Battery capacity prediction method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a battery capacity prediction method and device, electronic equipment and a storage medium. The method is applied to a battery capacity prediction system, and comprises the following steps: obtaining target feature data of a target battery cell in a battery manufacturing process, the target feature data being used for representing target feature data of parameter values of the battery cell in each process link in the battery manufacturing process; inputting the target characteristic data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell; the capacity prediction model is obtained by training first sample feature data of parameter values of each process link of the battery manufacturing process of the sample battery, the capacity prediction model is established based on ensemble learning, and the first sample feature data is training target sample feature data obtained by excluding abnormal data of the sample battery. By implementing the embodiment of the invention, the real capacity of the battery can be predicted in a low-economic-cost, low-time-cost and low-operation-and-maintenance-cost manner.

Description

Technical Field

[0001] The present application relates to the technical field of battery capacity prediction, and in particular to a battery capacity prediction method and device, electronic equipment, and storage medium. Background Art

[0002] With the popularity of renewable energy, the demand for lithium batteries in the energy storage market is also growing. The application of lithium batteries in the field of energy storage requires accurate capacity prediction to optimize the performance of the energy storage system. At present, the measurement methods for battery capacity mainly include constant current discharge method, constant power discharge method and deep discharge method. The basic idea of ​​these methods is to fully charge the battery first and then discharge it completely. However, these test methods are time-consuming and consume a lot of energy, especially for large-capacity power batteries. Battery manufacturers must be equipped with a large number of test equipment when producing power batteries, which occupies a large amount of factory area. In addition, in order to maintain the temperature of the test environment, air-conditioning equipment must be installed, which consumes a lot of electricity. Therefore, a new, non-manpower and capital-intensive test method is urgently needed to replace the existing traditional solutions. Summary of the invention

[0003] The embodiments of the present application disclose a battery capacity prediction method and device, an electronic device, and a storage medium, which can predict the actual capacity of a battery with low economic cost, low time cost, and low operation and maintenance cost, thereby ensuring the accuracy and reliability of the battery capacity prediction.

[0004] In a first aspect, an embodiment of the present application discloses a battery capacity prediction method, which is applied to a battery capacity prediction system. The method includes:

[0005] Acquire target characteristic data of a target battery cell in a battery manufacturing process, wherein the target characteristic data is used to characterize the target characteristic data of parameter values ​​of the battery cell in each process link in the battery manufacturing process;

[0006] Inputting the target characteristic data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell;

[0007] The capacity prediction model is obtained by training the first sample feature data of the parameter values ​​of the sample battery in each process link of the battery manufacturing process. The capacity prediction model is built based on ensemble learning. The first sample feature data is the training target sample feature data obtained after excluding the abnormal data of the sample battery.

[0008] As an optional implementation manner, in a first aspect of this embodiment, a method for obtaining the first sample characteristic data includes:

[0009] Obtaining sample characteristic data of parameter values ​​of the first sample battery cell in each process link in the battery manufacturing process, wherein the sample characteristic data of parameter values ​​in the process link at least includes: the weight of the sample battery cell, the weighing of the sample battery cell before and after the first injection of electrolyte and the first injection amount of electrolyte, the voltage and temperature at each stage of the formation process, and the formation capacity;

[0010] Screening the parameter value sample characteristic data to obtain preliminary screening sample characteristic data, wherein the preliminary screening sample characteristic data is within a preset data range;

[0011] An unsupervised algorithm is used to further screen the initial screening sample characteristic data of the sample battery cell to obtain the target sample characteristic data.

[0012] As an optional implementation manner, in the first aspect of this embodiment, obtaining target characteristic data of a target battery cell in a battery manufacturing process includes:

[0013] Obtaining characteristic data of parameter values ​​of the target battery cell in each process link in the battery manufacturing process, wherein the characteristic data of parameter values ​​in the process link at least includes: the weight of the target battery cell, the weighing of the target battery cell before and after the first injection of electrolyte and the first injection amount of electrolyte, the voltage and temperature at each stage of the formation process, and the formation capacity;

[0014] Screening the characteristic data of the parameter value to obtain preliminary screening characteristic data, wherein the preliminary screening characteristic data is within a preset data range;

[0015] An unsupervised algorithm is used to further screen the initial screening feature data of the target battery cell to obtain the target feature data.

[0016] As an optional implementation manner, in the first aspect of this embodiment, the method further includes:

[0017] Collect second sample characteristic data. When the data volume of the second sample characteristic data reaches a preset data volume value, input the second sample characteristic data into the capacity prediction model for training and updating, wherein the second sample characteristic data is sample characteristic data of parameter values ​​of updated sample cells in each process link of the battery manufacturing process.

[0018] As an optional implementation manner, in the first aspect of this embodiment, the method further includes:

[0019] When the continuous deviation between the predicted value and the true value of the capacity prediction model reaches a set threshold, the second sample feature data is input into the capacity prediction model for training and updating to obtain the capacity prediction model with updated model parameters.

[0020] As an optional implementation, in the first aspect of this embodiment, when there are multiple target battery cells, the target battery cells correspond to station identifications of battery cell production equipment, and the method further includes:

[0021] When the predicted capacity of the target battery cell is less than a preset first threshold, outputting the workstation identification information corresponding to the target battery cell; or,

[0022] Output comparison information of the target characteristic data of the first group of target battery cells and the second group of target battery cells, and identify significant differences, wherein the predicted capacity of the first group of target battery cells is lower than the preset first threshold, the predicted capacity of the second group of target battery cells is higher than the preset second threshold, and the second threshold is higher than or equal to the preset first threshold.

