represents the historical self-discharge K value corresponding to the sample input; in each round of training, performing the following steps: evaluating the model performance of the machine learning model based on the coefficient of determination and the root mean square error of a sample, and optimizing the model hyperparameters of the machine learning model using grid search based on the evaluation results, and continuing the next round of training using the optimized machine learning model; evaluating the trained machine learning model using the test data set, and optimizing the model hyperparameters of the machine learning model again based on the evaluation results to obtain a battery cell self-discharge prediction model.
[0010] In some embodiments, the determination coefficient is calculated using the following formula:
[0011]
[0012] And, the root mean square error is calculated using the following formula:
[0013]
[0014] Among them, R 2 is the determination coefficient, RMSE is the root mean square error, n is the number of samples, a i exp is the true label of the i-th sample, a i pred is the predicted value of the machine learning model for the i-th sample.
[0015] In a second aspect, the present application provides a battery cell self-discharge prediction device, comprising: a data collection module, used to collect process data and historical self-discharge data of a formation stage from a battery cell production line, and use the process data to construct a data set to be trained, wherein the formation stage includes multiple processes, and the data set to be trained corresponds to multiple formation features set for the formation stage, and each formation feature is associated with a category of data to be collected in the corresponding process; a feature screening module, used to perform correlation screening on the multiple formation features to obtain a relevant feature subset, and then perform mutual information screening on the relevant feature subset to obtain an optimal feature subset; a model training module, used to select the best input data in the process data according to the optimal feature subset, and use the optimal input data and the historical self-discharge data to train a machine learning model to obtain a battery cell self-discharge prediction model; a self-discharge prediction module, used to input the current process data of the battery cell production line into the battery cell self-discharge prediction model, so that the battery cell self-discharge prediction model outputs a prediction result.
[0016] In a third aspect, the present application provides an electronic device, including a processor and a memory, wherein the memory stores instructions, and when the instructions are called by the processor, the processor executes the method described in any one of the first aspects above.
[0017] In a fourth aspect, the present application provides a storage medium comprising computer program instructions, wherein the computer program instructions can be called by a computer to implement a method as described in any one of the first aspects above.
[0018] Compared with the prior art, the present invention has the following advantages:
[0019] The present application provides a method, device, electronic device and storage medium for predicting battery cell self-discharge. A trained battery cell self-discharge prediction model is used to quickly and efficiently predict the self-discharge level of a battery cell, thereby reducing time, manpower and material costs. The feature data used by the battery cell self-discharge prediction model are all derived from the process data of each production process, which improves the reliability of the prediction results and has strong generalization capabilities, which helps to optimize the process level of the battery cell production line. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The accompanying drawings are included to provide a further understanding of the present application. They are included and constitute a part of the present application. The accompanying drawings illustrate embodiments of the present application and together with the present specification serve to explain the principles of the present application. In the accompanying drawings:
[0021] Figure 1 This is a flowchart of a method for predicting self-discharge of a battery cell provided by way of example in this application;
[0022] Figure 2 This is a schematic diagram of the prediction effect of a battery cell self-discharge prediction model provided by the present application as an example;
[0023] Figure 3 It is a schematic diagram of the prediction effect of a comparative prediction model provided by the present application as an example;
[0024] Figure 4 is a schematic diagram of a battery cell self-discharge prediction device exemplarily provided in the present application;
[0025] Figure 5 This is a schematic diagram of an electronic device provided as an example in this application. DETAILED DESCRIPTION
[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.
[0027] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0028] In addition, it should be noted that the use of words such as "first" and "second" to define components is only for the convenience of distinguishing the corresponding components. If not otherwise stated, the above words have no special meaning and cannot be understood as limiting the scope of protection of this application. In addition, although the terms used in this application are selected from well-known and commonly used terms, some terms mentioned in the specification of this application may be selected by the applicant at his or her discretion, and their detailed meanings are explained in the relevant parts of the description of this article. In addition, it is required to understand this application not only by the actual terms used, but also by the meaning implied by each term.
