System and method for analyzing process data of battery cells

Abstract

A battery cell process data analysis system according to one embodiment includes a process data acquisition unit that acquires process data for a battery cell, a predicted capacity estimation unit that estimates a predicted capacity of the battery cell based on the process data, an actual capacity measurement unit that measures an actual capacity of the battery cell, and a controller that calculates a difference between the predicted capacity of the battery cell and the actual capacity of the battery cell, and determines process factors that affect the difference in capacity using a trained artificial intelligence model.

Description

[Technical Field]

[0001] This invention claims priority under Korean Patent Application No. 10-2022-0037600 dated March 25, 2022, and all content disclosed in the said Korean Patent Application is incorporated herein by reference.

[0002] The embodiments disclosed herein relate to a process data analysis system for battery cells and a process data analysis method for battery cells using the same. [Background technology]

[0003] Lithium polymer batteries, a type of rechargeable battery, are batteries that can be repeatedly charged and discharged through the reversible interconversion between chemical and electrical energy. They have the advantages of a long lifespan and large capacity, and are widely used in portable electronic devices, electric vehicles, and other applications. Battery capacity, one of the most important performance indicators of a battery cell, is affected by numerous process factors involved in the manufacturing process. Traditionally, processes were managed by simply analyzing the correlation between battery cell design factors (i.e., variables controllable during the battery design process) and capacity. However, the actual manufacturing process of a battery cell consists of multiple discontinuous processes with different characteristics, such as electrode processes, assembly processes, and activation processes, and various process factors other than design factors can affect the battery cell's capacity. The influence of such process factors on battery cell capacity is generally irregular, and since multiple processes, not just a single process, affect it simultaneously, it has been difficult to clearly grasp these correlations. [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] One objective of the embodiments disclosed herein is to provide a battery cell process data analysis system and method that can identify process factors affecting the capacity of a battery cell by using artificial intelligence-based data analysis.

[0005] Another objective of the embodiments disclosed herein is to improve the capacity quality of produced battery cells by identifying and controlling process factors that have a positive effect on the capacity of the battery cells (e.g., making the actual capacity greater than the design capacity) and process factors that have a negative effect (e.g., making the actual capacity less than the design capacity).

[0006] The technical problems of the embodiments disclosed in this document are not limited to those mentioned above, and other technical problems not mentioned can be clearly understood by those skilled in the art from the following description. [Means for solving the problem]

[0007] A battery cell process data analysis system according to one embodiment for achieving the above objective includes: a process data acquisition unit that acquires process data of a battery cell; a predicted capacity estimation unit that estimates the predicted capacity of the battery cell based on the process data; an actual capacity measurement unit that measures the actual capacity of the battery cell; and a controller that calculates the difference between the predicted capacity of the battery cell and the actual capacity of the battery cell and determines process factors that affect the difference in capacity using a trained artificial intelligence model.

[0008] According to one embodiment, the predictive capacity estimation unit calculates the size of the electrodes of each monocell based on the measured values ​​of each monocell, and estimates the predicted capacity of a battery cell in which multiple monocells are stacked based on the size of the electrodes and the amount of electrode active material loaded on the electrodes.

[0009] According to one embodiment, the controller acquires distribution data of actual capacity relative to predicted capacity for a plurality of battery cells manufactured through a process, classifies the distribution data into two or more groups according to the magnitude of the actual capacity relative to the predicted capacity of the battery cells using a first artificial intelligence model, and calculates the importance of the influence of the process factor on the capacity of the battery cells by analyzing the distribution data belonging to each group and the corresponding process data of the battery cells using a second artificial intelligence model.

[0010] According to one embodiment, the second artificial intelligence model can be trained to calculate the importance of each process factor for two or more process factors based on the change in the distribution data depending on the presence or absence of the process factor.

[0011] According to one embodiment, the controller classifies the distribution data into a first group if the difference between the predicted capacity and the actual capacity of the input battery cell is greater than or equal to a first threshold, and classifies the distribution data into a second group if the difference between the predicted capacity and the actual capacity of the input battery cell is less than or equal to a second threshold. For the distribution data belonging to the first group, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as positive process factors, and for the distribution data belonging to the second group, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as negative process factors.

[0012] According to one embodiment, the process data of the battery cell includes data related to at least one of the electrode process, assembly process, and activation process. The process data acquisition unit generates integrated process data by concatenating process data related to different processes using a key index. The controller can use the integrated process data to determine process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell from among the process factors related to different processes.

[0013] A battery cell process data analysis method according to one embodiment for achieving the above objective includes the steps of: acquiring process data of the battery cell; estimating the predicted capacity of the battery cell based on the process data; measuring the actual capacity of the battery cell; calculating the difference between the actual capacity and the predicted capacity of the battery cell; and determining process factors that affect the difference in capacity using a trained artificial intelligence model.

[0014] According to one embodiment, the step of estimating the predicted capacity may include the steps of calculating the size of the electrodes of individual monocells based on the measured values ​​of each monocell, and estimating the predicted capacity of a battery cell in which a plurality of individual monocells are stacked based on the size of the electrodes and the amount of electrode active material loaded on the electrodes.

