System using autoencoder for detecting abnormal battery degradation

Abstract

Disclosed is a system using autoencoder for detecting abnormal battery degradation, the system according to an embodiment comprising a database for storing a first dataset that comprises multiple pieces of data received from an input unit, and a processor for training an artificial intelligence model and detecting abnormal battery degradation by means of the trained artificial intelligence model, wherein the processor comprises: a data preprocessing unit for converting the first dataset to mimic images; a feature training unit for training the artificial intelligence model to extract features on the basis of the images generated by the data preprocessing unit and generate virtual health indicators (VHIs) of a battery; and an abnormality detection unit for detecting abnormal battery degradation by generating, on the basis of the artificial intelligence model trained by the feature training unit, battery VHIs from a second dataset input by a user and clustering, on the basis of a clustering algorithm, the generated battery VHIs.

Description

Battery Abnormal Degradation Detection System Using an Autoencoder

[0001] One disclosed embodiment relates to a battery abnormal degradation detection system that detects abnormal degradation of a battery by training an artificial intelligence model based on battery data.

[0002] Lithium-ion batteries are characterized by high energy density and long lifespan, and are used critically in various fields such as electric vehicles, energy storage systems (ESS), and portable electronic devices. However, batteries undergo degradation during repeated charging and discharging cycles, which can lead to performance degradation and safety issues. Therefore, early detection of abnormal conditions during the initial degradation phase is a critical task for ensuring battery reliability and safety.

[0003] Conventional technologies for detecting abnormal battery degradation are broadly classified into threshold-based detection, statistical degradation models, and deep learning-based detection. Threshold-based detection distinguishes between normal and abnormal states by setting specific thresholds for battery data. While simple, this method has the disadvantage that threshold setting relies on empirical judgment and fails to reflect the diverse degradation patterns of the battery. Statistical degradation models detect anomalies by analyzing battery capacity degradation curves and degradation trajectories. However, this method is based on the assumption that battery degradation patterns are identical across all environments and suffers from a problem where accuracy degrades significantly depending on data quality and volume. Deep learning-based detection is a method that automatically extracts features from data and detects abnormal states. While this method holds high potential, it requires large-scale, high-quality data and has limitations, such as performance degradation in limited data environments and the inability to effectively model uncertainty.

[0004] These conventional technologies share several common drawbacks. First, most existing methods are manual, requiring humans to manually analyze data and extract features, which is time-consuming and labor-intensive. Second, their high reliance on labeled data leads to significantly degraded anomaly detection performance in limited data environments. Third, they fail to adequately account for the diverse usage environments and degradation mechanisms of batteries, making it difficult to generalize them to various battery types or conditions. Finally, they pose a risk of causing serious damage or failure by failing to detect abnormal conditions during the early degradation stages.

[0005] In order to overcome the limitations of the aforementioned prior art, one disclosed embodiment relates to a system that extracts features from battery data using an autoencoder and can rapidly and accurately detect abnormal degradation states of a battery even in environments where initial data or labels are scarce through unsupervised learning clustering.

[0006] A battery abnormal degradation detection system using an autoencoder according to a disclosed embodiment comprises: a database storing first set data including a plurality of data received from an input unit; and a processor that trains an artificial intelligence model and detects abnormal degradation of the battery through the trained artificial intelligence model. The processor comprises: a data preprocessing unit that converts the first set data into a patterned image; a feature learning unit that extracts features based on the image generated by the data preprocessing unit and proceeds with training the artificial intelligence model to generate a battery virtual health indicator (VHI); and an abnormality detection unit that generates a battery virtual health indicator (VHI) from second set data input by a user based on the artificial intelligence model trained by the feature learning unit, and detects abnormal degradation of the battery by performing clustering of the generated battery virtual health indicator (VHI) based on a clustering algorithm.

[0007] The first and second set data above may include battery data consisting of voltage, resistance, temperature, discharge capacity, charge capacity, and current.

[0008] The above data preprocessing unit receives the first set of data and can convert multidimensional data into patterned images while maintaining the temporal and spatial correlation of the data.

[0009] The above feature learning unit can train the artificial intelligence model using first set data including patterned images generated by the data preprocessing unit as input, and generate a battery virtual state indicator (VHI) based on the trained artificial intelligence model.

