Method of screening low voltage defective batteries

Abstract

Disclosed herein relates to a method for screening low voltage defective batteries, comprising: (a) a pre-filtering step of collecting data during an activation process of a plurality of secondary batteries and removing in real time secondary batteries having outlier data; (b) a clustering step of clustering based on similar characteristics or records of the plurality of pre-filtered secondary batteries; (c) a correction step of measuring a pressure drop amount (ΔOCV) for each cluster and correcting a scatter of the pressure drop amount (ΔOCV) according to a temperature or a period; (d) a screening step of screening low voltage defective batteries based on the corrected scatter of the pressure drop amount.

Description

Technical Field

[0001] This application claims the benefit of priority based on Korean Patent Application No. 10-2021-0175819, filed on December 9, 2021.

[0002] This invention relates to a method for screening low-voltage defective batteries with improved detection reliability. Background Technology

[0003] In general, unlike primary batteries which cannot be recharged, secondary batteries are batteries that can be recharged and discharged, and are widely used in electronic devices such as mobile phones, laptops, cameras, and electric vehicles. Specifically, lithium secondary batteries are rapidly expanding their applications because they have higher capacity and higher energy density per unit weight compared to nickel-cadmium or nickel-based energy storage batteries.

[0004] These lithium-ion secondary batteries primarily use lithium-based oxides and carbon materials as the positive and negative electrode active materials, respectively. The lithium-ion secondary battery comprises an electrode assembly and an external material. In the electrode assembly, positive and negative electrode plates, coated with the respective active materials, are arranged with a separator inserted between them. The external material is used to seal and contain the electrode assembly along with the electrolyte.

[0005] During the manufacturing process or during use, various types of defects may occur in lithium secondary batteries for various reasons. Specifically, some manufactured secondary batteries exhibit a voltage drop that is greater than their self-discharge rate, and this phenomenon is known as low voltage.

[0006] This low-voltage defect in secondary batteries is usually caused by internal metallic foreign objects. Specifically, when metallic foreign objects such as iron or copper are present in the positive electrode of a secondary battery, they may grow into dendrites in the negative electrode. Furthermore, these dendrites cause internal short circuits in the secondary battery, which can lead to battery malfunction or damage, and in severe cases, may become a cause of fire.

[0007] Meanwhile, the aforementioned low-voltage defects manifest as a relative increase in voltage drop, and these defects are detected during the aging process in the activation phase of the secondary battery. Specifically, during the (shipment) aging period, low-voltage defective batteries can be screened by comparing the measured drop in open-circuit voltage (OCV) with a reference value while monitoring the battery's open-circuit voltage (OCV).

[0008] Traditionally, when deriving such a reference value, a threshold is selected based on the distribution of voltage drop magnitude (ΔOCV) data accumulated during the battery manufacturing process. This threshold is then used to scale the remaining data using a pallet scale standard, thereby setting a specific sigma (e.g., 3σ or 4σ) as the reference value. Batteries exceeding this specific sigma are then detected as low-voltage defective batteries.

[0009] However, in such traditional methods, because specific thresholds do not reflect manufacturing deviations, there is a risk that defective products cannot be screened when manufacturing deviations occur, or that non-defective products are over-detected as defective. Furthermore, as manufacturing deviations accumulate in each unit process, the spread of open circuit voltage or voltage drop (ΔOCV) increases or abnormal spreads occur during low-voltage defect inspection, which can blur the boundary between good and defective products.

[0010] Therefore, when screening low-voltage defective batteries, it is necessary to develop a technology for screening low-voltage defective batteries that can reliably distinguish good products from defective products even when manufacturing deviations exist. Summary of the Invention

[0011] [Technical Issues]

[0012] The present invention aims to provide a method for screening low-voltage defective batteries, which improves the detectability of good and defective batteries by improving the distribution of voltage drop (ΔOCV) during screening of low-voltage defective batteries.

[0013] [Technical Solution]

[0014] A method for screening low-voltage defective batteries includes: (a) a pre-filtering step of collecting data during the activation process of multiple secondary batteries and removing secondary batteries with outlier data in real time;

[0015] (b) Clustering step, clustering based on the similar characteristics or records of multiple pre-filtered secondary batteries;

[0016] (c) The calibration step involves measuring the pressure drop (ΔOCV) for each cluster and correcting the dispersion of the pressure drop (ΔOCV) based on temperature or time period; and

[0017] (d) Screening step, which screens low-voltage defective batteries based on the distribution of the corrected voltage drop.