[0015] According to one embodiment, the step of determining the process factor may include: acquiring distribution data of actual capacity relative to predicted capacity for a plurality of battery cells manufactured through the process; classifying the distribution data into two or more groups according to the magnitude of the actual capacity relative to the predicted capacity of the battery cells using a first artificial intelligence model; and calculating the importance of the influence of the process factor on the capacity of the battery cells by analyzing the distribution data belonging to each group and the corresponding process data of the battery cells using a second artificial intelligence model.

[0016] According to one embodiment, the second artificial intelligence model can be trained to calculate the importance of each process factor based on the change in the distribution data depending on the presence or absence of two or more pre-set process factors.

[0017] According to one embodiment, in the step of classifying the distribution data into two or more groups, if the difference between the predicted capacity and the actual capacity of the input battery cell is greater than or equal to a first threshold, the distribution data is classified into a first group; if the difference between the predicted capacity and the actual capacity of the input battery cell is less than or equal to a second threshold, the distribution data is classified into a second group; and in the step of determining the process factors, for the distribution data belonging to the first group, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as positive process factors, and for the distribution data belonging to the second group, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as negative process factors.

[0018] According to one embodiment, the process data of the battery cell includes data related to at least one of an electrode process, an assembly process, and an activation process. In the step of obtaining the process data, key indexes are used to concatenate respective process data related to different processes to generate integrated process data. In the step of determining the process factors, the integrated process data can be used to determine process factors related to different processes that affect the difference between the predicted capacity and the actual capacity of the battery cell.

Effects of the Invention

[0019] According to the battery cell process data analysis system and method of the above embodiment, process factors that affect the capacity of the battery cell can be found through artificial intelligence-based data analysis. Also, by concatenating data of different processes using key indexes, a plurality of process factors that affect the battery capacity in each process step can be integrated and managed.

[0020] Furthermore, by finding process factors that have a positive effect on the capacity of the battery cell (for example, making the actual capacity larger than the predicted capacity) and process factors that have a negative effect (for example, making the actual capacity smaller than the predicted capacity), and managing the process factors, the capacity quality of the produced battery cells can be improved. In addition, various effects that can be directly or indirectly understood can be provided by this document.

[0021] To more clearly explain the embodiments disclosed in this document or the technical solutions of the prior art, the drawings necessary for the description of the embodiments are briefly introduced below. It should be understood that the following drawings are only for explaining the embodiments of this specification and are not for limiting purposes. Also, for the sake of clarity of the description, the representation of some components in the drawings may be exaggerated or omitted.

Brief Description of the Drawings

[0022] [Figure 1] A graph showing the predicted and actual distributions of measured actual capacity relative to the predicted capacity calculated for a plurality of battery cells. [Figure 2] A block diagram showing the configuration of a process data analysis system for battery cells according to one embodiment. [Figure 3] A schematic diagram showing that process factors of a plurality of process steps included in the manufacturing process of a battery cell affect the capacity of the completed battery cell. [Figure 4] A schematic diagram showing linking process data of a plurality of processes included in the manufacturing process of a battery cell using a key index. [Figure 5] Shows how to measure the size of an electrode to calculate the predicted capacity of a battery monocell. [Figure 6] A graph showing distribution data of measured actual capacity relative to the predicted capacity for a plurality of battery cells. [Figure 7] Shows classification of capacity distribution data of battery cells into two groups using an artificial intelligence model according to one embodiment. [Figure 8] Shows classification of capacity distribution data of battery cells into three groups using an artificial intelligence model according to one embodiment. [Figure 9] A flowchart showing a method for analyzing process data of battery cells according to one embodiment. [Figure 10] A flowchart showing in more detail the step of determining the process factor of FIG. 9.

Embodiments for Carrying Out the Invention

[0023] The embodiments disclosed in this document will be described in detail below with reference to illustrative drawings. It should be noted that, when assigning reference numerals to components in each drawing, the same reference numerals will be used for the same components whenever possible when they appear in other drawings. Furthermore, when describing the embodiments disclosed in this document, if a specific description of a related known configuration or function is deemed to hinder understanding of the embodiments disclosed in this document, such detailed description will be omitted.

[0024] The terminology used in this document has been selected to the greatest extent possible from commonly used terms, taking into consideration their function, although this may vary depending on the intentions or conventions of engineers in the field or the emergence of new technologies. In some cases, the applicant has arbitrarily selected terms, in which case their meanings will be described in the explanatory section of the specification. Therefore, it should be made clear that the terms used in this document are not merely names of terms, but must be interpreted based on their substantive meaning and the overall content of this document. Furthermore, the terms used in this document are used solely to describe specific embodiments and are not intended to limit the scope of other embodiments.

[0025] In this document, singular expressions may include plural expressions unless the context clearly indicates otherwise. Furthermore, expressions such as "first" and "second" are used to distinguish components from one another and do not imply any rank or order among the components.