[0010] The above anomaly detection unit receives second set data from a user based on an artificial intelligence model learned by the feature learning unit to generate a battery virtual state index (VHI), clusters the generated battery virtual state index (VHI) based on a clustering algorithm, and determines whether there is abnormal degradation of the battery based on the clustering result.

[0011] A method for detecting abnormal degradation of a battery using an autoencoder according to another disclosed embodiment comprises: storing a plurality of data received from an input unit as a first set of data; preprocessing the first set of data; inputting the preprocessed first set of data into the artificial intelligence model to perform training of the artificial intelligence model to generate a VHI; generating a battery virtual state indicator (VHI) from a second set of data input by a user based on the trained artificial intelligence model; performing clustering of the generated battery virtual state indicator (VHI) based on a clustering algorithm; and determining whether the battery is abnormal based on the clustering.

[0012] The above artificial intelligence model can learn a latent space representing the degradation state of a battery and generate a battery virtual state index (VHI) based on it.

[0013] The above artificial intelligence model may be characterized by using a mixture distribution (Laplace and Student's t-distribution) as a prior probability.

[0014] The above artificial intelligence model may include at least one of a Bayesian Convolutional Autoencoder (BCAE), a Variational Autoencoder (VAE), and Deep Embedding Clustering (DEC).

[0015] The above clustering algorithm may include at least one of the Gaussian Mixture Model (GMM), DBSCAN (Density-Based Spatial Clustering of Applications with Noise), or Spectral Clustering algorithm.

[0016] The above clustering algorithm can cluster battery virtual state indicators (VHI) and detect abnormal battery conditions based on cluster boundary values.

[0017] Determining whether the above battery is abnormal may include performing labeling to distinguish between a normal state and an abnormal state based on battery data included in the first set data, and determining each cluster of the battery virtual state indicator (VHI) as a normal state or an abnormal state based on the labeling.

[0018] A battery abnormal degradation detection system using an autoencoder according to another disclosed embodiment comprises: a database storing a first set of data including a plurality of data received from an input unit; a processor that preprocesses the first set of data, inputs the preprocessed first set of data into an artificial intelligence model to generate a VHI, performs clustering on the generated VHI based on a clustering algorithm, and detects abnormal degradation of the battery based on the clustering performed; and an output unit that outputs a battery abnormal degradation detection result from a second set of data input by a user based on the learned artificial intelligence model.

[0019] A battery abnormal degradation detection system according to the disclosed embodiment may include providing a warning about an abnormal condition when abnormal degradation of the battery is detected.

[0020] A battery abnormal degradation detection system according to the disclosed embodiment may include analyzing a battery degradation pattern over time based on the first set of data and the second set of data, and providing an analysis result.

[0021] A battery abnormal degradation detection system using an autoencoder according to a disclosed embodiment utilizes an Adaptive Mixture Distribution Bayesian Convolutional Autoencoder (AMDBAE) to automatically extract features from battery data and effectively detect abnormal degradation states through unsupervised learning-based clustering, thereby having the following advantages.

[0022] Compared to existing threshold-based detection methods, statistical degradation models, and deep learning-based detection methods, the disclosed battery abnormal degradation detection system enables learning even in environments with insufficient labels and can provide high accuracy and reliability using only initial data. While existing technologies rely on large-scale data and labeling or fail to adequately reflect complex degradation mechanisms, the disclosed battery abnormal degradation detection system significantly improves universality and adaptability by modeling various battery degradation patterns based on mixed distributions. Furthermore, while existing technologies had issues such as difficulty in generalization depending on the usage environment or failure to detect initial abnormal states, the disclosed battery abnormal degradation detection system enables stable detection by quantifying data uncertainty.

[0023] The disclosed battery abnormal degradation detection system can provide improved performance in terms of data analysis efficiency and accuracy compared to existing technologies. Since battery data is processed based on an autoencoder, user-dependent feature extraction processes are not required, thereby saving time and costs. Furthermore, through unsupervised learning clustering, it can automatically detect abnormal battery conditions even in unlabeled data. In particular, by automatically updating the ratio parameters of the mixture distribution during the training process, it efficiently handles data diversity and can provide reliable results even with small amounts of data.

[0024] Finally, the battery abnormal degradation detection system according to the disclosed embodiment has high versatility applicable to various battery types and usage environments. This characteristic can be effectively integrated into a Battery Management System (BMS) and a battery safety monitoring solution, and can provide an important technical solution for ensuring the reliability and safety of the battery.

