Battery service life prediction device and operation method therefor

Abstract

A battery service life prediction device according to one embodiment disclosed in the present document may comprise: an information acquisition unit for acquiring tabular data on each of a plurality of batteries, an image of each of the plurality of batteries, and the service life of each of the plurality of batteries after a first cycle; and a training unit for training a battery service life prediction model through multi-modal learning on the basis of the tabular data, the image, and the service life after the first cycle.

Description

Battery life prediction device and its operating method

[0001] Cross-citation with related applications

[0002] This application claims the benefit of priority to Republic of Korea Patent Application No. 10-2023-0105936, filed August 11, 2023, the entire contents of which are incorporated herein by reference.

[0003] Technology field

[0004] The embodiments disclosed in this document relate to a battery life prediction device and an operating method thereof.

[0005] Recently, research and development on secondary batteries has been actively conducted. Here, secondary batteries are rechargeable and include both conventional Ni / Cd and Ni / MH batteries, as well as recent lithium-ion batteries. Among secondary batteries, lithium-ion batteries have the advantage of having a much higher energy density than conventional Ni / Cd and Ni / MH batteries. Furthermore, lithium-ion batteries can be manufactured in a compact and lightweight form, making them suitable for use as power sources for mobile devices. Recently, their use has expanded to include power sources for electric vehicles, attracting attention as a next-generation energy storage medium.

[0006] Defect detection during the battery activation process is limited in accuracy, relying solely on voltage and current generated during the process. Furthermore, sophisticated defect detection methods are not designed to detect defects such as low-voltage battery failures. Therefore, a learning model is needed to detect life cycle defects early during the battery activation process and support decision-making regarding battery life quality.

[0007] One purpose of the embodiments disclosed in this document is to provide a battery life prediction device and an operating method thereof capable of predicting the life of a battery at an early stage based on tabula data of the battery and an image of the battery after an activation process.

[0008] One purpose of the embodiments disclosed in this document is to provide a life prediction model based on a multi-modal deep learning model, and to provide a battery life prediction device and its operating method capable of augmenting insufficient learning data.

[0009] The technical problems of the embodiments disclosed in this document are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the descriptions below.

[0010] A battery life prediction device according to one embodiment disclosed in the present document may include an information acquisition unit that acquires tabular data for each of a plurality of batteries, an image of each of the plurality of batteries, and a lifespan of each of the plurality of batteries after a first cycle, and a learning unit that learns a battery life prediction model through multi-modal learning based on the tabular data, the image, and the lifespan after the first cycle of each of the plurality of batteries.

[0011] In one embodiment, the method further includes a learning data augmentation unit that augments learning data through a generative adversarial network (GAN), wherein the generative model can be learned based on tabula data and images of each of the plurality of batteries.

[0012] In one embodiment, the learning data augmentation unit can reduce the dimension of the augmented tabula data through principal component analysis, and input the dimensionally reduced data into an image generation model to generate an image corresponding to each augmented tabula data.

[0013] In one embodiment, the learning unit may supervise learning the battery life prediction model through a neural network by embedding tabula data of each of the plurality of batteries through a first network, embedding images of each of the plurality of batteries through a second network, and concatenating the embedded tabula data and the embedded images to output the life after the first cycle.

[0014] In one embodiment, the first network may include Tabnet and the second network may include a vision transformer.

[0015] In one embodiment, the tabular data for each of the plurality of batteries may include information about each of the plurality of batteries when the activation process is completed, and the image of each of the plurality of batteries may include an X-ray image of each of the plurality of batteries after the activation process is completed.

[0016] In one embodiment, the tabula data may include at least one of a room temperature aging time of the battery, a high temperature aging time, a charge / discharge capacity, a charge / discharge voltage, a charge / discharge current, a charge / discharge temperature change, a constant current charging time, a sealing time, and an open circuit voltage.

[0017] In one embodiment, the battery life prediction model may further include a life prediction unit that predicts the life of the first battery after the first cycle based on tabula data and images of a first battery different from the plurality of batteries.

[0018] An operating method of a battery life prediction device according to one embodiment disclosed in this document may include an operation of obtaining tabular data for each of a plurality of batteries, an image of each of the plurality of batteries, and a lifespan of each of the plurality of batteries after a first cycle, and an operation of learning a battery life prediction model through multi-modal learning based on the tabular data, the image, and the lifespan after the first cycle of each of the plurality of batteries.