[0026] Preferred embodiments of a battery cell process data analysis system and a battery cell process data analysis method using the same will be described below with reference to the drawings. Figure 1 shows the distribution of predicted and actual capacities measured for multiple battery cells. As shown in the figure, the predicted distribution indicates a linear relationship between predicted and actual capacities, but in reality, the distribution tends to be wider due to the influence of various process factors. In other words, in the actual distribution, the measured actual capacity may be smaller (lower part of the predicted distribution range) or larger (upper part of the predicted distribution range) than the calculated predicted capacity due to the influence of unexpected process factors. The influence of such process factors on the capacity of battery cells is generally irregular, and since multiple processes, not just a single process, have an effect simultaneously, it has been difficult to clearly grasp these correlations.

[0027] Figure 2 is a block diagram showing the configuration of a battery cell process data analysis system according to one embodiment for solving the above problem. Referring to Figure 2, the battery cell process data analysis system 10 according to the embodiment may include a process data acquisition unit 110, a predicted capacity estimation unit 120, an actual capacity measurement unit 130, and a controller 140. The components shown in the block diagram are separated according to their respective functions and roles, and each block does not necessarily have to be implemented with independent hardware or software. For example, the separated components may actually be implemented with one device or program, or one component may be implemented as a combination of multiple devices and programs.

[0028] The process data acquisition unit 110 acquires process data for the battery cell. Here, the process data for the battery cell is data related to multiple processes included in the battery manufacturing process, such as the electrode process, assembly process, and activation process, and can include data on multiple process factors (i.e., process variables) that affect the capacity of the produced battery cell. For example, in the electrode process, the amount of electrode material loaded per unit area, the size and thickness of the electrodes are process factors; in the assembly process, the loading method of the positive electrode, separator, and negative electrode, and the amount of electrolyte are process factors; and in the activation process, the number of charge-discharge cycles of the cell, the temperature and pressure during charge-discharge, etc., can be process factors. The types of data listed are merely examples, and various process data related to process steps that can affect the capacity of the battery cell can be utilized.

[0029] Figure 3 is a schematic diagram illustrating how process factors in multiple process steps included in the battery cell manufacturing process affect the capacity of the finished battery cell. Referring to Figure 3, various processes (first to third processes) and process factors (process factors 1 to 6) involved in these processes influence the determination of the battery cell capacity. Among these process factors, the process data acquisition unit 110 can collect process data for each process in order to find the process factors that have a significant impact on the difference between the theoretically calculated predicted capacity and the actual measured capacity (or process factors that have a positive / negative impact).

[0030] According to one embodiment, the process data acquisition unit 110 can generate integrated process data by linking process data related to different processes using a key index. Conventionally, skilled engineers have relied on their personal experience to directly identify and manage design factors or process factors that affect the capacity of battery cells. While this method is effective for managing individual processes (e.g., one of the electrode process, assembly process, and activation process), it is difficult to grasp the relationships between process factors in multiple process steps, and therefore becomes less effective as the number of process steps becomes more complex. In other words, since some process factors can influence each other as dependent variables rather than independent variables for battery capacity, it is advantageous to integrate and manage process data from multiple process steps. For this purpose, it is necessary to link this data while considering the relationships between the process data of each process.

[0031] Figure 4 is a schematic diagram illustrating how process data from multiple steps included in the battery cell manufacturing process are linked using a key index. Referring to Figure 4, for multiple battery cells (AAA to ZZZ), process data from the first process (e.g., electrode process) (electrode 1, electrode 2, ..., electrode n), the second process (e.g., assembly process) (assembly 1, assembly 2, ..., assembly n), and the third process (e.g., activation process) (activation 1, activation 2, ..., activation n) are linked using key indices to generate integrated process data. In this process, the key indices can be set considering the relationships between each process.

[0032] The key index can vary depending on the form of data generated in the process. For example, in a monocell assembly process, each monocell is assigned an ID, so a key index can be set that links the IDs of multiple monocells to the ID of a single completed cell. In this way, in the assembly process, each data point exists discontinuously, so linking between processes is possible by tracking the IDs.

[0033] In the electrode process, a roll-to-roll process results in continuous time-series data under a single electrode ID. Therefore, when linking the electrode process and the assembly process, the process data can be linked by tracking the ID of the electrode fed into the assembly process and the data within a single electrode ID in chronological or reverse chronological order.

[0034] The integrated process data is input into an artificial intelligence model, which will be described later. Based on this integrated process data, the AI ​​model can determine process factors that significantly affect the capacity of the battery cells. By integrating and managing process data from multiple process steps in this way, the impact of each process factor on battery capacity can be easily and accurately understood.

[0035] The predicted capacity estimation unit 120 can estimate the predicted capacity of the battery cell based on the process data. Here, predicted capacity means the theoretical maximum capacity calculated based on the design variables of the battery cell. According to one embodiment, the size of the electrodes (e.g., the size of the positive or negative electrode) can be calculated from vision data obtained by photographing the upper and lower parts of the surface of a monocell (a unit cell consisting of a positive electrode, a negative electrode, and a separator), and the theoretical capacity that the battery cell can electrically possess can be estimated based on the size of the electrodes and the amount of electrode active material loaded onto the electrodes.