[0025] FIG. 1 is a diagram schematically illustrating a battery abnormal degradation detection system using a disclosed autoencoder.

[0026] FIG. 2 is a control block diagram of a battery abnormal degradation detection system using a disclosed autoencoder.

[0027] Figure 3 is a flowchart illustrating the process of a processor receiving aggregate data and detecting abnormal battery degradation.

[0028] Figure 4 is a diagram for specifically explaining the operation of the data preprocessing unit.

[0029] Figure 5 is a diagram for specifically explaining the operation of the feature learning unit.

[0030] Figure 6 is a diagram for specifically explaining the operation of the abnormal detection unit.

[0031] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. However, the technical concept of the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided to ensure that the disclosed content is thorough and complete and to sufficiently convey the concept of the present invention to those skilled in the art.

[0032] In this specification, when a component is described as being on another component, it means that it may be formed directly on the other component or that a third component may be interposed between them. Additionally, in the drawings, shapes and sizes are exaggerated for the effective illustration of the technical content.

[0033] Additionally, although terms such as first, second, third, etc., have been used to describe various components in the various embodiments of this specification, these components should not be limited by such terms. These terms are used merely to distinguish one component from another. Accordingly, what is referred to as the first component in one embodiment may be referred to as the second component in another embodiment. Each embodiment described and illustrated herein also includes its complementary embodiment. Furthermore, in this specification, "and / or" is used to mean including at least one of the components listed before and after it.

[0034] In the specification, singular expressions include plural expressions unless the context clearly indicates otherwise. Furthermore, terms such as "include" or "have" are intended to specify the existence of the features, numbers, steps, components, or combinations thereof described in the specification, and should not be understood as excluding the existence or addition of one or more other features, numbers, steps, components, or combinations thereof. Additionally, in this specification, "connection" is used to include both indirectly connecting multiple components and directly connecting them.

[0035] In addition, in describing the present invention below, if it is determined that a detailed description of related known functions or configurations could unnecessarily obscure the essence of the invention, such detailed description will be omitted.

[0036] FIG. 1 is a diagram schematically illustrating a battery abnormal degradation detection system using a disclosed autoencoder.

[0037] Referring to FIG. 1, a battery abnormal degradation detection system (1) using an autoencoder according to one disclosed embodiment can be implemented as a computer or portable terminal capable of collecting battery data consisting of voltage, resistance, temperature, discharge capacity, charge capacity, and current from external devices (3, 4) through a communication network (2). Here, the computer includes, for example, a laptop, desktop, laptop, tablet PC, slate PC, etc. equipped with a web browser, and the portable terminal may include, for example, a wireless communication device that ensures portability and mobility, all kinds of handheld-based wireless communication devices such as PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), WiBro (Wireless Broadband Internet) terminal, smartphone, etc., and wearable devices such as watches, rings, bracelets, anklets, necklaces, glasses, contact lenses, or head-mounted devices (HMDs).

[0038] The battery abnormal degradation detection system (1) can collect battery data consisting of voltage, resistance, temperature, discharge capacity, charge capacity, and current from a battery management system (BMS) (3) and a personal computer (4).

[0039] Specifically, the Battery Management System (BMS) is configured to measure battery data and can collect battery data including voltage, resistance, temperature, discharge capacity, charge capacity, and current data.

[0040] The personal computer (4) is configured to collect battery data held by users in their individual terminals, and is configured to explain that not only current but also past battery data can be used as learning data.

[0041] Battery data collected from each device (3, 4) is transmitted to the battery abnormal degradation detection system (1) via the communication network (2).

[0042] The communication network (2) is a channel for receiving data from the aforementioned configurations (3, 4), and the battery abnormal degradation detection system (1) receives data from the communication network (2). The multiple battery data received in this way are stored in the database (12, see FIG. 2) of the battery abnormal degradation detection system (1). The battery abnormal degradation detection system (1) can detect abnormal degradation of the battery by preprocessing the data stored in the database (2) and then training an artificial intelligence model. The specific operation and method by which the battery abnormal degradation detection system (1) detects abnormal degradation of the battery will be described later through other drawings.