[0018] In an exemplary embodiment of the present invention, the pre-filtering step (a) uses a time series anomaly detection algorithm to remove outliers. Here, the time series anomaly detection algorithm is selected from one of the following groups: Control Chart, Random Cut Forest, Dynamic Time Warping, TAnoGAN (Time Series Anomaly Detection with Generative Adversarial Networks), MAD-GAN (Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks), USAD (Unsupervised Anomaly Detection on Multivariate Time Series), LSTM (Long Short-Term Memory) + AE, and LSTM (Long Short-Term Memory) + CNN (Convolutional Neural Network).

[0019] In an exemplary embodiment of the present invention, the clustering step (b) clusters secondary batteries having a consistent criterion of groups selected from the following: LOT cells, tray cells, aging temperature, charge / discharge temperature, and aging period.

[0020] In an exemplary embodiment of the present invention, the correction step (c) includes the step of correcting the distribution of the pressure drop by means of linear regression or machine learning methods.

[0021] In an exemplary embodiment of the present invention, for each individual unit process, the pre-filtering step (a) includes the process of collecting the open-circuit voltage (OCV) or voltage drop of the secondary battery as data.

[0022] In an exemplary embodiment of the present invention, a single unit process includes at least one of an initial charging process, a room temperature aging process, and a high temperature aging process.

[0023] In an exemplary embodiment of the present invention, the screening step (d) uses an anomaly detection algorithm to screen for low-voltage defective batteries in the corrected scatter. Here, the anomaly detection algorithm is selected from one of the following groups: Local Outlier Factor (LOF), Isolation Forest, and OC Support Vector Machine (SVM).

[0024] In an exemplary embodiment of the invention, a screening step (d) is performed after aging for shipment is completed.

[0025] In an exemplary embodiment of the present invention, the screening step (d) screens low-voltage defective batteries for each cluster.

[0026] In an exemplary embodiment of the present invention, the screening step (d) screens low-voltage defective batteries after collecting the dispersion data of the voltage drop for each cluster, corrected by step (c).

[0027] [Beneficial Effects]

[0028] The method for detecting and classifying low voltage according to the present invention has excellent detectability and can reduce the time required to screen low voltage defective batteries. Attached Figure Description

[0029] Figure 1 This is a flowchart of a method for screening low-voltage defective batteries according to an exemplary embodiment of the present invention.

[0030] Figure 2 and Figure 3 This is a cumulative probability distribution diagram of the pressure drop before and after the pre-filtering step of the present invention.

[0031] Figure 4 It is a graph showing the distribution of the pressure drop without the correction steps according to the invention.

[0032] Figure 5 This is a graph showing the distribution of the pressure drop after the correction steps according to the present invention. Detailed Implementation

[0033] This invention can have various modifications and examples, and therefore specific examples are shown in the accompanying drawings and described in detail in the specification. However, it should be understood that the invention is not limited to the specific embodiments, but includes all modifications, equivalents, or substitutions falling within the spirit and scope of the invention.

[0034] The terminology used in this invention is for describing particular embodiments only and is not intended to limit the invention. Singular expressions include plural expressions unless the context clearly specifies otherwise. Terms such as “comprising” or “having” are used herein to specify the presence of features, numbers, steps, actions, components or elements or combinations thereof described in the specification, and it should be understood that the possibility of the presence or addition of one or more other features, numbers, steps, actions, components or elements or combinations thereof is not excluded in advance.

[0035] In this specification, the distribution of voltage drop refers to listing or plotting a graph of the voltage drop values ​​for each of a large number of secondary batteries.

[0036] Figure 1 This is a flowchart of a method for screening low-voltage defective batteries according to an exemplary embodiment of the present invention. (Reference) Figure 1 The method for screening low-voltage defective batteries according to the present invention includes: (a) a pre-filtering step, collecting data during the activation process of multiple secondary batteries and removing secondary batteries with outlier data in real time; (b) a clustering step, clustering multiple pre-filtered secondary batteries based on similar characteristics or records; (c) a correction step, measuring the voltage drop (ΔOCV) of each cluster and correcting the dispersion of the voltage drop (ΔOCV) according to temperature or time period; and (d) a screening step, screening low-voltage defective batteries based on the corrected dispersion of the voltage drop.