[0029] The technical concept of the present application is: in the related art, the manufacturer tests the battery cells through self-discharge test equipment before the battery cells are put on the market. Although this method can detect abnormal characteristics in the battery cells, the cost of the test equipment is relatively high, the test process requires manpower, material resources and financial resources, and the test time is long. Therefore, the self-discharge level of the battery cell can be predicted by combining the process data of the battery cell production line with the machine learning model. The prediction process is implemented by computer, which is faster and more efficient, reducing the cost of manpower and material resources. The collection of process data corresponds to multiple processes in the formation stage, so that the trained machine learning model can closely integrate data analysis with the production process of the battery cell production line, which helps to improve the reliability of the prediction results and the generalization ability of the model, and provides strong support for the process improvement of the battery cell production line.
[0030] refer to Figure 1 In a first aspect of an exemplary embodiment of the present application, a method for predicting self-discharge of a battery cell is provided, comprising:
[0031] S1, collect process data and historical self-discharge data of the formation stage from the battery production line, and use the process data to construct a training data set, where the formation stage includes multiple steps, and the training data set corresponds to multiple formation features set for the formation stage, and each formation feature is associated with a category of data that needs to be collected in the corresponding step.
[0032] S2, performing correlation screening on multiple formation features to obtain a related feature subset, and then performing mutual information screening on the related feature subset to obtain an optimal feature subset.
[0033] S3, selecting the best input data in the process data according to the best feature subset, and using the best input data and historical self-discharge data to train the machine learning model to obtain a battery cell self-discharge prediction model.
[0034] S4, inputting the current process data of the battery cell production line into the battery cell self-discharge prediction model, so that the battery cell self-discharge prediction model outputs a prediction result.
[0035] In some embodiments, the battery cell production line may refer to, for example, a production line for lithium iron phosphate energy storage cells. The historical self-discharge data may specifically include the historical self-discharge K value of the battery cell production line, so the output of the battery cell self-discharge prediction model may also be the predicted new self-discharge K value of the lithium iron phosphate energy storage cell. The current process data of the battery cell production line may include data that has not participated in the training process of the machine learning model, and of course, may also include process data that has participated in the above training process, without limitation.
[0036] The processes included in the formation stage of the current battery cell production line are generally: liquid injection process, high-temperature infiltration process, formation process, capacity division process and OCV process (i.e., open circuit voltage detection process). The data of one or more formation characteristics that can be recorded by relevant instruments in these processes can be used as process data in the formation stage. Specifically, for the liquid injection process, the formation characteristics may include: Hi-pot resistance (abbreviated as Hp), cell weighing before the first liquid injection (abbreviated as GoneB), cell weighing after the first liquid injection (abbreviated as GoneA), the first liquid injection volume (abbreviated as Gone), and the liquid retention volume (abbreviated as Gkeep) in any one or combination; for the high-temperature infiltration process, the formation characteristics may include: weighing before the second liquid injection (abbreviated as GtwoB), weighing after the second liquid injection (abbreviated as GtwoA), and the high-temperature infiltration time (abbreviated as Tjr) in any one or combination; for the formation process, the formation characteristics may include: formation liquid loss volume (abbreviated as Gloss), formation charging capacity (abbreviated as AhForC) any one or combination; for the capacity division process, the formation characteristics may include: capacity division charging capacity (abbreviated as AhVolC), capacity division discharge capacity (abbreviated as AhVolF), capacity division discharge average voltage (abbreviated as VaveF), capacity division discharge average temperature (abbreviated as TemAveF), first effect (abbreviated as Sx) any one or combination; for the OCV process, the formation characteristics may include: open circuit voltage (abbreviated as OCV1), internal resistance (abbreviated as IR1), test temperature (abbreviated as Tem) any one or combination. That is to say, in step S1, the collected process data may specifically include the Hi-pot resistance value of the battery cell production line in the formation stage, the weighing of the battery cell before the first liquid injection, the weighing of the battery cell after the first liquid injection, the first liquid injection volume, the liquid retention volume, the weighing before the second liquid injection, the weighing after the second liquid injection, the high-temperature immersion time, the formation liquid loss, the formation charging capacity, the divided capacity charging capacity, the divided capacity discharge capacity, the divided capacity discharge average voltage, the divided capacity discharge average temperature, the divided capacity first effect, the OCV open circuit voltage, the OCV internal resistance, and the OCV test temperature related data. One or more.