[0036] Figure 5 shows how the electrode size is measured to calculate the predicted capacity of a battery monocell. Referring to Figure 5, the monocell 20 consists of a positive electrode 210, a negative electrode 220, and a separator 230 inserted between the positive and negative electrodes to separate them. Using vision data captured from the upper and lower surfaces of the monocell, the distance from each part of the positive electrode 210 to the end of the separator 230 (see arrows in Figure 5) can be determined. Since the lateral length (W) of the separator 230 is already known as a design variable, the lateral length of the positive electrode 210 can be calculated by subtracting the lateral distance between the positive electrode 210 and the separator 230 (see arrow) from the lateral length (W) of the separator 230. Similarly, since the vertical length (H) of the separator 230 is also already known as a design variable, the vertical length of the positive electrode 210 can be calculated by subtracting the vertical distance between the positive electrode 210 and the separator 230 (see arrow) from the vertical length (H) of the separator 230.

[0037] The size of the positive electrode 210 can be calculated from the lateral and vertical lengths of the positive electrode 210 calculated in this way, and by substituting the amount of active material per unit area, the theoretical predicted capacity of each monocell 20 can be calculated. Furthermore, since one battery cell has a structure in which multiple monocells are stacked, it is possible to estimate the theoretical predicted capacity of the battery cell by adding up the predicted capacities of all the monocells. The method for calculating the size of the electrodes described above assumes that specified design variables are taken into consideration, but this is merely an example. According to the embodiment, a method can be considered in which the size of the electrodes is calculated by using an image processing algorithm to identify specified individuals (e.g., positive electrode, negative electrode, separator) in vision data captured from the upper and lower parts of the monocell surface, and calculating the size of the identified individuals.

[0038] Generally, in lithium-ion batteries, the capacity of a battery cell is determined by the size of the positive electrode and the amount of electrode active material loaded, so the size of the positive electrode was used, but it is not limited to this. For example, when the capacity of a battery cell is determined by the size of the negative electrode and the amount of electrode active material loaded according to the design, the predicted capacity estimation unit 120 can estimate the theoretical capacity of the battery cell based on the size of the negative electrode 220 and the corresponding amount of electrode active material.

[0039] The actual capacity measurement unit 130 measures the actual capacity of the completed battery cell. The capacity of the battery cell can generally be measured by measuring the cell current after the charge-discharge process, but there are no particular limitations on the timing and method of measurement.

[0040] The controller 140 calculates the difference between the predicted capacity of the battery cell estimated by the predicted capacity estimation unit 120 and the actual capacity of the battery cell measured by the actual capacity measurement unit 130, and can use a trained artificial intelligence model to determine the process factors that affect the difference in capacity. To this end, the controller 140 acquires distribution data of the actual capacity relative to the predicted capacity for multiple battery cells manufactured through the process.

[0041] Figure 6 is a graph showing the distribution data of measured actual capacity against predicted capacity for multiple battery cells. Referring to Figure 6, the horizontal axis of the graph represents the predicted capacity estimated based on monocell vision data, and the vertical axis represents the actual capacity measured by inspection of the completed battery cell. Each point represents the predicted and actual capacity of the battery cell. As illustrated, even battery cells that have gone through the same manufacturing process will have different predicted capacities depending on the size of the monocell electrodes, and the actual capacity will also have errors of varying magnitudes due to various process factors that influence the process as it goes through multiple stages. In particular, if the actual capacity is significantly lower than the predicted capacity (i.e., points located relatively low on the graph), it can lead to a decrease in product quality, so it is necessary to identify and manage the process factors that affect such results.

[0042] To determine the process factors that affect the difference between actual capacity and predicted capacity, the controller 140 can use a clustering algorithm and / or a first artificial intelligence model to classify the data into two or more groups according to pre-established classification criteria (for example, depending on the ratio of the actual capacity to the predicted capacity of the battery cells on the distribution data).

[0043] According to one embodiment, the clustering algorithm may be the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm. Here, the DBSCAN algorithm is an algorithm for classifying points into core points, border points, or noise points based on whether a specified number of points are included within a specified radius from each point representing the predicted capacity and actual capacity of a battery cell. Core points may be points that contain a specified number or more points within the radius. Border points may be points that contain fewer than the specified number of points within the radius, and that contain core points. Noise points may be points that contain fewer than the specified number of points within the radius, and that do not contain core points. Here, the size and number of the radius can be predefined by the user. According to the embodiment, two or more groups classified according to the ratio of the actual capacity to the predicted capacity of a battery cell on the distribution data may be clusters composed of different core points and border points. Noise points may not be included in the two or more groups. Referring to Figure 7, points classified as "GOOD" may be a first cluster consisting of core points and boundary points, while points classified as "NOT GOOD" may be a second cluster consisting of other core points and other boundary points. Here, the first and second clusters may refer to two or more groups.