[0023] As an optional implementation, in the first aspect of this embodiment, the prediction model includes a first-layer base learner and a second-layer meta-learner, the first-layer base learner includes multiple regression models, and the target feature data is input into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell, including:

[0024] The first subset of the target feature data subsets is used as a test set, and the remaining subsets are used as training sets to train a first regression model; wherein the target sample feature data subsets are multiple independent subsets obtained by dividing the target feature data;

[0025] The second subset of the target feature data subset is used as a test set, and the remaining subsets are used as training sets to train the second regression model, and so on, to train the Nth regression model, where N is greater than or equal to 3 and less than or equal to the number of regression models included in the first layer base learner;

[0026] Perform three-fold cross validation on each trained regression model;

[0027] A preliminary prediction is performed on the target feature data subset as the test set after three-fold cross-validation. The preliminary prediction result is input as feature data to the second-layer meta-learner, and the output value of the second-layer meta-learner is the predicted battery capacity.

[0028] A second aspect of an embodiment of the present application discloses a battery capacity prediction device, the device comprising:

[0029] A data acquisition module, used to acquire target characteristic data of a target battery cell in a battery manufacturing process, wherein the target characteristic data is used to characterize the target characteristic data of parameter values ​​of the battery cell in each process link in the battery manufacturing process;

[0030] The capacity prediction module is used to input the target feature data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell.

[0031] A third aspect of an embodiment of the present application discloses an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements the method described above.

[0032] A fourth aspect of an embodiment of the present application discloses a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method described above when executed by a processor.

[0033] Compared with the related art, the embodiments of the present application include at least the following beneficial effects:

[0034] The embodiments of the present application disclose a battery capacity prediction method and device, electronic device, and storage medium. By acquiring the target feature data of the target battery cell in the battery manufacturing process and inputting the target feature data into a pre-trained capacity prediction model, the prediction result of the battery capacity of the target battery cell can be obtained. Since the present application can predict the capacity of the battery based on the battery production data and in combination with the capacity prediction model, there is no need to perform charge and discharge experiments on all batteries, so that the actual capacity of the battery can be predicted with low economic cost, low time cost, and low operation and maintenance cost, ensuring the accuracy and reliability of the battery capacity prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 A flowchart of a battery capacity prediction method in one embodiment;

[0037] Figure 2 is a flowchart of a method for identifying abnormal data in one embodiment;

[0038] Figure 3 A technical roadmap for abnormal data identification in one embodiment;

[0039] Figure 4 A flow chart of a battery capacity prediction model update method in one embodiment;

[0040] Figure 5 A flow chart of a battery capacity prediction model update method in another embodiment;

[0041] Figure 6 A schematic diagram of battery capacity prediction result statistics in one embodiment;

[0042] Figure 7 A schematic diagram of a battery capacity prediction model in an embodiment;

[0043] Figure 8 is a block diagram of a battery capacity prediction device in one embodiment;

[0044] Fig. 9 FIG. 4 is a structural block diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0046] In order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first" and "second" are used to distinguish the same items or similar items with substantially the same functions and effects. For example, the first instruction and the second instruction are to distinguish different user instructions, and their order is not limited. Those skilled in the art can understand that the words "first" and "second" do not limit the quantity and execution order, and the words "first" and "second" do not necessarily limit them to be different.

[0047] It should be noted that, in this application, words such as "exemplarily" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplarily" or "for example" in this application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplarily" or "for example" is intended to present related concepts in a specific way.

[0048] In addition, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can mean: a, or b, or c, or a and b, or a and c, or b and c, or a, b and c, where a, b, c can be single or multiple.

[0049] In addition, the terms "including" and "having" and any variations thereof in the embodiments of the present application and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products, or devices.

[0050] As the country's requirements for greening the industry continue to increase and a good development environment is provided, the new energy industry has developed rapidly. Emerging energy supply methods such as lithium batteries and photovoltaics have gradually replaced traditional fuel supply methods such as coal and oil. In particular, the current market sales rate of new energy vehicles has exceeded 30%. Lithium-ion batteries have become the mainstream power batteries in the market due to their high energy density, long cycle life, and reliable safety. In addition, lithium-ion batteries have also made significant progress in portable devices and energy storage.

[0051] At present, there are many indicators used to evaluate battery quality, among which battery capacity is an important criterion. Battery groups are paired to form battery packs, and battery packs are further paired to form battery modules. The size of the battery capacity is directly related to the length of time the battery can be used after a standard charge. When the single cells are assembled into a battery pack, the capacity of the single cells is crucial to the discharge capacity and service life of the battery pack. Therefore, screening the electrochemical consistency of the battery on the basis of consistent capacity is the key to making high-quality battery packs. Therefore, whether in the production process or in the market circulation and sales process, lithium-ion batteries need to pay attention to, test and supervise the capacity of the battery.

[0052] In the related technology, the measurement methods for battery capacity mainly include constant current discharge method, constant power discharge method and deep discharge method, etc. The traditional method of capacity measurement is to test the capacity of the battery by full charge and discharge experimental method. This test method requires all batteries to be charged and discharged, which means that this test scheme has the disadvantages of high energy consumption, high duration, and high equipment occupancy rate.

[0053] The embodiments of the present application disclose a battery capacity prediction method and device, electronic device, and storage medium, which can predict the actual capacity of a battery with low economic cost, low time cost, and low operation and maintenance cost, thereby ensuring the accuracy and reliability of the battery capacity prediction. The following are detailed descriptions.

[0054] See also Figure 1 , Figure 1 Flow chart of a battery capacity prediction method in one embodiment, which is applied to a battery capacity prediction system, such as Figure 1 As shown, the method may include the following steps:

[0055] Step 101 , obtaining target characteristic data of a target battery cell in a battery manufacturing process, wherein the target characteristic data is used to characterize the target characteristic data of parameter values ​​of the battery cell in each process link in the battery manufacturing process.