[0037] Therefore, in this embodiment, the training data set can be completely constructed by relying on the process data and historical self-discharge data of the battery cell production line, which helps to closely integrate with the actual process of the battery cell production line.
[0038] In some embodiments, in step S2, multiple formation features are screened for correlation to obtain a correlated feature subset, including: using the Pearson correlation coefficient to calculate the linear correlation between the multiple formation features, and grouping the multiple formation features according to a preset correlation threshold to determine at least one corresponding high-correlation feature group and at least one low-correlation feature group; retaining each low-correlation feature group, and retaining one formation feature in each high-correlation feature group, deleting the remaining formation features, and obtaining a correlated feature subset.
[0039] Further, in step S2, mutual information screening is performed on the relevant feature subsets to obtain the optimal feature subset, including: using the mutual information method to calculate the importance between each formation feature in the relevant feature subset and the target value; sorting the relevant feature subsets in order of importance from large to small, retaining the formation features with the top 80% of the corresponding importance, deleting the remaining formation features, and obtaining the optimal feature subset.
[0040] It should be noted that in the above step S2, the correlation threshold can be adjusted according to the actual situation of the training process, such as the number of features to be retained. For example, the correlation threshold can be adjusted to 0.8. Similarly, the importance ratio in step S2 can also be adjusted based on 80% with reference to the actual situation, and there is no limit to this. The purpose of determining the optimal feature subset based on the Pearson correlation coefficient and mutual information screening is to filter out higher quality data from the process data to train the machine learning model, thereby reducing the impact of poor data sets on the machine learning model.
[0041] In some embodiments, in step S3, the machine learning model is trained using the best input data and historical self-discharge data to obtain a battery cell self-discharge prediction model, including the following steps: 1) Determine a sample data set and a test data set based on the best input data and the historical self-discharge data, and use the sample data set to perform multiple rounds of iterative training on the machine learning model, wherein the sample data set includes multiple samples, each sample includes a sample input obtained from the best input data and a true label obtained from the historical self-discharge data, and the true label represents the historical self-discharge K value corresponding to a sample input. 2) In each round of training, perform the following steps: Evaluate the model performance of the machine learning model based on the coefficient of determination and root mean square error of a sample, and optimize the model hyperparameters of the machine learning model using grid search based on the evaluation results, and continue the next round of training using the optimized machine learning model. 3) Evaluate the trained machine learning model using the test data set, and optimize the model hyperparameters of the machine learning model again based on the evaluation results to obtain a battery cell self-discharge prediction model.
[0042] In step 1), the sample data set and the test data set can be obtained by dividing the optimal input data and the historical self-discharge data in a certain ratio. For example, the combined data of the optimal input data and the historical self-discharge data are divided in a certain ratio, 80% of which is divided into the sample data set, and the remaining 20% is divided into the test data set. The sample data set is mainly used for multiple rounds of iterative training of the machine learning model, and the test data set is mainly used to evaluate the prediction effect of the model after the training is completed.
[0043] In step 2), the coefficient of determination can be calculated using the following formula:
[0044]
[0045] And, the root mean square error is calculated using the following formula:
[0046]
[0047] Among them, R 2 is the coefficient of determination, RMSE is the root mean square error, n is the number of samples, is the true label of the i-th sample, is the predicted value of the machine learning model for the i-th sample.
[0048] The above-mentioned determination coefficient and root mean square error can be used as indicators for evaluating model performance during the training process, so that the model can be optimized according to the indicators to ensure that the battery cell self-discharge prediction model has a certain reliable prediction ability.