[0037] Because the method for screening low-voltage defective batteries according to the present invention removes potential defective batteries exhibiting outlier data in the individual cell process constituting the activation process, clusters the batteries with common factors after removing such potential defective batteries, measures the voltage drop of each cluster, and corrects the dispersion of the measured voltage drop, it can enhance the detectability of low-voltage defective batteries and reduce the risk of over-detection or under-detection due to manufacturing deviations when screening low-voltage defective batteries by obtaining an improved dispersion of the total voltage drop.

[0038] The pre-filtering step (a) is a process that pre-removes potentially defective cells, collects data during the secondary cell activation process, and removes secondary cells exhibiting outliers from the collected data in real time. By doing so, the distribution of voltage drop during defective cell screening can be improved.

[0039] The activation process of a secondary battery is the step of activating the battery so that the assembled lithium secondary battery can be used. The assembly process of a secondary battery involves housing the electrode assembly, including the positive electrode, separator, and negative electrode, inside the battery casing and sealing it after injecting electrolyte. The completion of this assembly process does not mean the battery is immediately usable. To make the battery usable, it is necessary to perform processes such as inducing electrochemical reactions between the electrodes and electrolyte through charging, releasing the gases generated during this process to the outside of the battery, and stabilizing the battery; this series of processes is called the activation process.

[0040] The activation process of this secondary battery includes individual unit processes, such as a pre-aging process that ages the battery so that the electrolyte is immediately immersed in the electrode assembly after the secondary battery is assembled, an initial charging process that charges the battery until a set SOC is reached, an aging process that ages the initially charged battery, and a degassing process that releases gases generated during the initial charging and aging processes to the outside of the battery. The aging process may include a high-temperature aging process or a room-temperature aging process, or both.

[0041] In the pre-filtering step (a) of the present invention, in order to remove potentially defective batteries in advance while performing such a separate cell process, it includes a process of collecting data on the open-circuit voltage (OCV) of the secondary battery or the voltage drop of the battery and removing secondary batteries indicating outlier data in real time.

[0042] In a specific example, the individual unit processes for collecting data and removing outlier data may include at least one of an initial charging process and an aging process. To illustrate this with a specific example, when performing an initial charging process or an aging process on multiple secondary batteries, the open-circuit voltage of the secondary batteries is measured and collected as data, secondary batteries exhibiting outliers in the collected data are removed in real time, and subsequent processes are performed on the remaining secondary batteries after removal.

[0043] In the pre-filtering step (a), the method used to remove outliers employs a time series anomaly detection algorithm. Time series anomaly detection is a technique for discovering rare patterns or target objects that deviate from or show signs of deviation from the general patterns of other data at past or similar points in time series data. Such anomaly detection algorithms can use known algorithms, and in embodiments of the invention, can be selected from one of the following groups: control graphs, randomized forests, dynamic time warps, TAnoGAN (time series anomaly detection via generative adversarial networks), MAD-GAN (multivariate anomaly detection on time series data via generative adversarial networks), USAD (unsupervised anomaly detection on multivariate time series), LSTM (Long Short-Term Memory) + AE, and LSTM (Long Short-Term Memory) + CNN (Convolutional Neural Network).

[0044] Figure 2 and Figure 3 The diagram illustrates the cumulative probability distribution of voltage drop (ΔOCV) over time before and after a secondary battery exhibiting outlier data is removed via a pre-filtering step, according to an exemplary embodiment of the present invention.

[0045] refer to Figure 2 and Figure 3 After the pre-filtering step, it can be confirmed that the distribution of the pressure drop is improved, as clearly shown at the point where the slope of the cumulative probability distribution changes.