[0044] According to one embodiment, the first artificial intelligence model is a machine learning-based classification model that can discontinue and / or classify adjacent data in order to improve classification performance. For example, the first artificial intelligence model can utilize a DBSCAN-based model, but is not limited to this. For example, if the first artificial intelligence model is DBSCAN-based, the first artificial intelligence model can be trained to perform supervised learning based on training data. For example, the first artificial intelligence model can be trained to minimize the difference between core points, boundary points, or noise points output by the first artificial intelligence model and core points, boundary points, or noise points labeled in the training data.

[0045] Figure 7 shows the classification of battery cell capacity distribution data into two groups using an artificial intelligence model according to one embodiment. In Figure 7, points classified as "GOOD" refer to battery cells where the difference between the theoretically estimated predicted capacity and the actual capacity is relatively small, while points classified as "NOT GOOD" refer to battery cells where the difference between the predicted capacity and the actual capacity is relatively large. As mentioned above, a large difference between the predicted capacity and the actual capacity can lead to a decrease in product quality, so it is necessary to identify and manage process factors that affect such results.

[0046] The controller 140 uses a second artificial intelligence model to analyze the distribution data belonging to each group (for example, a group classified as "GOOD" or a group classified as "NOT GOOD") and the corresponding battery cell process data, thereby calculating the importance of each process factor influencing the capacity of the battery cell.

[0047] According to one embodiment, a second artificial intelligence model can be trained to calculate the importance of each process factor based on changes in the distribution data due to changes in or the presence or absence of two or more pre-set process factors. The second artificial intelligence model can be based on, for example, a model that utilizes an ensemble-based boosting method (e.g., gradient boost for classification, XGBoost (e.g., XGBClassifier, LightGBMClassifier)). Here, the ensemble may be a technique that generates prediction results by combining multiple models (e.g., a decision tree).

[0048] In one embodiment, the second artificial intelligence model can combine decision trees for each of multiple process factors to understand the importance of each process factor. Here, each decision tree may contain two leaves branching from one decision node.

[0049] In one embodiment, a second artificial intelligence model can be trained to minimize the difference between the capacity difference of a battery cell classified as the final leaf and the predicted value of that final leaf (i.e., the pseudo-residual (or error) value). Here, the second artificial intelligence model can consist of a decision tree of a specified depth. The specified depth can be set to an integer between 8 and 32.

[0050] The second artificial intelligence model can be trained by utilizing the process data from the first to third processes as input data and using the capacity difference as the predicted value. The second artificial intelligence model can be trained by the calculation process of the pseudo-residuals of each leaf of the current decision tree, the generation process of the child decision tree of the current decision tree, and the generation process of the predicted values ​​of each leaf included in the child decision tree.

[0051] Each leaf can indicate a pseudo-residual (or error) value. Here, the pseudo-residual of each leaf of the nth decision tree is the capacity difference of the battery cells classified into each leaf of the nth decision tree and the predicted value (P n-1 ) determined by the previous decision tree, and can be calculated based on the difference value therebetween.

[0052] Based on the pseudo-residual of each leaf of the nth decision tree, a predicted value for the capacity difference of the battery cells included in each leaf of the nth decision tree can be calculated. The initial predicted value (P0) can be set to the average value of the capacity differences of all battery cells. The predicted value (P n ) of each leaf of the nth decision tree may be the sum of the predicted value (P n-1 ) of the previous decision tree and the value obtained by multiplying the output value (O n ) of the leaf of the current decision tree itself by a weight (η (for example, 0.1)) (P n = P n-1 + η × O n ). Here, the output value (O n ) of the leaf of the current decision tree itself can be calculated based on the pseudo-residual of the battery cells included in its own leaf. For example, the output value (O n ) may be the average value of the pseudo-residuals. That is, the output value (O n ) of each leaf of the nth decision tree may be the value (S = R / M) obtained by dividing the sum value (R = r1 + r2 + … + r M ) of the pseudo-residuals of the battery cells included in each leaf of the nth decision tree by the number (M) of the battery cells included in that leaf. For example, the output value of the first leaf and the output value of the second leaf of the nth decision tree may be different.

[0053] Decision trees branching from a previous decision tree can be determined based on gain values. For example, decision trees for each of the two leaves included in the previous decision tree can be determined based on gain values. That is, the decision trees for each of the two leaves may be different. For example, among the candidate decision trees for the first leaf of the previous decision tree, which are decision trees classifying the presence or absence of process factor 2, decision trees classifying the presence or absence of process factor 3, ..., decision trees classifying the presence or absence of process factor N, the decision tree with the highest gain value can be determined as the decision tree for the first leaf of the previous decision tree. Here, the gain value of each candidate decision tree is calculated based on the similarity score of each candidate decision tree, and the similarity score (S) is the sum of the pseudo-residuals of the battery cells included in each leaf included in each decision tree (R = r1 + r2 + ... + r M The similarity score may also be the value obtained by dividing the square of (R^2) by the number of battery cells (M) (S = (R^2) / M). Here, the similarity score of one decision tree may include the similarity score of the first leaf (S1) and the similarity score of the second leaf (S2). The gain value may also be the value obtained by subtracting the similarity score of the parent node from the sum of the similarity scores of the first leaf (S1) and the second leaf (S2).