[0049] In step 3), the trained machine learning model is evaluated again using the test data set, and the performance of the model on the training data set and the test data set is compared to optimize the model again, and the final battery cell self-discharge prediction model is obtained. The output of the battery cell self-discharge prediction model can be the self-discharge K value of the new lithium iron phosphate energy storage battery cell mentioned above.
[0050] The prediction effect of the above-mentioned battery cell self-discharge prediction model will be demonstrated below with reference to specific experimental examples.
[0051] Referring to step S1, the process data of the formation stage collected from the battery production line and the historical self-discharge data are used to construct a training data set, wherein the historical self-discharge data uses the self-discharge K value of the corresponding battery cell, and some of the data are shown in Table 1 below:
[0052]
[0053]
[0054] Table 1
[0055] Among them, the training data set corresponds to 18 formation features, including: Hi-pot resistance (abbreviated as Hp), cell weighing before the first injection (abbreviated as GoneB), cell weighing after the first injection (abbreviated as GoneA), the first injection volume (abbreviated as Gone), the liquid retention volume (abbreviated as Gkeep), the weighing before the second injection (abbreviated as GtwoB), the weighing after the second injection (abbreviated as GtwoA), the high temperature immersion time (abbreviated as Tjr), the formation liquid loss (abbreviated as Gloss), the formation charging capacity (abbreviated as AhForC), the divided capacity charging capacity (abbreviated as AhVolC), the divided capacity discharge capacity (abbreviated as AhVolF), the divided capacity discharge average voltage (abbreviated as VaveF), the divided capacity discharge average temperature (abbreviated as TemAveF), the first effect (abbreviated as Sx), the open circuit voltage (abbreviated as OCV1), the internal resistance (abbreviated as IR1), and the test temperature (abbreviated as Tem).
[0056] Further, referring to step S2, the aforementioned 18 formation characteristics are screened for correlation to obtain a relevant feature subset, and then the relevant feature subset is screened for mutual information to obtain an optimal feature subset. The optimal feature subset is shown in Table 2 below:
[0057] GtwoA Hp AhForC Gloss Gkeep Tj Sx AhVolF IR1 Tem
[0058] Table 2
[0059] Further, referring to step S3, the best input data in the process data is selected according to the best feature subset, and the machine learning model is trained using the best input data and historical self-discharge data to obtain a cell self-discharge prediction model, wherein some of the best input data are shown in Table 3 below:
[0060]
[0061]
[0062] Table 3
[0063] Further, referring to step S4, the current process data of the battery production line is input into the battery self-discharge prediction model, so that the battery self-discharge prediction model outputs a prediction result. The prediction result is as follows: Figure 2 Reference Figure 2 , where the horizontal axes of the upper and lower figures are the test values of the self-discharge K value of the battery cell, and the vertical axes are the predicted values of the self-discharge K value of the battery cell. The upper figure shows the results of the training data set, and the lower figure shows the results of the test data set. The determination coefficients of the self-discharge prediction model for the battery cell on the training data set and the test data set are 0.9477 and 0.8573, respectively, and the root mean square errors are 0.00079 and 0.00169, respectively.
[0064] The above experimental example designs a comparative example. The difference between the comparative example and the experimental example is that in step S3, the machine learning model is trained using the training data set that has not been through step S2 to obtain a corresponding comparative prediction model. The comparative prediction model uses the same current process data to obtain the prediction results as shown in FIG. Figure 3 Reference Figure 3 , where the horizontal axes of the upper and lower figures are the test values of the self-discharge K value of the battery cell, and the vertical axes are the predicted values of the self-discharge K value of the battery cell. The upper figure shows the results of the training data set, and the lower figure shows the results of the test data set. Under the premise that the division ratio of the training data set and the test data set remains unchanged, the determination coefficients of the comparative prediction model on the corresponding training data set and test data set are 0.9244 and 0.7679, respectively, and the root mean square errors are 0.00215 and 0.00215, respectively.