[0046] Furthermore, with the pre-filtering step, the distribution of voltage drop over both 5 days (5-day ΔOCV) and 3 days (3-day ΔOCV) is improved, thus low-voltage screening can be performed even after 3 days. However, without pre-filtering, such as Figure 3 As shown, when 3 days have passed, the points where the slope of the cumulative probability distribution changes are displayed at two points, thus blurring the boundary between good and defective products. Therefore, the method for detecting and screening low voltage according to the present invention not only has excellent detectability but also reduces the time required to screen low-voltage defective batteries.

[0047] Clustering step (b) is a step of clustering secondary batteries that have similar characteristics or similar records in the secondary battery manufacturing process. Clustering multiple secondary batteries based on similar characteristics or records means that there is no manufacturing deviation in the same cluster, thus improving the problem of increased dispersion caused by manufacturing deviation and blurring the boundary between good and defective products.

[0048] In a specific example, clustering step (b) can cluster secondary batteries with consistent criteria selected from groups consisting of: LOT cells, tray cells, aging temperature, charge / discharge temperature, and aging period. Because these factors are manufacturing deviations that affect the distribution of voltage drop, it is preferable to cluster secondary batteries based on these factors.

[0049] The calibration step (c) measures the pressure drop of each cluster and corrects the distribution of the pressure drop based on temperature or time period.

[0050] For each cluster, the correction step (c) includes a process of deriving the distribution of voltage drop by monitoring the open-circuit voltage (OCV) while aging multiple cells under constant temperature conditions, and a process of correcting the distribution of voltage drop.

[0051] Specifically, the process of deriving the distribution of voltage drop can be as follows: measure the open circuit voltage (OCV1) of the battery at the start of aging, and measure the open circuit voltage (OCV2) of the battery after a certain period of time T has passed, thereby calculating the voltage drop over a certain period of time (ΔOCV=OCV1-OCV2), and thus plotting the voltage drop of each of the multiple batteries.

[0052] However, it is known that the voltage drop of a battery is affected by the measurement period (the time interval between measuring OCV1 and OCV2) and the aging temperature. That is, even with the same battery, when the time interval T between voltage drop measurements becomes longer, or when the voltage drop is measured under high aging temperature conditions, the voltage drop will appear to be less than the actual value.

[0053] Therefore, in order to eliminate the influence of pressure drop measurement conditions (temperature, time period) on the pressure drop, the present invention includes the following process: correcting the difference in pressure drop based on the difference in aging temperature and measurement time period (the time interval for measuring OCV1 and OCV2).

[0054] The aging process can be one or more selected from the aging processes included throughout the battery manufacturing process. Specifically, it can be one or more selected from the aging process prior to the degassing process, the aging process after full charge and full discharge, and the aging process after charging for shipment.

[0055] In one embodiment of the invention, linear regression or machine learning methods can be used as a specific method for correcting the dispersion of pressure drop. Linear regression is an analytical method that calculates and measures a model between observed variables based on a large amount of back data, and is one of the representative statistical analysis methods used when deriving appropriate predicted values ​​based on changes in variables. Machine learning is one of the specific methods of artificial intelligence, in which a computer learns from a large amount of back data and discovers patterns and correlations in large-scale data through experience. The technical configurations of linear regression and machine learning are well known in the art, and therefore detailed descriptions are omitted.

[0056] Figure 4 It is a graph showing the distribution of the pressure drop without the correction steps according to the invention, and Figure 5 These are graphs showing the distribution of the voltage drop after the correction steps according to the invention. Referring to these graphs, it can be seen that the deviation of the voltage drop (ΔOCV) is significantly reduced after the correction steps according to the invention.

[0057] Even for batteries of the same model, voltage drop is affected by temperature or time of day. Using linear regression or machine learning to study the correlation between temperature or time of day and voltage drop, and by making appropriate corrections, the dispersion of voltage drop can be reduced. Furthermore, as a result of reducing voltage drop dispersion in this way, outliers in the voltage drop are better revealed, which can aid in the detection of low-voltage batteries.

[0058] Screening step (d) is a step of screening low-voltage defective batteries based on the distribution of the corrected voltage drop.

[0059] In a specific example, the screening step (d) is a statistical method such as a cumulative probability distribution, which derives a threshold within the spread of voltage drop and screens secondary batteries with voltage drops exceeding the threshold as low-voltage defective batteries. The spread of voltage drop is data listing the voltage drop of each of the large number of secondary batteries obtained through steps (a) to (c).