[0056] The battery cell is the smallest energy storage unit in the battery system. It is a basic unit composed of positive electrode materials, negative electrode materials, electrolytes and diaphragms. The main function of the battery cell is to store and release electrical energy. It is the basis of battery modules and battery packs. The performance of the battery cell directly affects the energy density, charge and discharge cycle life, power density and safety performance of the entire battery system. In a battery module or battery pack, multiple batteries can be combined in series (increase voltage) or in parallel (increase capacity) to meet the required voltage and capacity requirements. Therefore, the total capacity of the battery system is the sum of the capacities of all batteries. In the battery manufacturing process, the current host computer system saves the timing data of each process of a single battery cell as a CSV file, including winding, injection, formation, SOC and OCV process segments. The host computer system is usually a computer system in an industrial control system.

[0057] In some embodiments, the timing data of each process of a single battery cell may include the winding speed, winding tension, pole piece alignment, and temperature of the winding machine in the winding process section; and the injection time, injection amount, electrolyte type and concentration, battery cell weight, battery cell weight before and after injection, and injection environment temperature in the injection process section; and may also include the formation charge and discharge cycle number, charging voltage and current, temperature during the formation process, battery cell capacity after formation, voltage and current changes during the formation process, etc. SOC is the ratio of the actual charge amount of the battery to the rated capacity, that is, the ratio of the amount of electricity that can be provided by discharging to the termination voltage at the rated current in the current state to the rated capacity. The timing data of the SOC process section includes real-time SOC value, SOC change trend, SOC record during charging and discharging, SOC calibration data, etc. OCV represents open circuit voltage, that is, the voltage between the positive and negative electrodes of the battery when there is no load. OCV timing data may include OCV test time, OCV value under different SOC, temperature and humidity of the OCV test environment, and battery cell state after OCV test. The parameter values ​​of the battery cell in each process link in the battery manufacturing process may include at least one of the above-mentioned time series data, which is not specifically limited here.

[0058] In an embodiment of the present application, the target feature data is the time series data saved in each process of battery production. After feature extraction, the missing data is interpolated or deleted according to the actual situation. All dimensions of the processed feature data can represent the data representation of a battery cell in each production process. The purpose of feature extraction is to filter out features that have an important impact on battery performance, production process control, etc. from the original data, and provide a basis for subsequent data analysis and model establishment. In some embodiments, there are many methods for feature extraction. For example, features can be extracted by calculating the statistics of time series data, such as mean, median, variance, standard deviation, minimum and maximum values, etc., and mathematical models can be used to fit the time series data, and the coefficients or parameters of the model are used as features. The dynamic change characteristics of the data in the battery production process can be captured by fitting the mathematical model. In some embodiments, statistical methods can be used to fill in the missing data, such as linear interpolation, time series prediction models, etc., to ensure that the time series data after interpolation filling or deletion operations can still accurately reflect the production process of the battery cell.

[0059] Step 102, input the target feature data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell, wherein the capacity prediction model is obtained by training the first sample feature data of the parameter values ​​of the sample battery in each process link of the battery manufacturing process, and the capacity prediction model is built based on ensemble learning, and the first sample feature data is the training target sample feature data obtained after excluding abnormal data of the sample battery.

[0060] In some embodiments, the sample battery generally refers to a portion of batteries extracted from a battery production batch, which is used to represent the entire batch for various tests and analyses. The data of the sample battery is used to train the capacity prediction model. In addition to lithium-ion batteries, the sample battery can also be other types of batteries such as nickel-metal hydride batteries, nickel-chromium batteries, and dry batteries. The sample feature data is the feature data of the time series data saved in each process of the sample battery production after feature extraction and interpolation filling or deletion. The sample feature data is further eliminated by the abnormal data recognition system to obtain the first sample feature data for training the capacity prediction model.

[0061] In the actual battery cell production scenario, unqualified products may be produced in each link. For example, in the battery cell injection process, due to reasons such as the injection equipment, the amount of electrolyte introduced into a very small number of battery cells may not meet the process requirements. These battery cells usually need to undergo measures such as artificial rehydration to make them qualified and enter the next process. However, the data from manual operations cannot be accurately synchronized to the system in real time, resulting in a large deviation between the recorded data of these battery cells and the actual situation. In order to prevent these abnormal data from misleading the training of the capacity prediction model, such data needs to be eliminated during the model training process. In the subsequent real-time prediction of battery capacity, such data will not be involved in the capacity prediction, but the entire data will be directly marked as abnormal.

[0062] In some embodiments, the capacity prediction model can be built based on ensemble learning. When predicting the battery capacity, the ensemble learning model can improve the accuracy and robustness of the prediction by combining the prediction results of multiple base learners. These models can be based on different algorithms, such as Bagging, Boosting, and Stacking, and use different base learners, such as decision trees, support vector machines, or neural networks. In addition, the ensemble learning model can also improve the prediction performance by combining deep learning techniques. For example, the capacity data of the battery is modeled by using a long short-term memory network (LSTM) and a deep belief network (DBN), and the outputs of multiple models are integrated to obtain more accurate prediction results.

[0063] By adopting the above-mentioned embodiment, the capacity of the battery can be predicted based on the battery production data and in combination with the capacity prediction model, so there is no need to conduct charge and discharge experiments on all batteries, so that the real capacity of the battery can be predicted with low economic cost, low time cost and low operation and maintenance cost, ensuring the accuracy and reliability of the battery capacity prediction. In addition, the battery capacity prediction system in the embodiment of the present application has an unsupervised abnormal data detection function, which can identify normal data patterns according to continuous production conditions, and identify and eliminate abnormal data with rework anomalies or record deviations, thereby avoiding the negative impact of abnormal data on model training and ensuring the accuracy and reliability of the model.

[0064] Since both sample feature data and target feature data need to be excluded from abnormal data before they can be used for subsequent model training and capacity prediction, the abnormal data identification system and abnormal data identification process in the battery capacity prediction system will be introduced below.