[0065] It can be seen from the prediction results of the above experimental examples and comparative examples that in step S2, multiple formation characteristics are screened for correlation to obtain a relevant feature subset, and then the relevant feature subset is screened for mutual information to obtain the best feature subset, which helps to improve the quality of data input to the model. That is, this method can optimize the formation characteristics and reduce data noise, so that the battery cell self-discharge prediction model has better prediction performance than the ordinary model.
[0066] refer to Figure 4 The present application provides a battery cell self-discharge prediction device, comprising: a data collection module 401, a feature screening module 402, a model training module 403 and a self-discharge prediction module 404, wherein:
[0067] The data collection module 401 is used to collect process data and historical self-discharge data of the formation stage from the battery production line, and use the process data to construct a training data set, wherein the formation stage includes multiple processes, and the training data set corresponds to multiple formation features set for the formation stage, and each formation feature is associated with a category of data that needs to be collected in the corresponding process.
[0068] The feature screening module 402 is used to perform correlation screening on multiple formation features to obtain a related feature subset, and then perform mutual information screening on the related feature subset to obtain an optimal feature subset.
[0069] The model training module 403 is used to select the best input data in the process data according to the best feature subset, and use the best input data and historical self-discharge data to train the machine learning model to obtain a battery cell self-discharge prediction model.
[0070] The self-discharge prediction module 404 is used to input the current process data of the battery cell production line into the battery cell self-discharge prediction model so that the battery cell self-discharge prediction model outputs a prediction result.
[0071] In some embodiments, a plurality of formation features are screened for correlation to obtain a subset of correlated features, and the feature screening module 402 is used to: calculate the linear correlation between the plurality of formation features using the Pearson correlation coefficient, and group the plurality of formation features according to a preset correlation threshold to determine at least one corresponding high-correlation feature group and at least one low-correlation feature group; retain each low-correlation feature group, retain one formation feature in each high-correlation feature group, and delete the remaining formation features to obtain a subset of correlated features.
[0072] In some embodiments, mutual information screening is performed on relevant feature subsets to obtain an optimal feature subset. The feature screening module 402 is used to: use the mutual information method to calculate the importance between each formation feature in the relevant feature subset and the target value; sort the relevant feature subsets in order of importance from large to small, retain the formation features with the top 80% of the corresponding importance, delete the remaining formation features, and obtain the optimal feature subset.
[0073] In some embodiments, the machine learning model is trained using the best input data and the historical self-discharge data to obtain a battery cell self-discharge prediction model, and the model training module 403 is used to: determine a sample data set and a test data set based on the best input data and the historical self-discharge data, and use the sample data set to perform multiple rounds of iterative training on the machine learning model, wherein the sample data set includes multiple samples, each sample includes a sample input obtained from the best input data and a true label obtained from the historical self-discharge data, and the true label represents the historical self-discharge K value corresponding to a sample input; in each round of training, the following steps are performed: the model performance of the machine learning model is evaluated based on the determination coefficient and the root mean square error of a sample, and based on the evaluation results, the model hyperparameters of the machine learning model are optimized using grid search, and the optimized machine learning model is used to continue the next round of training; the trained machine learning model is evaluated using the test data set, and the model hyperparameters of the machine learning model are optimized again based on the evaluation results to obtain a battery cell self-discharge prediction model.
[0074] Further, if Figure 5 An exemplary embodiment of the present application further provides an electronic device, including a memory 501 and a processor 502, wherein the memory 501 stores instructions, and when the instructions are executed by the processor 502, the processor 502 executes any method of the first aspect.
[0075] It should be understood that the processor mentioned in the embodiments of the present application may be a CPU, or other general-purpose processors, DSPs, ASICs, FPGAs or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
[0076] It should also be understood that the memory mentioned in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory, dynamic random access memory, synchronous dynamic random access memory, double data rate synchronous dynamic random access memory, enhanced synchronous dynamic random access memory, synchronously connected dynamic random access memory, and direct memory bus random access memory.