[0060] In another specific example, the screening step (d) can be to screen low-voltage defective batteries from the corrected scatter by using an anomaly detection algorithm, and the anomaly detection algorithm can be one of the following: Local Outlier Factor (LOF), Isolation Forest, and OC Support Vector Machine (SVM).

[0061] In one embodiment of the invention, screening step (d) screens low-voltage defective cells for each cluster. In this case, an anomaly detection algorithm can be used to screen defective cells based on the voltage drop distribution data corrected in step (c).

[0062] In another exemplary embodiment of the invention, the screening step (d) screens low-voltage defective batteries after collecting the dispersion data of voltage drop for each cluster, corrected by step (c). In this case, the dispersion data of each voltage drop for each cluster, corrected in step (c), is collected as a single set, and defective batteries are screened based on the collected dispersion data using an anomaly detection algorithm. In this embodiment, because defective batteries are screened all at once based on the collected dispersion data using an anomaly detection algorithm, it has the effect of screening defective batteries faster than screening defective batteries by cluster.

[0063] The screening step (d) of the present invention can be performed after the aging (the next step after the activation process) for shipment is completed and by steps (a) to (c), and because the dispersion of the voltage drop of the battery is improved, defective batteries are screened by an anomaly detection algorithm based on the improved dispersion, thus improving the reliability of the detection.

[0064] As described above, this disclosure and accompanying drawings disclose preferred embodiments of the invention. Although specific terminology is used, it is only for the general purpose of readily explaining the technical content of the invention and aiding in understanding the invention, and is not intended to limit the scope of the invention. Besides the embodiments disclosed herein, it will be apparent to those skilled in the art that other modifications based on the inventive concept can be implemented.

Claims

1. A method for screening low-voltage defective batteries, comprising: A pre-filtering step collects data during the activation process of multiple secondary batteries and removes secondary batteries with outlier data in real time. The clustering step involves clustering based on the similar characteristics or records of multiple pre-filtered secondary batteries; The calibration step involves measuring the pressure drop for each cluster and correcting the distribution of the pressure drop based on temperature or time period. The screening step uses the distribution of the corrected voltage drop to screen for low-voltage defective batteries.

2. The method for screening low-voltage defective batteries according to claim 1, wherein... The pre-filtering step uses a time series anomaly detection algorithm to remove outliers.

3. The method for screening low-voltage defective batteries according to claim 1, wherein... The clustering step clusters secondary batteries that have consistent criteria selected from the following groups: LOT cells, tray cells, aging temperature, charge / discharge temperature, and aging period.

4. The method for screening low-voltage defective batteries according to claim 1, wherein... The correction step includes correcting the distribution of the pressure drop using linear regression or machine learning methods.

5. The method for screening low-voltage defective batteries according to claim 1, wherein... For each individual cell process, the pre-filtering step includes the process of collecting the open-circuit voltage or voltage drop of the secondary battery as data.

6. The method for screening low-voltage defective batteries according to claim 5, wherein... The individual unit process includes at least one of the following: an initial charging process, a room temperature aging process, and a high temperature aging process.

7. The method for screening low-voltage defective batteries according to claim 2, wherein... The time series anomaly detection algorithm is selected from one of the following groups: control graph, random pruning forest, dynamic time warp, TAnoGAN, MAD-GAN, USAD, LSTM+AE, and LSTM+CNN.

8. The method for screening low-voltage defective batteries according to claim 1, wherein... The screening step uses an anomaly detection algorithm to screen for low-voltage defective batteries in the corrected scattering.

9. The method for screening low-voltage defective batteries according to claim 8, wherein... The anomaly detection algorithm is selected from one of the following groups: local outlier factor, isolated forest, and OC support vector machine.

10. The method for screening low-voltage defective batteries according to claim 1, wherein... The screening steps are performed after the aging process for shipment is completed.

11. The method for screening low-voltage defective batteries according to claim 1, wherein... The screening step filters for low-voltage defective batteries for each cluster.

12. The method for screening low-voltage defective batteries according to claim 1, wherein... The screening step filters low-voltage defective batteries after collecting scatter data of the voltage drop for each cluster, corrected by the correction step.

Images