[0054] In one embodiment, a decision node can classify each battery cell into one of two leaves by classifying whether or not each process factor is present, or by classifying based on a criterion value for each process factor. For example, a first decision node can query whether or not process factor 1 is present in the first process. The first decision node can then assign cell IDs where process factor 1 is present in the first process to the first leaf of the decision node, and assign cell IDs where process factor 1 is not present in the first process to the second leaf of the first decision node. A second decision node can then query whether or not process factor 2 is present in the first process. The second decision node can then assign cell IDs where process factor 2 is present in the first process to the first leaf of the second decision node, and assign cell IDs where process factor 2 is not present in the first process to the second leaf of the second decision node. Here, the first decision node and the second decision node can be linked to each other. For example, the decision process of the second decision node can be performed on battery cells assigned to the first leaf of the first decision node. In this case, the first decision node may be the parent node or a previous node of the second decision node.

[0055] In one embodiment, the second artificial intelligence model can calculate the importance of each process factor using the permutation importance method based on process data classified into two or more groups. Here, the importance of a process factor can be based on the gain value of the decision tree in which the process factor is utilized. For example, the sum of the gain values ​​of the decision trees that query the presence or absence of the first process factor can be evaluated as the importance of the first process factor. However, the type of model and the method of utilization are merely examples and are not limited thereto.

[0056] For example, the second artificial intelligence model can compare the capacity difference (difference between predicted capacity and actual capacity) calculated excluding the intervention of process factor 1 from the process data with the capacity difference calculated excluding the intervention of process factor 2, thereby determining which factor, process factor 1 or process factor 2, had a greater influence on the difference between predicted and actual capacity. By using this method, the importance of each process factor can be calculated by comparing all process factors included in the process data with the capacity difference (distribution data) with and without the intervention of that process factor.

[0057] In this way, the controller 140 can determine the process factors that affect the capacity of the battery cell based on the importance of each process factor calculated using an artificial intelligence model. Once the process factors are determined, the capacity quality of the battery cell can be improved by controlling these process factors through appropriate feedback.

[0058] In a further embodiment, the controller 140 can classify the distribution data of the actual capacity relative to the predicted capacity of the battery cells into three or more groups. Unlike Figure 7, which classifies the distribution data into two groups, Figure 8 shows the classification of the battery cell capacity distribution data into three groups ("GOOD", "OK", and "NOT GOOD") using an artificial intelligence model according to one embodiment.

[0059] Specifically, the controller 140 can classify the distribution data into the first group, "GOOD," if the difference between the predicted capacity and the actual capacity of the input battery cell is greater than or equal to a first threshold; classify the distribution data into the second group, "NOT GOOD," if the difference between the predicted capacity and the actual capacity of the input battery cell is less than or equal to a second threshold; and classify it into "OK" in all other cases.

[0060] Here, battery cells belonging to the first group ("GOOD") can be considered to be of relatively superior quality because their actual capacity is greater than their theoretically predicted capacity. Therefore, for the distribution data belonging to the first group, the controller 140 can determine process factors that significantly affect the capacity difference of the battery cells as positive process factors. Conversely, battery cells belonging to the second group ("NOT GOOD") can be considered to be of relatively lower quality because their actual capacity is smaller than their theoretically predicted capacity. Therefore, for the distribution data belonging to the second group, the controller 140 can determine process factors that significantly affect the capacity difference of the battery cells as negative process factors. Furthermore, by increasing the influence of positive process factors and decreasing the influence of negative process factors in subsequent battery manufacturing processes, the capacity quality of the battery cells can be improved.

[0061] Figure 9 is a flowchart showing a process data analysis method for a battery cell according to one embodiment. Each step of the method of this embodiment can be performed by any one or a combination thereof of the components of the battery cell process data analysis system 10 described with reference to Figure 1, namely the process data acquisition unit 110, the predicted capacity estimation unit 120, the actual capacity measurement unit 130, and the controller 140, but the system 10 is not essential to perform this method.

[0062] Referring to Figure 9, the first step (S100) is to acquire process data for the battery cell. According to one embodiment, the process data for the battery cell may include data related to multiple processes included in the battery manufacturing process, such as the electrode process, assembly process, and activation process.

[0063] Furthermore, in step (S100), integrated process data can be generated by linking process data related to different processes using a key index. Using this integrated process data, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell can be determined from among the process factors related to different processes. By integrating and managing process data from multiple process steps in this way, the influence of each process factor on battery capacity can be easily and accurately grasped.

[0064] Next, a step (S200) is performed to estimate the predicted capacity of the battery cell based on the process data. Predicted capacity refers to the theoretical maximum capacity calculated based on the design variables of the battery cell. According to one embodiment, step (S200) may include a step of calculating the size of the electrodes of an individual mono cell based on the measured values ​​of the mono cell, and a step of estimating the predicted capacity of a battery cell in which a plurality of individual mono cells are stacked based on the size of the electrodes and the amount of electrode active material loaded on the electrodes. Specifically, the size of the electrodes (e.g., the size of the positive or negative electrode) can be calculated from vision data obtained by photographing the upper and lower parts of the surface of a mono cell (a unit cell consisting of a positive electrode, a negative electrode, and a separator), and the theoretical capacity that the battery cell can electrically have can be estimated based on this. The process of calculating the size of the positive electrode from the vision data of a mono cell is as described above with reference to Figure 5. Also, as described above, depending on the design, the theoretical capacity of the battery cell can be estimated based on the size of the negative electrode and the corresponding amount of electrode active material.