[0065] See also Figure 2 , Figure 2 FIG. 1 is a flow chart of a method for identifying abnormal data in one embodiment. Figure 2 As shown, taking abnormal data identification of sample feature data as an example, the method may include the following steps:

[0066] Step 201, obtaining sample characteristic data of parameter values ​​of the first sample battery cell in each process link in the battery manufacturing process, wherein the sample characteristic data of parameter values ​​in the process link at least includes: sample battery cell weight, sample battery cell weighing before and after the first injection of electrolyte and the first electrolyte injection volume, voltage and temperature at each stage of the formation process, and formation capacity.

[0067] In an embodiment of the present application, the sample feature data of the first sample battery cell after the abnormalities are eliminated is the initial training data set of the capacity prediction model. Optionally, in the subsequent model update stage, the second sample battery cell, the third sample battery cell, etc. for model update can also be obtained, and their corresponding sample feature data are subjected to the same abnormal data identification system to eliminate abnormal data. The types of parameter value sample feature data of the first sample battery cell in each process link in the battery manufacturing process are not repeated here. It should be noted that the label data in the initial training data set is the battery capacity of the first sample battery cell measured by experimental means. Each segment of sample feature data of the first sample battery cell will have a corresponding battery capacity value as the corresponding data label, which is used for model training in the supervised learning process.

[0068] Step 202, screening the parameter value sample characteristic data to obtain preliminary screening sample characteristic data, the preliminary screening sample characteristic data is within a preset data range.

[0069] In some embodiments, screening rules can be set according to battery design and performance requirements to identify abnormal data with large deviations from normal data, and data that do not meet the preset range are eliminated, and those data within the set range are retained as sample feature data for preliminary screening. As an optional embodiment, in the embodiment of the present application, the 3Sigma (mean ± 3 times standard deviation) principle can be determined according to actual data and test results to screen the key features such as sample cell weight, sample cell weight before and after the first injection of electrolyte, and the first electrolyte injection volume, voltage and temperature at each stage in the formation process, and formation capacity, and data beyond the range are considered abnormal and excluded from the capacity prediction model. Specifically, the mean and standard deviation are calculated for each type of sample feature data, and then the principle of mean ± 3 times standard deviation is used to determine the acceptance range of each data. In other embodiments, the range screening of abnormal data can also adopt a distance-based method, such as K nearest neighbor, by calculating the average distance between each data point and its K nearest neighbor points, if the average distance of a data point is much larger than other data points, it is considered as an outlier. In addition, the Z-score method can be used to identify outliers by calculating the distance of each data point from the mean (in units of standard deviation). If the absolute value of the Z-score of a data point is greater than the set threshold (such as 3), it is considered an outlier.

[0070] Step 203: Use an unsupervised algorithm to further screen the initial screening sample characteristic data of the sample battery cells to obtain target sample characteristic data.

[0071] In some embodiments, the abnormal data identification system includes an abnormal data detection model, and an unsupervised algorithm can be used to implement abnormal data detection. Unsupervised algorithms are a type of learning method without pre-labeled or classified data. They discover anomalies or regularities in the data by exploring the inherent structure and pattern of the data. The abnormal data detection model is a specific unsupervised learning model that is specifically used to identify anomalies or outliers in a data set. The target sample feature data obtained after further eliminating abnormal data through the abnormal data detection model is the training data set of the capacity prediction model.

[0072] In an embodiment of the present application, the feature data after range screening is input into the abnormal data detection model to further identify abnormal data. As an optional implementation, the abnormal data detection model can be an isolation forest algorithm. The basic idea of ​​the isolation forest is that the number of outliers (outliers) is relatively small, and they are usually sparsely distributed in the data set, so they are easier to be "isolated". The algorithm constructs multiple isolation trees, each of which recursively splits the data by randomly selecting features and split points until each data point is isolated in its own node. For outliers, since they are relatively rare in the data, they usually require fewer splitting steps to be isolated, so they will reach the leaf node earlier in the tree. When the abnormal data detection model identifies the abnormal data, it further outputs an abnormal data tag to mark the entire data. For the marked data, no subsequent capacity prediction will be performed, nor will it participate in the training and updating of the capacity prediction model. When marking outliers, a new column or label can be added to the data set to indicate whether each data point is marked as abnormal.

[0073] In the battery capacity prediction stage, the method flow for identifying and eliminating abnormal data of the characteristic data of the target battery cell is consistent with the above embodiment and will not be repeated here.

[0074] Figure 3 The technical roadmap for abnormal data identification in the above embodiments is as follows: Figure 3 As shown, firstly, the sample feature data or the target feature data is range-screened to exclude the data outside the range, and then the data within the range is input into the abnormal data detection model to further identify and exclude the abnormal data, and finally the obtained normal data is input into the capacity prediction model for model training or battery capacity prediction.

[0075] By adopting the above-mentioned embodiment, the abnormal data can be identified and processed by the abnormal data identification system, so as to improve the overall quality of the data set, thereby providing more accurate input data for the capacity prediction model and improving the prediction ability and accuracy of the capacity prediction model. This process not only improves the prediction accuracy, but also optimizes the production efficiency and the reliability of data management.

[0076] In the actual battery capacity prediction process, if the input feature data changes, especially when the new data has a significant difference in distribution from the data used to train the model before, or if the application scenario or task requirements of the model change, the capacity prediction model may need to be updated to adapt to these changes, such as adjusting the structure or parameters of the model to meet the new requirements, which helps maintain the accuracy and effectiveness of the model. Therefore, the following will explain how to update the capacity prediction model.

[0077] See also Figure 4 , Figure 4FIG. 1 is a flow chart of a battery capacity prediction model update method in one embodiment. Figure 4 As shown in the figure, when the capacity prediction system detects that the input feature data has changed, such as the product specifications of the battery cell have changed, the capacity prediction model is updated. The update method is as follows:

[0078] The second sample characteristic data is collected. When the data volume of the second sample characteristic data reaches a preset data volume value, the second sample characteristic data is input into the capacity prediction model for training and updating, wherein the second sample characteristic data is the sample characteristic data of the parameter values ​​of the updated sample battery cells in each process link of the battery manufacturing process.