[0077] The present application also provides a storage medium storing computer program instructions, which, when executed by a processor of a computer, enables the computer to execute the steps of any of the methods mentioned above.
[0078] A computer-readable medium may include a propagated data signal containing computer program code, such as in baseband or as part of a carrier wave. The propagated signal may have a variety of manifestations, including electromagnetic, optical, etc., or a suitable combination. A computer-readable medium may be any computer-readable medium other than a computer-readable storage medium, which may be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit a program for use. The program code on the computer-readable medium may be propagated via any suitable medium, including radio, cable, fiber optic cable, radio frequency signal, or similar medium, or any combination of the above mediums.
[0079] The basic concepts have been described above. Obviously, for those skilled in the art, the above invention disclosure is only used as an example and does not constitute a limitation of the present application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements and amendments to the present application. Such modifications, improvements and amendments are suggested in the present application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary embodiments of the present application.
[0080] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.
[0081] Some aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". The processor may be one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DAPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, or combinations thereof. In addition, various aspects of the present application may be expressed as computer products located in one or more computer-readable media, which include computer-readable program codes. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, tapes ...), optical disks (e.g., compact disks CDs, digital versatile disks DVDs ...), smart cards, and flash memory devices (e.g., cards, sticks, key drives ...).
[0082] A computer-readable medium may include a propagated data signal containing computer program code, such as in baseband or as part of a carrier wave. The propagated signal may have a variety of manifestations, including electromagnetic, optical, etc., or a suitable combination. A computer-readable medium may be any computer-readable medium other than a computer-readable storage medium, which may be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit a program for use. The program code on the computer-readable medium may be propagated via any suitable medium, including radio, cable, fiber optic cable, radio frequency signal, or similar medium, or any combination of the above mediums.
[0083] Similarly, it should be noted that in order to simplify the description of the disclosure of this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the object of this application are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.
[0084] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of the present application are approximate values, in specific embodiments, the setting of such numerical values is as accurate as possible within the feasible range.
[0085] Although the present application has been described with reference to the current specific embodiments, ordinary technicians in this technical field should recognize that the above embodiments are only used to illustrate the present application, and various equivalent changes or substitutions may be made without departing from the spirit of the present application. Therefore, as long as the changes and modifications to the above embodiments are within the essential spirit of the present application, they will fall within the scope of the claims of the present application.
Claims
1. A method for predicting self-discharge of a battery cell, characterized in that: include: Collect process data and historical self-discharge data of the formation stage from the battery production line, and use the process data to construct a training data set, wherein the formation stage includes multiple processes, and the training data set corresponds to multiple formation features set for the formation stage, and each formation feature is associated with a category of data to be collected in the corresponding process; Performing correlation screening on the plurality of formation characteristics to obtain a correlated feature subset, and then performing mutual information screening on the correlated feature subset to obtain an optimal feature subset; Selecting the best input data from the process data according to the best feature subset, and using the best input data and the historical self-discharge data to train a machine learning model to obtain a cell self-discharge prediction model; and The current process data of the battery cell production line is input into the battery cell self-discharge prediction model so that the battery cell self-discharge prediction model outputs a prediction result.
2. The method according to claim 1, characterized in that The formation stage includes a liquid injection process and a high-temperature infiltration process, and the formation characteristics include any one or a combination of the following: Hi-pot resistance value; Weigh the battery cell before filling once; Weigh the battery cell after one injection; Amount of liquid injected at one time; Fluid retention; Weigh before secondary injection; Weigh after the second injection; High temperature soaking time.
3. The method according to claim 2, characterized in that The formation stage includes a formation process, a volume separation process and an OCV process, and the formation characteristics include any one or a combination of the following: The amount of fluid loss; Formation charging capacity; Differentiated charging capacity; Discharge capacity by volume; Average voltage of capacity-divided discharge; Average temperature of the discharge at different capacities; The first effect of volume division; OCV open circuit voltage; OCV internal resistance; OCV test temperature.