[0065] Next, a step (S300) is performed to measure the actual capacity of the completed battery cell. As mentioned above, the capacity of the battery cell can generally be measured by measuring the cell current after the charge-discharge process, but there are no particular limitations on the timing and method of measurement.

[0066] Next, a step (S400) is performed to calculate the difference between the actual capacity and the predicted capacity of the battery cell. The calculated capacity difference can be used in the next step to determine the main process factors that affect the capacity difference of the battery cell.

[0067] Next, a step (S500) is performed to determine the process factors that affect the difference in capacity using a trained artificial intelligence model. Even battery cells that have gone through the same manufacturing process will have different predicted capacities depending on the size of the electrodes of the monocell and the amount of electrode active material loaded. Furthermore, as the cells go through multiple processes, various process factors will have an effect, resulting in errors of varying magnitudes in the actual capacity. In particular, if the actual capacity is significantly lower than the predicted capacity, it can lead to a decline in product quality, so it is necessary to find and manage the process factors that affect such results.

[0068] Figure 10 is a flowchart that shows in more detail the steps for determining the process factors in Figure 9. Referring to Figure 10, step (S500) may include: acquiring distribution data of actual capacity relative to predicted capacity for multiple battery cells manufactured through the process (S510); classifying the distribution data into two or more groups according to the magnitude of the actual capacity relative to the predicted capacity of the input battery cells using a first artificial intelligence model (S520); and calculating the importance of each process factor influencing the capacity of the battery cells by analyzing the distribution data belonging to each group and the corresponding process data of the battery cells using a second artificial intelligence model (S530).

[0069] According to one embodiment, the first artificial intelligence model is a machine learning-based classification model that can discontinue adjacent data to improve classification performance. For example, DBSCAN can be used, but is not limited to this. Furthermore, the second artificial intelligence model can be trained to calculate the importance of each process factor based on the degree of change in the distribution data due to changes in the process factors for two or more pre-set process factors. More specifically, the importance of each process factor can be calculated by comparing all process factors included in the process data with the capacity difference (distribution data) with and without intervention of the process factor. The second artificial intelligence model can, for example, use classification models such as XGBClassifier and LightGBMClassifier to find process data that can be classified into two or more groups, and calculate the importance of each process factor using the permutation importance method. However, the type of model and the method of use are merely examples and are not limited to these.

[0070] In this way, based on the importance of each process factor calculated using an artificial intelligence model, it is possible to determine the process factors that affect the capacity of the battery cell, and furthermore, by controlling these process factors through appropriate feedback, the capacity quality of the battery cell can be improved.

[0071] In a further embodiment, in the step of classifying the distribution data into two or more groups (S520), if the difference between the predicted capacity of the input battery cells and the actual capacity is greater than or equal to a first threshold, the distribution data can be classified into a first group (for example, the group belonging to "GOOD" in Figure 8), and if the difference between the predicted capacity of the input battery cells and the actual capacity is less than or equal to a second threshold, the distribution data can be classified into a second group (for example, the group belonging to "NOT GOOD" in Figure 8).

[0072] Based on the classification results of step (S520), in step (S500) for determining process factors, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell can be determined as positive process factors for distribution data belonging to the first group (for example, the group belonging to "GOOD" in Figure 8), and process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell can be determined as negative process factors for distribution data belonging to the second group (for example, the group belonging to "NOT GOOD" in Figure 8). Furthermore, by increasing the influence of positive process factors and decreasing the influence of negative process factors in the subsequent battery manufacturing process, the capacity quality of the battery cell can be improved.

[0073] The battery cell data analysis method according to the above embodiment can be implemented in an application or in the form of program instructions that can be executed via various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., individually or in combination.

[0074] According to the above embodiments, a battery cell process data analysis system and method are provided that can identify process factors affecting the capacity of a battery cell through artificial intelligence-based data analysis. Furthermore, by linking data from different processes using a key index, multiple process factors affecting battery capacity at each process step can be integrated and managed.

[0075] By utilizing the proposed system and method, it is possible to improve the capacity quality of produced battery cells by identifying and controlling process factors that have a positive impact on battery cell capacity (i.e., making the actual capacity greater than the predicted capacity) and process factors that have a negative impact (i.e., making the actual capacity smaller than the predicted capacity).

[0076] The above description is merely illustrative of the technical concept disclosed in this document, and any person with ordinary skill in the art to which the embodiments disclosed in this document belong can make various modifications and variations without departing from the essential characteristics of the embodiments disclosed in this document.