[0079] In some embodiments, the second sample feature data may be sample feature data corresponding to a sample battery cell with the same or different specifications as the first sample battery cell, and the specifications of the battery cell product include size, capacity, voltage, internal resistance, charging current, discharging current, operating temperature range, etc. When the specifications of the first sample battery cell and the second sample battery cell are the same, only 10% of the batteries in the production process are retained for the same type of battery cell, and the capacity values ​​are still measured by the experimental scheme as the data labels required for updating the prediction model, which are used to update the capacity prediction model, and the remaining 90% of the batteries can directly predict their capacity values ​​through the capacity model. The process of obtaining the second sample feature data and handling abnormal data is the same as that of the first sample feature data, and will not be repeated here. As an optional implementation, the preset value of the data volume of the second sample feature data can be 1000 data, 2000 data or other values, which are not specifically limited here. If the amount of data used to update the prediction model reaches the preset value, the second sample feature data is input into the capacity prediction model for training and updating to obtain an updated capacity prediction model. If the amount of data used to update the prediction model does not reach the preset value, the update data continues to be collected. The update of the model can be the adjustment of model parameters, such as learning rate, regularization coefficient, number of network layers, number of nodes, etc., or the update of model optimization algorithms, such as Adam, RMSprop, etc., using the updated data to retrain or fine-tune the model. The model can learn new feature data and adapt to the new distribution to adapt to new application scenarios or task requirements.

[0080] The embodiment of the present application also provides a prediction method for other products with the same principle but different specifications. When the product specifications are switched, as long as a certain amount of production data is accumulated, the real-time capacity prediction can be carried out online after the model training is completed, thus meeting the requirements of high efficiency and intelligence in modern battery production.

[0081] See also Figure 5 , Figure 5 FIG. 1 is a flow chart of a battery capacity prediction model update method in another embodiment. Figure 5As shown in the figure, when the capacity prediction system detects that the prediction accuracy of the capacity prediction model is deviated and the continuous deviation reaches the set threshold, the capacity prediction model is updated. The update method is as follows:

[0082] When the continuous deviation between the predicted value and the true value of the capacity prediction model reaches a set threshold, the second sample feature data is input into the capacity prediction model for training and updating to obtain the capacity prediction model with updated model parameters.

[0083] In some embodiments, due to factors such as the complexity of the data and the limitations of the model, the prediction accuracy of the model may be affected and a large continuous deviation may occur. To solve this problem, in the embodiment of the present application, it is determined whether the model needs to be updated based on the current model prediction deviation and trend. When the continuous deviation of the model's prediction accuracy reaches a set threshold (such as 10 times), the capacity prediction model will start the self-training function. Optionally, the prediction accuracy deviation of the model can be measured by using an experimental scheme for 10% of the cells in the production process to obtain the capacity value as the true value, and the difference with the model prediction value is calculated to obtain the prediction deviation result of the model. The second sample feature data is the sample feature data corresponding to the current time node when the continuous deviation of the model prediction reaches the set threshold and the time node when the deviation occurs, and the real-time production of the cell data during this time period. As an optional implementation method, the prediction accuracy of the model can be continuously monitored during the operation of the model. When the continuous deviation of the prediction accuracy reaches the set threshold, it indicates that the prediction performance of the model has decreased, and the second sample feature data is used to train and update the capacity prediction model to improve the prediction performance of the model.

[0084] In some other embodiments, the automatic update method of the capacity prediction model can also be: when the running time of the prediction model reaches a preset time or a specific time node is set to trigger the training update of the prediction model. The automatic update of the model can be triggered when the number of second sample feature data reaches a preset value, or the number of consecutive deviations of the model prediction accuracy reaches a set threshold, and the running time of the set model reaches a specific time period. As long as one of the above three conditions is met, the automatic update of the model can be triggered.

[0085] In some embodiments, in addition to automatic updating, the capacity prediction model can also be updated manually. Manual updating means updating the model parameters after the battery product specifications change or due to subjective needs. Manual model update means that the model update is triggered manually. When the capacity prediction model receives a manually triggered update instruction, the second sample feature data is collected to train and update the model.

[0086] By adopting the above-mentioned embodiment, through automatic and manual updating of the model, the model can adapt to new data distribution and characteristics, maintain high prediction accuracy, and the activation of the model self-training function can timely discover and solve the problem of reduced model prediction performance, thereby improving the stability of the model.

[0087] The battery capacity prediction system in the embodiment of the present application supports outputting production statistics, performing statistics on the distribution interval of the current battery cell capacity, and outputting statistical results. The process diagram can be found in Figure 6 .like Figure 6 As shown, the process of the battery capacity prediction system outputting capacity prediction result statistics is as follows:

[0088] When there are multiple target battery cells to be predicted, the capacity prediction results of the target battery cells are divided into two groups, and the target battery cells correspond to the workstation identifications of the battery cell production equipment. The predicted capacity of the first group of target battery cells is lower than the preset first threshold. When the predicted capacity of the target battery cells is lower than the preset first threshold, the workstation identification information corresponding to the target battery cells is output, and the target characteristic data of the first group of target battery cells is output. The predicted capacity of the second group of target battery cells is higher than the preset second threshold, and the target characteristic data of the second group of target battery cells is output. Afterwards, the comparison information of the target characteristic data of the first group of target battery cells and the second group of target battery cells is output to identify significant differences, wherein the preset second threshold is higher than or equal to the preset first threshold.