[0019] In one embodiment, the method further comprises augmenting training data using a generative adversarial network (GAN), wherein the generative model can be trained based on tabula data and images of each of the plurality of batteries.

[0020] In one embodiment, the operation of augmenting learning data through the generative model (GAN) may include an operation of reducing the dimensionality of the augmented tabula data through principal component analysis and an operation of inputting the dimensionally reduced data into an image generation model to generate an image corresponding to each of the augmented tabula data.

[0021] In one embodiment, the operation of learning a battery life prediction model through multi-modal learning based on tabula data, images, and life after the first cycle of each of the plurality of batteries may include the operation of embedding tabula data of each of the plurality of batteries through a first network, the operation of embedding images of each of the plurality of batteries through a second network, and the operation of supervised learning the battery life prediction model to output the life after the first cycle by concatenating the embedded tabula data and the embedded images through a neural network.

[0022] In one embodiment, the first network may include Tabnet and the second network may include a vision transformer.

[0023] In one embodiment, the tabular data for each of the plurality of batteries may include information about each of the plurality of batteries when the activation process is completed, and the image of each of the plurality of batteries may include an X-ray image of each of the plurality of batteries after the activation process is completed.

[0024] In one embodiment, the tabula data may include at least one of a room temperature aging time of the battery, a high temperature aging time, a charge / discharge capacity, a charge / discharge voltage, a charge / discharge current, a charge / discharge temperature change, a constant current charging time, a sealing time, and an open circuit voltage.

[0025] In one embodiment, the battery life prediction model may further include an operation of predicting the life of the first battery after the first cycle based on tabula data and images of the first battery, which are different from the plurality of batteries.

[0026] A battery life prediction device and its operating method according to one embodiment disclosed in this document can learn a life prediction model based on tabula data and images of a battery after an activation process.

[0027] A battery life prediction device and its operating method according to one embodiment disclosed in this document can learn tabula data and images of a battery based on a multi-modal deep learning model.

[0028] The battery life prediction device and its operating method according to one embodiment disclosed in this document can secure insufficient data by augmenting data required for learning.

[0029] In addition, various effects may be provided, either directly or indirectly, through this document.

[0030] FIG. 1 is a drawing showing a functional block diagram of a battery life prediction device according to one embodiment disclosed in this document.

[0031] FIG. 2 is a block diagram specifically showing a battery life prediction device according to one embodiment disclosed in this document.

[0032] FIG. 3 is a diagram showing an example of a battery life prediction device augmenting data according to one embodiment disclosed in this document.

[0033] FIG. 4 is a diagram showing an example of a battery life prediction device learning a battery life prediction model according to one embodiment disclosed in this document.

[0034] FIG. 5 is a diagram showing an operation method of a battery life prediction device according to one embodiment disclosed in this document.

[0035] FIG. 6 and FIG. 7 are drawings specifically showing an operation method of a battery life prediction device according to one embodiment disclosed in this document.

[0036] Hereinafter, embodiments disclosed in this document will be described in detail with reference to exemplary drawings. When designating components in each drawing, it should be noted that, where possible, identical components are given the same reference numerals, even if they appear in different drawings. Furthermore, when describing embodiments disclosed in this document, detailed descriptions of related known structures or functions will be omitted if they are deemed to hinder understanding of the embodiments disclosed in this document.

[0037] In describing the components of the embodiments disclosed in this document, terms such as first, second, A, B, (a), (b), etc. may be used. These terms are only intended to distinguish the components from other components and do not limit the nature, order, or sequence of the components. In addition, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which the embodiments disclosed in this document belong. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning they have in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined in this application.

[0038] FIG. 1 is a drawing showing a functional block diagram of a battery life prediction device according to one embodiment disclosed in this document.

[0039] Referring to FIG. 1, a battery life prediction device (200, see FIG. 2) according to one embodiment disclosed in the present document may include a processor (110), a memory (120), and a communication circuit (130). According to an embodiment, the battery life prediction device (200) may further include a display (140).