[0043] Meanwhile, in addition to the configuration (3, 4) illustrated in FIG. 1, the battery abnormal degradation detection system (1) can receive battery data from various devices capable of storing data, such as smartphones, laptops, or tablet PCs, and can also receive various types of battery data from web or cloud servers. Additionally, the battery abnormal degradation detection system (1) can collect voice data directly from a user through a peripheral device such as a USB (Universal Serial Bus) without using a communication network (2).

[0044] FIG. 2 is a control block diagram of a battery abnormal degradation detection system using a disclosed autoencoder.

[0045] Referring to FIG. 2, the battery abnormal degradation detection system (1) includes an input unit (9) for collecting battery data, a communication unit (11) for communicating with a communication network (2), a database (12) for storing a set of data (hereinafter referred to as the first set of data) containing a plurality of battery data collected from the input unit (9) or the communication unit (11) and an artificial intelligence model for detecting battery abnormal degradation from the first set of data, a processor (10) for preprocessing the first set of data stored in the database (12), training the artificial intelligence model, and generating a battery virtual health indicator (VHI) based on the trained artificial intelligence model, and an output unit (13) for outputting a battery abnormal degradation detection result based on newly received set of data (hereinafter referred to as the second set of data) by a user.

[0046] Specifically, the input unit (9) may include hardware devices such as various buttons or switches, pedals, keyboards, mice, trackballs, various levers, handles or sticks to receive user input.

[0047] For example, the input unit (9) can receive whether to train an artificial intelligence model using the first set of data, or to use the trained artificial intelligence model to detect abnormal degradation of the battery based on the second set of data.

[0048] The communication unit (11) may include various configurations that allow the battery abnormal degradation detection system (1) to communicate with external devices (3, 4 in FIG. 1), and may include, for example, at least one of a short-range communication module, a wired communication module, and a wireless communication module.

[0049] The short-range communication module may include various short-range communication modules that transmit and receive signals using a wireless communication network at short range, such as a Bluetooth module, an infrared communication module, an RFID (Radio Frequency Identification) communication module, a WLAN (Wireless Local Access Network) communication module, an NFC communication module, and a Zigbee communication module.

[0050] Wired communication modules may include various wired communication modules such as Local Area Network (LAN) modules, Wide Area Network (WAN) modules, or Value Added Network (VAN) modules, as well as various cable communication modules such as USB (Universal Serial Bus), HDMI (High Definition Multimedia Interface), DVI (Digital Visual Interface), RS-232 (recommended standard 232), power line communication, or POTS (plain old telephone service).

[0051] In addition to Wi-Fi modules and WiBro (Wireless broadband) modules, the wireless communication module may include wireless communication modules that support various wireless communication methods such as GSM (global System for Mobile Communication), CDMA (Code Division Multiple Access), WCDMA (Wideband Code Division Multiple Access), UMTS (universal mobile telecommunications system), TDMA (Time Division Multiple Access), and LTE (Long Term Evolution).

[0052] The database (12) stores not only various set data collected by the input unit (9) or the communication unit (11), but also an algorithm for preprocessing the set data, an artificial intelligence model for generating a battery virtual state indicator (VHI) from the preprocessed set data, an algorithm for clustering the battery virtual state indicator (VHI), an algorithm for labeling to distinguish between normal and abnormal states, an artificial intelligence model to learn through the set data, and an artificial intelligence model that has completed learning.

[0053] The database (12) may be implemented in at least one of a non-volatile memory device such as a cache, ROM (Read Only Memory), PROM (Programmable ROM), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), and flash memory, a volatile memory device such as RAM (Random Access Memory), or a storage medium such as a hard disk drive (HDD) and CD-ROM, but is not limited thereto. The database (12) may be a memory implemented as a separate chip from the processor (10) as shown in FIG. 2, but may also be implemented as a single chip with the processor (10) as needed.

[0054] The output unit (13) outputs data including the result of detecting abnormal degradation of the battery based on the second set data, which is the result estimated by the processor (10). For example, the output unit (13) can output the virtual state index (VHI) of the battery derived from the battery data through a display, and the result including the battery degradation status and abnormality analyzed by the artificial intelligence model through a user interface.

[0055] For the aforementioned operation, the output unit (13) may include various hardware devices such as a Digital Light Processing (DLP) panel, a Plasma Display Panel, a Liquid Crystal Display (LCD) panel, an Electro Luminescence (EL) panel, an Electrophoretic Display (EPD) panel, an Electrochromic Display (ECD) panel, a Light Emitting Diode (LED) panel, or an Organic Light Emitting Diode (OLED) panel.