[0089] In some embodiments, the battery cell will be bound to the corresponding production equipment number and work station number. The battery capacity prediction system will perform equipment-side work station statistics on the battery cells whose predicted capacity values ​​are lower than the preset first threshold value, so as to timely check the reduction in the quality of the battery cell products caused by the damage of individual battery cell production stations. The first threshold and the second threshold can be set based on historical data, production experience and product specifications, and are not specifically limited here. The first threshold and the second threshold can be used to distinguish different levels of battery capacity. Batteries with a capacity lower than the first threshold are low-capacity batteries, and batteries with a capacity higher than the second threshold are high-capacity batteries. For high-capacity batteries, their capacity may be close to or equal to the nominal capacity of the battery cell. For high-capacity and low-capacity batteries, the comparison information of the target feature data of the two groups of target batteries is further output. Through comparison, the differences between the two groups of batteries in various feature dimensions can be clearly seen. At the same time, the significant difference points are output in combination with the feature importance of the model. These significant difference points can provide important reference basis for subsequent production improvements, quality control, etc., for reference and real-time adjustment by on-site personnel. For example, if it is found that the internal resistance of the first group of battery cells is generally higher than that of the second group of battery cells, then it may be necessary to adjust a certain link in the production process to reduce the internal resistance and increase the battery cell capacity.

[0090] In some embodiments, statistical analysis can be performed to determine the importance of model features. For example, the correlation coefficient between each feature and the cell capacity is calculated, and the stronger the correlation, the more important the feature. Or a variance analysis is performed to test whether the influence of different features on the cell capacity is significant. For the first group of low-capacity cells and the second group of high-capacity cells, the importance of their features is analyzed separately, and which features have significantly different degrees of importance in the two groups are compared. For example, if a feature is highly important in the second group of high-capacity cells, but less important in the first group of low-capacity cells, then this feature may be an important factor leading to the difference in capacity between the two groups of cells. The capacity prediction system determines significant difference points based on the comparison results of the feature importance. These significant difference points can be a single feature or a combination of multiple features. For example, if it is found that the internal resistance feature importance of the first group of cells is significantly higher than that of the second group, and the charging time feature importance is significantly lower than that of the second group, then the internal resistance and charging time may be significant difference points.

[0091] By adopting the above embodiment, by grouping the target cells according to the capacity prediction results and outputting the station identification information of the low-capacity group, the production links where problems may occur can be quickly and accurately located. This enables production managers to take timely measures to inspect and adjust specific stations to improve production efficiency and product quality. In addition, by combining the output of significant difference points based on the importance of model features, the root causes of the differences between cells in different capacity groups can be deeply analyzed, which helps to solve the problem in a targeted manner, avoid blindly adjusting the production process, and reduce the cost and time of troubleshooting.

[0092] See also Figure 7 , Figure 7 A flowchart for constructing a battery capacity prediction model in one embodiment is shown below. Figure 7 As shown, the battery capacity prediction model is constructed by an integrated learning model. The prediction model includes a first-layer base learner and a second-layer meta-learner. The first-layer base learner includes multiple regression models. The method for constructing a capacity prediction model includes the following steps:

[0093] Step 701, using the first subset of the target feature data subsets as a test set and the remaining subsets as training sets to train a first regression model; wherein the target sample feature data subsets are multiple independent subsets obtained by dividing the target feature data;

[0094] In some embodiments, the target feature data is a data set used to train and test the model, which includes multiple features related to battery capacity, which will not be described in detail here. A specific partitioning method is used to divide the target feature data into multiple independent subsets. The partitioning method can be selected according to the characteristics and needs of the data, such as random partitioning, hierarchical partitioning, etc., and the first subset is determined as the test set, and the remaining subsets are combined as training sets. For example, the target feature data can be divided into K subsets of equal size, of which K-1 subsets are used for training and the remaining subset is used for testing. This process will be performed K times, each time using a different subset as the test set, and the average result of K times is used as the evaluation standard for model performance. This method can more comprehensively evaluate the generalization ability of the model.

[0095] The training set data is used to train the regression model of the first layer base learner. The goal of the regression model is to find the mapping relationship between the feature data and the battery capacity so that the battery capacity can be predicted based on the input feature data. Optionally, the regression model can be a linear regression model, a polynomial regression model, a random forest model, a decision tree model, a support vector machine model, etc.

[0096] Step 702, using the second subset of the target feature data subset as a test set and the remaining subsets as training sets to train the second regression model, and so on, to train the Nth regression model, where N is greater than or equal to 3 and less than or equal to the number of regression models included in the first layer base learner;

[0097] In some embodiments, the second subset is selected as a test set, that is, when training the second regression model, the data of the second subset will not participate in the training process of the model, but will be used to evaluate its performance after the model training is completed. The remaining subsets except the second subset are used as training sets, and the data of these subsets will be used to train the second regression model. By adjusting the parameters of the model, the model can fit the characteristics and laws of the training set data as accurately as possible. By analogy, for the Nth regression model (where N is greater than or equal to 3 and less than or equal to the number of regression models contained in the first layer base learner), the above method is also followed. The Nth subset is selected as a test set, and all the remaining subsets (excluding the Nth subset) are used as training sets to train the Nth regression model. When the value of N is large, the model performance evaluation deviation caused by the randomness of data partitioning can be reduced. By training multiple regression models and evaluating them on different test sets, more stable performance indicators can be obtained, thereby better selecting the optimal model or performing model fusion. Combining the results of these models can reduce the uncertainty of a single model and improve the stability of the overall prediction.

[0098] Step 703, performing three-fold cross validation on each trained regression model;

[0099] In some embodiments, three-fold cross validation is a method for evaluating and comparing the performance of machine learning models. It divides the data set into three parts, and then performs three training and validation processes, each time selecting a different subset as the validation set, and the remaining subset as the training set. Optionally, in each iterative training and validation process, performance evaluation indicators such as mean square error, precision, and recall rate can be used to evaluate model performance, and finally the average of the model performance indicators in each iteration is calculated to obtain an overall performance estimate of the model. By repeating training and testing on different training sets and validation sets, the performance of the model on unknown data can be more accurately estimated, and the variance of the model evaluation results can be reduced through multiple iterations, making the evaluation results more stable.