[0040] According to an embodiment, the operation of the battery life prediction device (200) may be performed by a BMS (Battery management system) in a vehicle, as well as by various devices such as a server, cloud, charger, or charger / discharger.

[0041] The processor (110) may control the overall operation of the battery life prediction device (200). In various embodiments, the processor (110) may include a single processor core or a plurality of processor cores. For example, the processor (110) may include a multi-core such as a dual-core, a quad-core, or a hexa-core. According to embodiments, the processor (110) may further include a cache memory located internally or externally. According to embodiments, the processor (110) may be configured with one or more processors. For example, the processor (110) may include at least one of an application processor, a communication processor, or a graphical processing unit (GPU).

[0042] All or part of the processor (110) may be electrically or operatively coupled with or connected to other components (e.g., memory (120), communication circuitry (130), or display (140)) within the battery life prediction device (200). The processor (110) may receive commands from other components of the battery life prediction device (200), interpret the received commands, and perform calculations or process data according to the interpreted commands. The processor (110) may interpret and process messages, data, commands, or signals received from the memory (120), communication circuitry (130), or display (140). The processor (110) may generate new messages, data, commands, or signals based on the received messages, data, commands, or signals. The processor (110) may provide processed or generated messages, data, instructions, or signals to the memory (120), communication circuitry (130), or display (140).

[0043] The processor (110) can process data or signals generated or produced by a program. For example, the processor (110) can request instructions, data, or signals from the memory (120) to execute or control the program. The processor (110) can record (or store) or update instructions, data, or signals in the memory (120) to execute or control the program.

[0044] The memory (120) may store commands, control command codes, control data, or user data that control the battery life prediction device (200). For example, the memory (120) may include at least one of an application program, an operating system (OS), middleware, or a device driver.

[0045] The memory (120) may include one or more of volatile memory or non-volatile memory. The volatile memory may include dynamic random access memory (DRAM), static RAM (SRAM), synchronous DRAM (SDRAM), phase-change RAM (PRAM), magnetic RAM (MRAM), resistive RAM (RRAM), ferroelectric RAM (FeRAM), etc. The non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, etc.

[0046] The memory (120) may further include a non-volatile medium such as a hard disk drive (HDD), a solid state disk (SSD), an embedded multi media card (eMMC), or a universal flash storage (UFS).

[0047] The communication circuit (130) can support the establishment of a wired or wireless communication connection between the battery life prediction device (200) and an external electronic device (e.g., a battery information collection device, a battery information processing device, a server, a cloud server), and the performance of communication through the established connection. According to one embodiment, the communication circuit (130) includes a wireless communication circuit (e.g., a cellular communication circuit, a short-range wireless communication circuit, or a GNSS (global navigation satellite system) communication circuit) or a wired communication circuit (e.g., a LAN (local area network) communication circuit, or a power line communication circuit), and can communicate with an external electronic device through a short-range communication network such as Bluetooth, WiFi direct, or IrDA (infrared data association), or a long-range communication network such as a cellular network, the Internet, or a computer network using the corresponding communication circuit. The various types of communication circuits (130) described above can be implemented as one chip or can be implemented as separate chips. According to an embodiment, since the communication circuit (130) is not an essential component of the battery life prediction device (200), the battery life prediction device (200) can be implemented excluding the communication circuit (130).

[0048] The display (140) can output content, data, or signals. In various embodiments, the display (140) can display image data processed by the processor (110). According to embodiments, the display (140) may be configured with an integrated touch screen by being combined with a plurality of touch sensors (not shown) that can receive touch inputs, etc. When the display (140) is configured with a touch screen, the plurality of touch sensors may be arranged above the display (140) or below the display (140).

[0049] Hereinafter, the battery life prediction device (200) will be specifically described with reference to FIGS. 2 to 4.

[0050] FIG. 2 is a block diagram specifically showing a battery life prediction device according to one embodiment disclosed in this document.

[0051] Referring to FIG. 2, a battery life prediction device (200) according to an embodiment disclosed in the present document may include an information acquisition unit (210) and a learning unit (220). According to an embodiment, the battery life prediction device (200) may further include a learning data augmentation unit (230). According to an embodiment, the battery life prediction device (200) may further include a life prediction unit (240). According to an embodiment, the battery life prediction device (200) may further include a learning data augmentation unit (230) and a life prediction unit (240).