[0056] Meanwhile, the output unit (13) may include a device that is a GUI (Graphical User Interface), i.e., software, such as a touch pad, for user input. The touch pad may be implemented as a touch screen panel (TSP) and form a layered structure with the input unit (9).

[0057] The processor (10) controls the overall system of the battery abnormal degradation detection system (1). To this end, the processor (10) may execute an algorithm or a program that reproduces the algorithm to control the configuration shown in FIG. 2. That is, the processor (10) refers to a data processing device embedded in hardware having a physically structured circuit to perform a function expressed by code or instructions included in the program. Examples of such data processing devices embedded in hardware include, but are not limited to, processing devices such as a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), and a Graphics Processing Unit (GPU). The processor (10) may be implemented as one or more chips.

[0058] The processor (10) can be divided into a data preprocessing unit (110) that converts a first set of data into a patterned image in software form, a feature learning unit (120) that extracts features based on the image generated by the data preprocessing unit and proceeds with learning an artificial intelligence model to generate a battery virtual health indicator (VHI), and an anomaly detection unit (130) that generates a battery virtual health indicator (VHI) from a second set of data input by a user based on the artificial intelligence model learned by the feature learning unit and detects abnormal degradation of the battery by performing clustering of the battery virtual health indicator (VHI) based on a clustering algorithm.

[0059] A detailed description of the operation of the software-separated processor (10) will be provided later through other drawings below.

[0060] Meanwhile, the disclosed battery abnormal degradation detection system (1) may include various additional configurations in addition to the configuration described above, and each configuration of FIG. 2 may be changed or combined in various ways depending on the required operation.

[0061] FIG. 3 is a flowchart illustrating the process in which a processor (10) of a disclosed battery abnormal degradation detection system (1) receives and preprocesses aggregate data, generates a battery virtual health indicator (VHI) through an artificial intelligence model, and clusters the generated battery virtual health indicator to detect battery abnormal degradation.

[0062] To avoid redundant explanations below, the process of the battery abnormal degradation detection system (1) preprocessing aggregate data is explained by referring to FIG. 3 and FIG. 4 simultaneously.

[0063] Referring to FIG. 3, the battery abnormal degradation detection system (1) receives aggregate data (100).

[0064] Here, the set data (hereinafter referred to as the first set data) includes battery data consisting of voltage, resistance, temperature, discharge capacity, charge capacity, and current received from the input unit (9) or the communication unit (11). The first set data may be data measured by each sensor that measures voltage, resistance, temperature, discharge capacity, charge capacity, and current in real time, and is used to train an artificial intelligence model in the feature learning unit (120).

[0065] Specifically, the first set of data can be collected from data measured in real time through measuring equipment such as a battery management system (BMS) (3). The battery management system (BMS) utilizes various sensors to measure battery data such as voltage, resistance, temperature, discharge capacity, charge capacity, and current. Voltage is measured through a voltage sensor, resistance through an impedance analyzer or a constant current circuit, temperature through a thermocouple, an RTD (Resistance Temperature Detector), or an NTC thermistor (Negative Temperature Coefficient Thermistor), discharge and charge capacities through a current sensor and a data logger, and current through a Hall Effect Current Sensor or a Shunt Resistor.

[0066] The battery abnormal degradation detection system (1) preprocesses the received set data (101).

[0067] To avoid redundant explanations below, the process of preprocessing the set data received by the battery abnormal degradation detection system (1) is described in detail with reference to FIG. 3 and FIG. 4.

[0068] The battery abnormal degradation detection system (1) preprocesses the battery data among the first set of data by normalizing it so that more accurate and useful information can be obtained when extracting a feature vector from the battery data (101).

[0069] Specifically, the normalization process adjusts multidimensional data such as battery voltage, resistance, temperature, discharge capacity, charge capacity, and current to the same scale, maintaining balance between data and improving model training efficiency.

[0070] In this process, each data point can be processed through Min-Max Normalization or Z-Score Standardization, and voltage, resistance, and temperature data are normalized after removing outliers using a noise removal filter.

[0071] Normalized battery data is converted into patterned images (Mimic Images) or low-dimensional feature vectors used for training artificial intelligence models, enabling accurate analysis of battery degradation status and detection of abnormalities.