[0100] In some other embodiments, multi-fold cross validation can also be selected, such as 5-fold, 10-fold, etc., depending on the size of the data set and the complexity of the model. For example, for smaller data sets, more folds can be used, while for larger data sets, fewer folds may be more appropriate. The larger the fold, the larger the amount of data in each training set, the smaller the deviation of the model, but the computational cost will also increase accordingly. In the integrated learning model in the embodiment of the present application, the performance of the model can be comprehensively evaluated by three-fold cross validation, and the base learner with the best performance can be selected for integration.

[0101] Step 704, a preliminary prediction is made on the target feature data subset as the test set, using the regression model after three-fold cross validation, and the preliminary prediction result is input as feature data to the second-layer meta-learner, and the output value of the second-layer meta-learner is the predicted battery capacity.

[0102] In some embodiments, each regression model after three-fold cross validation is used to predict the corresponding test set, and the prediction results generated by all base learners are summarized to form a new data set, and the labels of the prediction results are combined into new feature data and input into the second-layer meta-learner for training, and the prediction value of the final meta-model is used as the final output. In the subsequent real-time battery capacity prediction process, for new or unknown target feature data, the base learner of the first layer is first used to predict it to generate secondary features, and then these secondary features are input into the meta-learner to obtain the final prediction result, that is, the predicted battery capacity. Optionally, the meta-learner can be any type of machine learning model, including but not limited to linear regression, ridge regression, support vector machine, decision tree, random forest or neural network. By combining the prediction results of multiple regression models, the meta-learner can reduce the possible deviations of a single model and improve the accuracy of the prediction results.

[0103] By using the above embodiment, the capacity prediction model built based on the ensemble learning model can combine multiple different regression models to improve the overall battery capacity prediction performance. This method can reduce the deviation and variance that may exist in a single model, thereby improving the accuracy and robustness of battery capacity prediction.

[0104] See also Figure 8 , Figure 8 8 is a block diagram of a battery capacity prediction device in one embodiment. The battery capacity prediction device 800 may include a data acquisition module 810 and a prediction module 820 .

[0105] The data acquisition module 810 is used to acquire target characteristic data of the target battery cell in the battery manufacturing process, where the target characteristic data is used to characterize the target characteristic data of the parameter values ​​of the battery cell in each process link in the battery manufacturing process;

[0106] The capacity prediction module 820 is used to input the target feature data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell.

[0107] In some embodiments, the data acquisition module 810 is also used to obtain sample characteristic data of parameter values ​​of the first sample battery cell in each process link in the battery manufacturing process. The sample characteristic data of parameter values ​​in the process link include at least: the weight of the sample battery cell, the weighing of the sample battery cell before and after the first injection of electrolyte and the amount of the first electrolyte injection, the voltage and temperature at each stage of the formation process, and the formation capacity.

[0108] The data acquisition module 810 is further used for:

[0109] Obtain characteristic data of parameter values ​​of the target battery cell in each process link in the battery manufacturing process, wherein the characteristic data of parameter values ​​in the process link at least include: the weight of the target battery cell, the weighing of the target battery cell before and after the first injection of electrolyte and the first injection amount of electrolyte, the voltage and temperature at each stage of the formation process, and the formation capacity;

[0110] The second sample characteristic data is obtained, where the second sample characteristic data is sample characteristic data of parameter values ​​of the updated sample battery cell in each process link of the battery manufacturing process.

[0111] In some embodiments, the capacity prediction module 820 is further used to input the second sample feature data into the capacity prediction model for training and updating when the data volume of the second sample feature data reaches a preset data volume value.

[0112] The capacity prediction module 820 is further configured to:

[0113] When the continuous deviation between the predicted value and the true value of the capacity prediction model reaches a set threshold, the second sample feature data is input into the capacity prediction model for training and updating, so as to obtain the capacity prediction model after the model parameters are updated;

[0114] When the predicted capacity of the target battery cell is less than a preset first threshold, outputting the workstation identification information corresponding to the target battery cell; or,

[0115] Output comparison information of target characteristic data of the first group of target cells and the second group of target cells, and identify significant differences, wherein the predicted capacity of the first group of target cells is lower than a preset first threshold, the predicted capacity of the second group of target cells is higher than a preset second threshold, and the second threshold is higher than or equal to the preset first threshold.

[0116] In some embodiments, the capacity prediction module 820 is further configured to:

[0117] A preliminary prediction is made on the regression model after three-fold cross-validation for the target feature data subset input as the test set, and the preliminary prediction result is input as feature data to the second-layer meta-learner. The output value of the second-layer meta-learner is the predicted battery capacity.

[0118] In the embodiment of the present application, by obtaining the target characteristic data of the target battery cell in the battery manufacturing process and inputting the target characteristic data into the pre-trained capacity prediction model, the prediction result of the battery capacity of the target battery cell can be obtained. Since the present application can predict the capacity of the battery based on the battery production data and in combination with the capacity prediction model, there is no need to perform charge and discharge experiments on all batteries, so that the real capacity of the battery can be predicted with low economic cost, low time cost and low operation and maintenance cost, ensuring the accuracy and reliability of the battery capacity prediction.

[0119] Fig. 9 FIG. 1 is a structural block diagram of an electronic device in an embodiment. Fig. 9 As shown, the terminal electronic device 9 may include: a memory 910 storing executable program codes and a processor 920 coupled to the memory 910 .

[0120] The processor 920 calls the executable program code stored in the memory 910 to execute any one of the battery capacity prediction methods disclosed in the embodiments of the present application. Those skilled in the art can understand that Fig. 9 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0121] The processor 920 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. It executes various functions of the electronic device and processes data by running or executing software programs and / or modules stored in the memory 910, and calling data stored in the memory 910, so as to monitor the electronic device as a whole. Optionally, the processor 920 may include one or more processing units; preferably, the processor 920 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 920.