[0052] According to an embodiment, each operation of the information acquisition unit (210), learning unit (220), learning data augmentation unit (230), and life prediction unit (240) can be performed through the processor (110) of FIG. 1.

[0053] The information acquisition unit (210) can acquire tabular data for each of the plurality of batteries, an image of each of the plurality of batteries, and the lifespan of each of the plurality of batteries after the first cycle.

[0054] In an embodiment, the tabula data for each of the plurality of batteries may include information about each of the plurality of batteries when the activation process is completed. For example, the tabula data for each of the plurality of batteries may include at least one of a room temperature aging time, a high temperature aging time, a charge / discharge capacity, a charge / discharge voltage, a charge / discharge current, a charge / discharge temperature change, a constant current charging time, a sealing time, and an open circuit voltage.

[0055] In an embodiment, each image of the plurality of batteries may include an X-ray image of each of the plurality of batteries after the activation process is completed. However, the present invention is not limited thereto, and each image of the plurality of batteries may include at least one image related to each of the plurality of batteries.

[0056] The learning unit (220) can learn a battery life prediction model through multi-modal learning based on tabula data, images, and lifespan after the first cycle of each battery. For example, multi-modal learning may refer to a model utilizing two types of data.

[0057] According to an embodiment, the learning unit (220) can learn a battery life prediction model with improved prediction performance because it can preserve information about interactions between data through multi-modal learning based on tabula data and images of each of a plurality of batteries.

[0058] The learning data augmentation unit (230) can augment learning data using a generative adversarial network (GAN). For example, the generative model can be trained based on tabula data and images of each of multiple batteries.

[0059] According to an embodiment, the learning data augmentation unit (230) can reduce the dimensionality of the augmented tabula data through principal component analysis and input the dimensionally reduced data into an image generation model to generate an image corresponding to each augmented tabula data. For example, when generating an image corresponding to each augmented tabula data, the learning data augmentation unit (230) can additionally input a Latent Vector to increase the randomness of the image.

[0060] A generative adversarial network (GAN) is a generative model in which two architectures, a generator and a discriminator, interact. The generator generates data to fool the discriminator, and the discriminator detects falsehoods in the images generated by the generator, thereby enabling learning. In some embodiments, the GAN may include a DCGAN (Deep Convolutional GAN), a Conditional GAN, an InfoGAN, a WGAN (Wasserstein GAN), a WGAN-GP, an LSGAN, an Attention GAN, a cycle GAN, a PGGAN (Progressive GAN), a Style GAN, a BiGAN, a BEGAN, a TGAN, a CTGAN, and the like. However, the present invention is not limited thereto, and the generative model (GAN) may include various well-known GAN models.

[0061] According to an embodiment, the learning data augmentation unit (230) can augment tabula data of each of a plurality of batteries through TGAN or CTGAN.

[0062] According to an embodiment, the learning data augmentation unit (230) can generate an image corresponding to each piece of tabula data augmented through CGAN. For example, when generating an image, the learning data augmentation unit (230) can input the augmented tabula data together with a latent vector so that the image is generated according to the characteristics of the tabula data.

[0063] According to an embodiment, the learning data augmentation unit (230) may learn a tabula data augmentation model based on the tabula data of each of the plurality of batteries. In addition, the learning data augmentation unit (230) may learn an image augmentation model based on the tabula data and images of each of the plurality of batteries. In this case, the tabula data of each of the plurality of batteries may be tabula data whose dimensionality has been reduced through principal component analysis.

[0064] The learning unit (220) may embed tabula data of each of the plurality of batteries through a first network and embed images of each of the plurality of batteries through a second network. In this case, the learning unit (220) may perform supervised learning on a battery life prediction model through a neural network to output the life after the first cycle by concatenating the embedded tabula data and embedded images. For example, supervised learning may mean that the learning unit (220) trains a battery life prediction model to output the life after the first cycle by considering the life after the first cycle as the correct answer. According to an embodiment, the first network may include Tabnet, and the second network may include a Vision Transformer (ViT).