[0072] The battery abnormal degradation detection system (1) trains an artificial intelligence model based on preprocessed data (102) and generates a battery virtual health indicator (VHI) from the trained artificial intelligence model (103).

[0073] To avoid redundant explanations below, with reference to FIG. 3 and FIG. 5, the process of a battery abnormal degradation detection system (1) training an artificial intelligence model (102) and generating a battery virtual state indicator (VHI) (103) is described in detail.

[0074] The battery abnormal degradation detection system (1) trains an artificial intelligence model based on preprocessed battery data (102).

[0075] The artificial intelligence model learns the temporal and spatial characteristics of battery data to derive feature vectors necessary for generating the Battery Virtual Health Indicator (VHI).

[0076] In the disclosed embodiment, the artificial intelligence model is a feature extraction model that compresses battery data into a low-dimensional latent space and learns degradation patterns in the space, and may be at least one of a Bayesian Convolutional Autoencoder (BCAE), a Variational Autoencoder (VAE), and Deep Embedding Clustering (DEC).

[0077] In the disclosed embodiment, the artificial intelligence model may apply a modified Bayes by Backprop algorithm for parameter optimization during the learning process.

[0078] The modified Bayes by Backprop algorithm utilizes Monte Carlo (MC) sampling and the Gumbel-Softmax technique simultaneously to approximate the Kullback-Leibler (KL) divergence between the prior and backward distributions of the mixture distribution, thereby efficiently learning network parameters and enabling robust feature extraction by quantifying the uncertainty of the model even in a limited data environment.

[0079] In the disclosed embodiment, the learning rate and mixture weights of the artificial intelligence model are dynamically optimized during the learning process. The mixture weights are set based on the Laplace distribution and Student's t-distribution, and are adjusted to effectively reflect the complex degradation patterns and distributions of the battery data. This makes the boundary between the normal state and the abnormal state more clearly defined and facilitates the detection of abnormal data.

[0080] In the disclosed embodiment, the artificial intelligence model can use the Mean Squared Error Loss (MSE Loss) and the KL-Divergence Loss simultaneously as loss functions during the learning process.

[0081] MSE Loss is defined as a method of calculating the average by squaring the difference between each element of two data points, and can minimize the difference between the input data and the reconstructed data.

[0082] KL-Divergence Loss regularizes the distribution of data in the latent space to approximate the prior distribution, thereby effectively reflecting the structural patterns of the data in the latent space.

[0083] In other words, MSE Loss focuses on minimizing numerical differences during the reconstruction process, while KL-Divergence Loss helps the model learn various data distributions while maintaining its generalization performance.

[0084] In the disclosed embodiment, the artificial intelligence model learns the key features of the input data through a combination of MSE Loss and KL-Divergence Loss, and at the same time improves reconstruction accuracy, models the distribution of the data through a latent space.

[0085] Through this, the battery abnormal degradation detection system (1) can quantitatively evaluate the degradation state of the battery and effectively classify the normal state and the abnormal state.

[0086] The battery abnormal degradation detection system (1) generates a battery virtual state indicator (VHI) from a second set of data input by a user based on an artificial intelligence model learned in this way (103).

[0087] The Battery Virtual Health Indicator (VHI) is an indicator that quantitatively represents the degradation state of a battery and plays an important role in distinguishing between normal and abnormal states.

[0088] For example, the Battery Virtual Health Indicator (VHI) is quantified by reflecting degradation patterns and anomalies in battery data, and is designed to accurately detect abnormal conditions even in the early stages of degradation.

[0089] The battery abnormal degradation detection system (1) clusters the generated battery virtual state indicator (VHI) (104) and detects battery abnormal degradation (105).

[0090] To avoid redundant explanations below, with reference to FIG. 3 and FIG. 6, the process of a battery abnormal degradation detection system (1) clustering battery virtual state indicators (VHI) (104) and detecting battery abnormal degradation (105) is described in detail.

[0091] The battery abnormal degradation detection system (1) inputs the battery virtual health indicator (VHI) into a clustering algorithm to distinguish and cluster the battery's condition (104).

[0092] For example, data from normal-state batteries tends to have low VHI values ​​and be clustered in the normal-state cluster, whereas data from degraded batteries has high VHI values ​​and may be included in the abnormal-state cluster. Clustering algorithms can effectively detect the degradation state of batteries by learning these differences in distribution.