[0122] The memory 910 can be used to store software programs and modules, and the processor 920 executes various functional applications and data processing of the electronic device by running the software programs and modules stored in the memory 910. The memory 910 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 910 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0123] An embodiment of the present application discloses a computer-readable storage medium storing a computer program, wherein the computer program implements the methods described in the above embodiments when executed by a processor.

[0124] An embodiment of the present application discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program can be executed by a processor to implement the methods described in the above embodiments.

[0125] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a ROM, etc.

[0126] As used herein, any reference to memory, storage, database, or other medium may include nonvolatile and / or volatile memory. Suitable nonvolatile memory may include ROM, programmable ROM (Programmable ROM, PROM), erasable PROM (Erasable PROM, EPROM), electrically erasable PROM (Electrically Erasable PROM, EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which is used as an external cache memory. By way of illustration and not limitation, RAM may be in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus direct RAM (RDRAM) and direct memory bus dynamic RAM (DRDRAM).

[0127] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. Those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application. It should be noted that "multiple" in the present application includes "two or more".

[0128] In the various embodiments of the present application, it should be understood that the size of the serial numbers of the above-mentioned processes does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0129] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0131] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] The above is a detailed introduction to a battery capacity prediction method and device, electronic device, and storage medium disclosed in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those of ordinary skill in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A battery capacity prediction method, characterized in that: Applied to a battery capacity prediction system, the method comprises: Acquire target characteristic data of a target battery cell in a battery manufacturing process, wherein the target characteristic data is used to characterize the target characteristic data of parameter values ​​of the battery cell in each process link in the battery manufacturing process; Inputting the target characteristic data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell; The capacity prediction model is obtained by training the first sample feature data of the parameter values ​​of the sample battery in each process link of the battery manufacturing process. The capacity prediction model is built based on ensemble learning, and the first sample feature data is the training target sample feature data obtained after excluding the abnormal data of the sample battery.

2. The method according to claim 1, characterized in that The method for obtaining the first sample characteristic data includes: Obtaining sample characteristic data of parameter values ​​of the first sample battery cell in each process link in the battery manufacturing process, wherein the sample characteristic data of parameter values ​​in the process link at least includes: the weight of the sample battery cell, the weighing of the sample battery cell before and after the first injection of electrolyte and the first injection amount of electrolyte, the voltage and temperature at each stage of the formation process, and the formation capacity; Screening the parameter value sample characteristic data to obtain preliminary screening sample characteristic data, wherein the preliminary screening sample characteristic data is within a preset data range; An unsupervised algorithm is used to further screen the initial screening sample characteristic data of the sample battery cell to obtain the target sample characteristic data.

3. The method according to claim 1 or 2, characterized in that: Obtain target characteristic data of the target cell in the battery manufacturing process, including: Obtaining characteristic data of parameter values ​​of the target battery cell in each process link in the battery manufacturing process, wherein the characteristic data of parameter values ​​in the process link at least includes: the weight of the target battery cell, the weighing of the target battery cell before and after the first injection of electrolyte and the first injection amount of electrolyte, the voltage and temperature at each stage of the formation process, and the formation capacity; Screening the characteristic data of the parameter value to obtain preliminary screening characteristic data, wherein the preliminary screening characteristic data is within a preset data range; An unsupervised algorithm is used to further screen the initial screening feature data of the target battery cell to obtain the target feature data.

4. The method according to claim 2, characterized in that: The method further comprises: Collect second sample characteristic data. When the data volume of the second sample characteristic data reaches a preset data volume value, input the second sample characteristic data into the capacity prediction model for training and updating, wherein the second sample characteristic data is sample characteristic data of parameter values ​​of updated sample cells in each process link of the battery manufacturing process.

5. The method according to claim 4, characterized in that The method further comprises: When the continuous deviation between the predicted value and the true value of the capacity prediction model reaches a set threshold, the second sample feature data is input into the capacity prediction model for training and updating to obtain the capacity prediction model with updated model parameters.

6. The method according to claim 1, characterized in that When there are multiple target battery cells, the target battery cells correspond to station identifications of battery cell production equipment, and the method further includes: When the predicted capacity of the target battery cell is less than a preset first threshold, outputting the workstation identification information corresponding to the target battery cell; or, Output comparison information of the target characteristic data of the first group of target battery cells and the second group of target battery cells, and identify significant differences, wherein the predicted capacity of the first group of target battery cells is lower than the preset first threshold, the predicted capacity of the second group of target battery cells is higher than the preset second threshold, and the second threshold is higher than or equal to the preset first threshold.

7. The method according to claim 1, characterized in that The prediction model includes a first-layer base learner and a second-layer meta learner, the first-layer base learner includes multiple regression models, and the target feature data is input into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell, including: The first subset of the target feature data subsets is used as a test set, and the remaining subsets are used as training sets to train a first regression model; wherein the target sample feature data subsets are multiple independent subsets obtained by dividing the target feature data; The second subset of the target feature data subset is used as a test set, and the remaining subsets are used as training sets to train the second regression model, and so on, to train the Nth regression model, where N is greater than or equal to 3 and less than or equal to the number of regression models included in the first layer base learner; Perform three-fold cross validation on each trained regression model; A preliminary prediction is performed on the target feature data subset as the test set after three-fold cross-validation. The preliminary prediction result is input as feature data to the second-layer meta-learner, and the output value of the second-layer meta-learner is the predicted battery capacity.

8. A battery capacity prediction device, characterized in that: include: A data acquisition module, used to acquire target characteristic data of a target battery cell in a battery manufacturing process, wherein the target characteristic data is used to characterize the target characteristic data of parameter values ​​of the battery cell in each process link in the battery manufacturing process; The capacity prediction module is used to input the target feature data into a pre-trained capacity prediction model to obtain a prediction result of the battery capacity of the target battery cell.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images