4. The method according to any one of claims 1 to 3, characterized in that: The correlation screening of the plurality of formation features to obtain a subset of related features includes: Calculating the linear correlation between the plurality of formation features using the Pearson correlation coefficient, and grouping the plurality of formation features according to a preset correlation threshold to determine corresponding at least one high-correlation feature group and at least one low-correlation feature group; Each low-correlation feature group is retained, and one formation feature is retained in each high-correlation feature group, and the remaining formation features are deleted to obtain the relevant feature subset.
5. The method according to any one of claims 1 to 3, characterized in that: The mutual information screening of the relevant feature subsets to obtain the best feature subset includes: Calculate the importance between each formation feature and the target value in the relevant feature subset using the mutual information method; The relevant feature subsets are sorted in descending order of importance, the formation features with the top 80% of the corresponding importance are retained, and the remaining formation features are deleted to obtain the best feature subset.
6. The method according to any one of claims 1 to 3, characterized in that: The using the optimal input data and the historical self-discharge data to train the machine learning model to obtain a cell self-discharge prediction model includes: Determine a sample data set and a test data set according to the optimal input data and the historical self-discharge data, and use the sample data set to perform multiple rounds of iterative training on the machine learning model, wherein the sample data set includes a plurality of samples, each sample includes a sample input obtained from the optimal input data and a true label obtained from the historical self-discharge data, and the true label represents the historical self-discharge K value corresponding to the sample input; In each round of training, the following steps are performed: the model performance of the machine learning model is evaluated according to the coefficient of determination and the root mean square error of a sample, and based on the evaluation result, the model hyperparameters of the machine learning model are optimized using grid search, and the optimized machine learning model is used to continue the next round of training; The trained machine learning model is evaluated using the test data set, and the model hyperparameters of the machine learning model are optimized again according to the evaluation results to obtain a battery cell self-discharge prediction model.
7. The method according to claim 6, characterized in that The determination coefficient is calculated using the following formula: And, the root mean square error is calculated using the following formula: Among them, R 2 is the determination coefficient, RMSE is the root mean square error, n is the number of samples, is the true label of the i-th sample, is the predicted value of the machine learning model for the i-th sample.
8. A battery cell self-discharge prediction device, characterized in that: include: A data collection module, used to collect process data and historical self-discharge data of the formation stage from the battery production line, and use the process data to construct a training data set, wherein the formation stage includes multiple processes, and the training data set corresponds to multiple formation features set for the formation stage, and each formation feature is associated with a category of data to be collected in the corresponding process; A feature screening module, used for performing correlation screening on the plurality of formation features to obtain a related feature subset, and then performing mutual information screening on the related feature subset to obtain an optimal feature subset; A model training module, used to select the best input data in the process data according to the best feature subset, and use the best input data and the historical self-discharge data to train the machine learning model to obtain a cell self-discharge prediction model; The self-discharge prediction module is used to input the current process data of the battery cell production line into the battery cell self-discharge prediction model so that the battery cell self-discharge prediction model outputs a prediction result.
9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores instructions, and when the instructions are called by the processor, the processor executes the battery cell self-discharge detection method as described in any one of claims 1-7.
10. A storage medium, characterized in that: The method comprises computer program instructions, wherein the computer program instructions can be called by a computer to implement the method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Lithium battery self-discharge detection method, device and system
CN111965545A
Prediction model generation method and self-discharge voltage drop prediction method
CN114371415A
Self-discharge monitoring model training method and device, electronic equipment and storage medium
CN114462194A
Battery self-discharge detection method, device and system
CN117054884A
Battery shell surface defect detection method and system based on deep learning
CN117110305A
Cited By
Battery cell abnormal self-discharge screening method and related equipment
CN121071521A
Battery cell abnormal self-discharge screening method and related device
CN121071521B