[0093] In the disclosed embodiment, the clustering algorithm may be at least one of the Gaussian Mixture Model (GMM), DBSCAN (Density-Based Spatial Clustering of Applications with Noise), or Spectral Clustering algorithm.

[0094] The Gaussian Mixture Model (GMM) approximates the distribution of battery data as a Gaussian mixture model to calculate the probability that each data point belongs to a specific cluster.

[0095] DBSCAN is a density-based clustering algorithm that forms clusters using differences in the density of data points and can separate noise and outliers into separate clusters.

[0096] Spectral Clustering uses a graph-based approach to learn similarities between data and can effectively distinguish clusters even in complex data structures.

[0097] In the disclosed embodiment, the clustering algorithm is based on unsupervised learning and can identify abnormal degradation behavior of the battery early by analyzing the distribution of battery data without labeled data and classifying it into normal state clusters and abnormal state clusters.

[0098] The battery abnormal degradation detection system (1) determines whether the battery is abnormally degraded by analyzing the cluster to which each battery virtual state indicator (VHI) belongs based on the clustering results (105).

[0099] In the disclosed embodiment, the artificial intelligence model that has completed training can distinguish between normal data and abnormal data based on the Reconstruction Error.

[0100] Specifically, Reconstruction Error is calculated as the error between the input data and the reconstructed data, and can be analyzed using metrics such as Precision, Recall, and F1-Score to evaluate the model's anomaly detection performance.

[0101] A battery abnormal degradation detection system (1) can use a labeling function to classify each cluster as "normal state" or "abnormal state" based on the result of clustering battery virtual health indicators (VHI).

[0102] In the disclosed embodiments, the labeling function can label clusters or data points according to a predefined threshold or specific mathematical rules.

[0103] For example, the labeling function operates based on the battery virtual health indicator (VHI) value and clustering results, and specifically, can label it as "normal state" if the VHI value is below a certain threshold and as "abnormal state" if it exceeds the threshold.

[0104] As another example, labeling functions may be designed not only based on simple rule methods but also to reflect the statistical distribution of the data or the structural characteristics of the clusters.

[0105] Specifically, data in a normal state tends to have low VHI values ​​and be concentrated in the center of the clustering results, whereas data in an anomalous state exhibits characteristics such as high VHI values, location at cluster boundaries, or low density; therefore, a labeling function can perform accurate and reliable labeling by reflecting these distributional characteristics.

[0106] The battery abnormal degradation detection system (1) outputs the result of detecting abnormal degradation of the battery based on an artificial intelligence model that has completed training (106).

[0107] In the disclosed embodiment, the output unit (13) can visually provide battery status information determined to be normal or fault-detected through a display screen or notification device based on a user interface (UI).

[0108] Specifically, the output unit (13) displays a time series graph of the battery virtual state indicator (VHI) value along with battery data including the battery voltage, resistance, temperature, etc., and can output information related to the battery state including the time and cause when an abnormal state is detected.

[0109] Additionally, the output unit (13) may provide a visual or auditory warning to the user when the battery condition is detected to be abnormal, and the warning message may include additional information such as a detailed analysis result explaining the cause of the abnormal battery condition (excessive temperature rise) and necessary measures (battery replacement).

[0110] Finally, the output unit (13) may provide information related to the battery's usage history, including detected results and changes in state, in the form of a graph, CSV file, or report so that it can be used for long-term management of the battery's state.

[0111] The operation of the output unit (13) as described above is merely one example, and various variations are possible.

[0112] The battery abnormal degradation detection system (1) disclosed therein can improve the reliability and safety of the battery by enabling efficient battery replacement and maintenance based on the degradation state by evaluating the battery's condition in real time based on the battery virtual health indicator (VHI) derived from battery data and clustering results, and by detecting the battery's degradation behavior early.

Claims

1. A database storing first set data including a plurality of data received from an input unit; A processor that trains an artificial intelligence model and detects abnormal degradation of a battery through the trained artificial intelligence model; The above processor is, A data preprocessing unit that converts the above-mentioned first set data into a patterned image; A feature learning unit that extracts features based on an image generated by the data preprocessing unit and performs training of the artificial intelligence model to generate a battery virtual health indicator (VHI); and A battery abnormal degradation detection system comprising: an abnormality detection unit that detects abnormal degradation of a battery by generating a battery virtual state index (VHI) from a second set of data input by a user based on an artificial intelligence model learned by the above feature learning unit, and performing clustering of the battery virtual state index (VHI) generated based on a clustering algorithm.