[0065] According to an embodiment, the learning unit (220) can supervise learning of a battery life prediction model by inputting tabula data and augmented tabula data of each of a plurality of batteries by concatenating them and using them as tabula data, and by inputting images and augmented images of each of a plurality of batteries by concatenating them and using them as images.

[0066] According to an embodiment, the battery life prediction model may include a plurality of neural network layers. Each of the plurality of neural network layers may be a linear layer or a non-linear layer. For example, the linear layers may include a convolution layer, a fully connected layer, and a deconvolution layer, and the non-linear layers may include a sampling layer, a pooling layer, and an activation layer, but are not limited to the aforementioned layers.

[0067] According to an embodiment, the learning unit (220) can learn a battery life prediction model through a deep learning network. For example, the deep learning network may include various types of networks such as a CNN (Convolutional Neural Network) such as GoogleNet, AlexNet, VGG Network, etc., a R-CNN (Region with Convolutional Neural Network), an RPN (Region Proposal Network), an RNN (Recurrent Neural Network), an S-DNN (Stacking-based deep Neural Network), an S-SDNN (State-Space Dynamic Neural Network), a Deconvolution Network, a DBN (Deep Belief Network), an RBM (Restricted Boltzman Machine), an FCN (Fully Convolutional Network), an LSTM (Long Short-Term Memory) Network, an FPN (Feature Pyramid Network), an SPP (Spatial PyramidPooling), etc., but is not limited to the above-described networks. For example, the deep learning network may be based on a CNN (Convolutional Neural Network) network.

[0068] In some embodiments, the deep learning network may be a network based on end-to-end learning. End-to-end learning refers to learning without human intervention between the network's input and output. End-to-end learning has the advantage of eliminating the need for human intervention, such as re-executing a specific process, once learning begins.

[0069] In an embodiment, the battery life prediction device (200) according to one embodiment disclosed in the present document may further include a life prediction unit (240). For example, the life prediction unit (240) may predict the life of the first battery after the first cycle based on tabula data and images of a first battery different from a plurality of batteries through a battery life prediction model. For example, the life prediction unit (240) may input tabula data and images after the activation process of the first battery into the battery life prediction model and predict the life of the first battery after the first cycle as an output.

[0070] A battery life prediction device (200) according to one embodiment disclosed in this document can learn a life prediction model based on tabula data and images of a battery after an activation process.

[0071] A battery life prediction device (200) according to one embodiment disclosed in this document can learn tabula data and images of a battery based on a multi-modal deep learning model.

[0072] The battery life prediction device (200) according to one embodiment disclosed in this document can secure insufficient data by augmenting data required for learning.

[0073] FIG. 3 is a diagram showing an example of a battery life prediction device augmenting data according to one embodiment disclosed in this document.

[0074] Referring to FIG. 3, the information acquisition unit (210) of the battery life prediction device (200) according to one embodiment disclosed in the present document can acquire tabula data (310) of each of the plurality of batteries and images (340) of each of the plurality of batteries.

[0075] The learning data augmentation unit (230) can input tabula data (310) of each of a plurality of batteries into a tabula data generation model (320) and output augmented tabula data (330). According to an embodiment, the tabula data generation model (320) can be trained based on the tabula data (310) of each of a plurality of batteries.

[0076] The learning data augmentation unit (230) can input the images (340), augmented tabula data (330), and Latent Vector (360) of each of the plurality of batteries into the image generation model (350) to output an augmented image (370). For example, the image generation model (350) can be trained based on the tabula data (310) of each of the plurality of batteries and the images (340) of each of the plurality of batteries. According to an embodiment, the Latent Vector (360) can be used when training the image generation model (350), and may not be used when outputting the augmented image.

[0077] According to an embodiment, the augmented tabula data (330) and the augmented image (370) may correspond to each other. In addition, the tabula data (310) of each of the plurality of batteries and the image (340) of each of the plurality of batteries may correspond to each other.

[0078] FIG. 4 is a diagram showing an example of a battery life prediction device learning a battery life prediction model according to one embodiment disclosed in this document.

[0079] The learning data augmentation unit (230) can generate tabula data (415) for model learning by concatenating the tabula data (405) of each of the plurality of batteries and the augmented tabula data (410). According to an embodiment, the tabula data (405) of each of the plurality of batteries can be substantially identical to the tabula data (310) of each of the plurality of batteries of FIG. 3, and the augmented tabula data (410) can be substantially identical to the augmented tabula data (330) of FIG. 3.