[0077] Therefore, the embodiments disclosed herein are for illustrative purposes only, not to limit, the technical ideas disclosed herein, and such embodiments do not limit the scope of the technical ideas disclosed herein. The scope of protection of the technical ideas disclosed herein shall be interpreted in accordance with the claims set forth below, and all technical ideas within an equivalent scope shall be interpreted as being included in the scope of rights of this document.

Claims

1. A process data acquisition unit that acquires process data for battery cells, A predictive capacity estimation unit that estimates the predicted capacity of the battery cell based on the process data, A real capacity measuring unit for measuring the actual capacity of the battery cell, A controller that calculates the difference between the predicted capacity of the battery cell and the actual capacity of the battery cell, and uses the calculated difference and a trained artificial intelligence model to determine the process factors that affect the difference. A battery cell process data analysis system, including the following.

2. The battery cell process data analysis system according to claim 1, characterized in that the predicted capacity estimation unit calculates the size of the electrodes of each monocell based on the measured values ​​of each monocell, and estimates the predicted capacity of a battery cell in which a plurality of monocells are stacked based on the size of the electrodes and the amount of electrode active material loaded on the electrodes.

3. The aforementioned controller, Distribution data of actual capacity relative to predicted capacity is obtained for multiple battery cells manufactured through the process. Using the first artificial intelligence model, the distribution data is classified into two or more groups according to the difference between the actual capacity and the predicted capacity of the battery cell. A battery cell process data analysis system according to claim 1 or 2, characterized in that it uses a second artificial intelligence model to analyze distribution data belonging to each group and the corresponding battery cell process data to calculate the importance of the process factor influencing the capacity of the battery cell.

4. The battery cell process data analysis system according to claim 3, characterized in that the second artificial intelligence model is trained to calculate the importance of each process factor for two or more process factors based on the change in the distribution data depending on the presence or absence of the process factor.

5. The aforementioned controller, If the difference between the predicted capacity and the actual capacity of the input battery cell is greater than or equal to a first threshold, the distribution data is classified into a first group; if the difference between the predicted capacity and the actual capacity of the input battery cell is less than or equal to a second threshold, the distribution data is classified into a second group. The battery cell process data analysis system according to claim 4, characterized in that process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as positive process factors for distribution data belonging to the first group, and process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as negative process factors for distribution data belonging to the second group.

6. The process data for the battery cell includes data related to at least one of the following processes: electrode process, assembly process, and activation process. The aforementioned process data acquisition unit generates integrated process data by linking process data related to different processes using a key index. The battery cell process data analysis system according to claim 1 or 2, characterized in that the controller uses the integrated process data to determine process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell, among process factors related to different processes.

7. Steps to acquire process data for battery cells, A step of estimating the predicted capacity of the battery cell based on the process data, The steps include measuring the actual capacity of the battery cell, The steps include: calculating the difference between the actual capacity and the predicted capacity of the battery cell; A step of determining process factors that affect the difference using the calculated difference and a trained artificial intelligence model, A method for analyzing process data of battery cells, including the process data analysis method for battery cells.

8. The step of estimating the predicted capacity is: A step of calculating the size of the electrodes of the monocell based on the measured values ​​of each monocell, The battery cell process data analysis method according to claim 7, characterized by comprising the step of estimating the predicted capacity of a battery cell in which a plurality of individual monocells are stacked, based on the size of the electrode and the amount of electrode active material loaded on the electrode.

9. The step of determining the aforementioned process factors is: A step of obtaining distribution data of actual capacity relative to predicted capacity for multiple battery cells manufactured through a process, A first artificial intelligence model is used to classify the distribution data into two or more groups according to the difference between the actual capacity and the predicted capacity of the battery cells. A method for analyzing battery cell process data according to claim 7 or 8, characterized by comprising the step of using a second artificial intelligence model to analyze distribution data belonging to each group and the corresponding battery cell process data to calculate the importance of the process factors influencing the capacity of the battery cells.

10. The battery cell process data analysis method according to claim 9, characterized in that the second artificial intelligence model is trained to calculate the importance of each process factor based on the change in the distribution data depending on the presence or absence of two or more pre-set process factors.

11. In the step of classifying the distribution data into two or more groups, if the difference between the predicted capacity and the actual capacity of the input battery cells is greater than or equal to a first threshold, the distribution data is classified into the first group; if the difference between the predicted capacity and the actual capacity of the input battery cells is less than or equal to a second threshold, the distribution data is classified into the second group. The battery cell process data analysis method according to claim 10, characterized in that, in the step of determining the process factors, process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as positive process factors for the distribution data belonging to the first group, and process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell are determined as negative process factors for the distribution data belonging to the second group.

12. The process data for the battery cell includes data related to at least one of the following processes: electrode process, assembly process, and activation process. In the step of acquiring the aforementioned process data, integrated process data is generated by linking the process data of different processes together using a key index. The battery cell process data analysis method according to claim 7 or 8, characterized in that, in the step of determining the process factors, the integrated process data is used to determine process factors that affect the difference between the predicted capacity and the actual capacity of the battery cell, among process factors related to different processes.

Images