2. In Paragraph 1, A battery abnormal degradation detection system characterized by the fact that the above first and second set data include battery data composed of voltage, resistance, temperature, discharge capacity, charge capacity, and current.

3. In Paragraph 1, The above data preprocessing unit is, A battery abnormal degradation detection system that receives the above-mentioned first set data and converts multidimensional data into a patterned image while maintaining the temporal and spatial correlation of the data.

4. In Paragraph 1, The above feature learning unit is, The artificial intelligence model is trained using first set data including patterned images generated in the above data preprocessing unit as input, and A battery abnormal degradation detection system that generates a battery virtual state indicator (VHI) based on the above-mentioned learned artificial intelligence model.

5. In Paragraph 1, The above abnormality detection unit is, Based on the artificial intelligence model learned by the above feature learning unit, a second set of data is received from the user to generate a battery virtual state index (VHI), and Clustering of the above-mentioned battery virtual state indicators (VHI) generated based on a clustering algorithm is performed, and A battery abnormal degradation detection system that determines whether the battery is abnormally degraded based on the results of performing the above clustering.

6. Store multiple data received from the input unit as the first set data; Preprocess the above first set data; The preprocessed first set of data is input into the artificial intelligence model to train the artificial intelligence model to generate VHI; Based on the above-mentioned learned artificial intelligence model, a battery virtual state indicator (VHI) is generated from the second set of data input by the user; Clustering of the above-mentioned battery virtual state indicators (VHIs) generated based on a clustering algorithm is performed; A method for detecting abnormal degradation of a battery, comprising determining whether the battery is abnormal based on the clustering above.

7. In Paragraph 6, The above artificial intelligence model is, A method for detecting abnormal degradation of a battery by learning a latent space representing the degradation state of the battery and generating a battery virtual state index (VHI) based thereon.

8. In Paragraph 7, A method for detecting abnormal degradation of a battery, characterized in that the above artificial intelligence model uses a mixture distribution (Laplace and Student's t-distribution) as a prior probability.

9. In Paragraph 7, The above artificial intelligence model is, A method for detecting abnormal degradation of a battery comprising at least one of a Bayesian Convolutional Autoencoder (BCAE), a Variational Autoencoder (VAE), and Deep Embedding Clustering (DEC).

10. In Paragraph 7, The above artificial intelligence model is, A method for detecting abnormal degradation of a battery, characterized by optimizing network parameters using a modified Bayes by Backprop algorithm, wherein the algorithm simultaneously utilizes a Monte Carlo Sampling module and a Gumbel-Softmax module.

11. In Paragraph 6, A method for detecting abnormal degradation of a battery, wherein the clustering algorithm comprises at least one of a Gaussian Mixture Model (GMM), DBSCAN (Density-Based Spatial Clustering of Applications with Noise), or Spectral Clustering algorithm.

12. In Paragraph 11, The above clustering algorithm is a battery abnormal degradation detection method that clusters battery virtual state indicators (VHI) and detects abnormal battery conditions based on cluster boundary values.

13. In Paragraph 6, Determining whether the above battery is abnormal is, Based on the battery data included in the first set of data above, labeling is performed to distinguish between normal and abnormal states, and A method for detecting abnormal degradation of a battery, which determines each cluster of the battery virtual state indicator (VHI) as normal or abnormal based on the above labeling.

14. A database storing first set data including a plurality of data received from an input unit; A processor that preprocesses the first set of data, inputs the preprocessed first set of data into an artificial intelligence model to generate a battery virtual state index (VHI), performs clustering of the battery virtual state index (VHI) based on a clustering algorithm, and detects abnormal degradation of the battery based on the result of the clustering; and A battery abnormal degradation detection system comprising: an output unit that outputs a battery abnormal degradation detection result from a second set of data input by a user based on a learned artificial intelligence model.

15. In Paragraph 12, A battery abnormal degradation detection system comprising: providing a warning of an abnormal condition when abnormal degradation of the battery is detected.

16. In Paragraph 12, A battery abnormal degradation detection system comprising: analyzing a battery degradation pattern over time based on the first set of data and the second set of data, and providing the analysis results.

Images