[0080] The learning data augmentation unit (230) can generate an image (430) for model learning by concatenating the image (420) of each of the plurality of batteries and the augmented image (425). According to an embodiment, the image (420) of each of the plurality of batteries can be substantially identical to the image (340) of each of the plurality of batteries of FIG. 3, and the augmented image (425) can be substantially identical to the augmented image (370) of FIG. 3.

[0081] The learning unit (220) can embed (440) tabula data (415) for model learning through Tabnet (435). In addition, the learning unit (220) can embed (450) images (430) for model learning through a vision transformer (ViT, 445).

[0082] The learning unit (220) can learn the battery life prediction model (460) by concatenating embedded tabula data and embedded images and inputting them into the battery life prediction model (460). For example, the learning unit (220) can learn the battery life prediction model (460) to output the lifespan (455) of each of a plurality of batteries after the first cycle.

[0083] According to an embodiment, the lifespan (455) of each of the plurality of batteries after the first cycle may be provided corresponding to the tabula data (405) of each of the plurality of batteries and the image (420) of each of the plurality of batteries. Additionally, the lifespan (455) of each of the plurality of batteries after the first cycle may be generated corresponding to the augmented tabula data (410) and the augmented image (425).

[0084] FIG. 5 is a diagram illustrating an operation method of a battery life prediction device according to one embodiment disclosed in this document. The operations illustrated in FIG. 5 can be performed through the battery life prediction device (200) of FIG. 2.

[0085] Referring to FIG. 5, in operation 510, the information acquisition unit (210) can acquire tabular data for each of the plurality of batteries, an image of each of the plurality of batteries, and the lifespan of each of the plurality of batteries after the first cycle.

[0086] In an embodiment, the tabula data for each of the plurality of batteries may include information about each of the plurality of batteries when the activation process is completed. For example, the tabula data for each of the plurality of batteries may include at least one of a room temperature aging time, a high temperature aging time, a charge / discharge capacity, a charge / discharge voltage, a charge / discharge current, a charge / discharge temperature change, a constant current charging time, a sealing time, and an open circuit voltage.

[0087] In an embodiment, each image of the plurality of batteries may include an X-ray image of each of the plurality of batteries after the activation process is completed. However, the present invention is not limited thereto, and each image of the plurality of batteries may include at least one image related to each of the plurality of batteries.

[0088] In operation 520, the learning data augmentation unit (230) can augment learning data using a generative adversarial network (GAN). For example, the generative model can be trained based on tabula data and images of each of the multiple batteries.

[0089] According to an embodiment, the learning data augmentation unit (230) can reduce the dimensionality of the augmented tabula data through principal component analysis and input the dimensionally reduced data into an image generation model to generate an image corresponding to each augmented tabula data. For example, when generating an image corresponding to each augmented tabula data, the learning data augmentation unit (230) can additionally input a Latent Vector to increase the randomness of the image. According to an embodiment, operation 520 can be omitted.

[0090] According to an embodiment, the learning data augmentation unit (230) can augment tabula data of each of a plurality of batteries through TGAN or CTGAN.

[0091] According to an embodiment, the learning data augmentation unit (230) can generate an image corresponding to each piece of tabula data augmented through CGAN. For example, when generating an image, the learning data augmentation unit (230) can input the augmented tabula data together with a latent vector so that the image is generated according to the characteristics of the tabula data.

[0092] In operation 530, the learning unit (220) can learn a battery life prediction model through multi-modal learning based on tabula data, images, and lifespan after the first cycle of each battery. For example, multi-modal learning may refer to a model utilizing two types of data.

[0093] According to an embodiment, the learning unit (220) can learn a battery life prediction model with improved prediction performance because it can preserve information about interactions between data through multi-modal learning based on tabula data and images of each of a plurality of batteries.

[0094] According to an embodiment, the learning unit (220) can supervise learning of a battery life prediction model by inputting tabula data and augmented tabula data of each of a plurality of batteries by concatenating them and using them as tabula data, and by inputting images and augmented images of each of a plurality of batteries by concatenating them and using them as images.

[0095] In operation 540, the life prediction unit (240) can predict the life of the first battery after the first cycle based on tabula data and images of the first battery and a plurality of batteries through a battery life prediction model.

[0096] FIGS. 6 and 7 are diagrams specifically illustrating an operating method of a battery life prediction device according to one embodiment disclosed in this document. The operations illustrated in FIGS. 6 and 7 can be performed through the battery life prediction device (200) of FIG. 2.

[0097] Referring to FIG. 6, in operation 610, the learning data augmentation unit (230) can reduce the dimension of the augmented tabula data through principal component analysis.

[0098] In operation 620, the learning data augmentation unit (230) can input dimensionally reduced data into an image generation model to generate an image corresponding to each augmented tabula data.

[0099] According to an embodiment, operations 610 and 620 may be performed as included in operation 520 of FIG. 5.

[0100] Referring to FIG. 7, in operation 710, the learning unit (220) can embed tabula data of each of a plurality of batteries through the first network.

[0101] In operation 720, the learning unit (220) can embed images of each of the plurality of batteries through the second network.

[0102] In operation 730, the learning unit (220) may perform supervised learning on a battery life prediction model to output the life after the first cycle by concatenating the embedded tabula data and embedded images through a neural network. For example, supervised learning may mean that the learning unit (220) trains the battery life prediction model to output the life after the first cycle by considering the life after the first cycle as the correct answer. According to an embodiment, the first network may include Tabnet, and the second network may include a Vision Transformer (ViT).

[0103] According to an embodiment, operations 710 to 730 may be performed as included in operation 530 of FIG. 5.

[0104] The above description is merely an example of the technical idea disclosed in this document, and those skilled in the art to which the embodiments disclosed in this document pertain may make various modifications and variations without departing from the essential characteristics of the embodiments disclosed in this document.

[0105] Accordingly, the embodiments disclosed in this document are not intended to limit the technical ideas disclosed in this document, but rather to explain them, and the scope of the technical ideas disclosed in this document is not limited by these embodiments. The protection scope of the technical ideas disclosed in this document should be interpreted according to the following scope, and all technical ideas within the equivalent scope should be interpreted as being included within the scope of the rights of this document.

[0106] [Explanation of symbols]

[0107] 110: Processor

[0108] 120: Memory

[0109] 130: Communication circuit

[0110] 140: Display

[0111] 200: Battery life prediction device

[0112] 210: Information Acquisition Department

[0113] 220: Learning Department

[0114] 230: Learning Data Augmentation

[0115] 240: Life Prediction Department

Claims

1. An information acquisition unit that acquires tabular data for each of a plurality of batteries, an image of each of the plurality of batteries, and the lifespan of each of the plurality of batteries after the first cycle; and A battery life prediction device, comprising: a learning unit that learns a battery life prediction model through multi-modal learning based on tabula data, images, and life after the first cycle of each of the plurality of batteries; 2. In paragraph 1, Further comprising a learning data augmentation unit that augments learning data through a generative adversarial network (GAN); A battery life prediction device, wherein the above generation model is learned based on tabula data and images of each of the plurality of batteries.

3. In paragraph 2, The above learning data augmentation unit is, The dimensionality of the augmented tabula data is reduced through principal component analysis, A battery life prediction device that inputs dimensionally reduced data into an image generation model to generate an image corresponding to each augmented tabula data.

4. In paragraph 1, The above learning department, Embedding the tabula data of each of the above plurality of batteries through the first network, Embedding each image of the plurality of batteries through a second network, A battery life prediction device that supervises learning of the battery life prediction model through a neural network by concatenating the embedded tabula data and the embedded image to output the life after the first cycle.

5. In paragraph 4, A battery life prediction device, wherein the first network includes Tabnet and the second network includes a vision transformer.

6. In paragraph 1, The tabular data for each of the plurality of batteries includes information about each of the plurality of batteries when the activation process is completed, A battery life prediction device, wherein each image of the plurality of batteries includes an X-ray image of each of the plurality of batteries after the activation process is completed.

7. In paragraph 1, The above tabula data is, A battery life prediction device comprising at least one of a room temperature aging time, a high temperature aging time, a charge / discharge capacity, a charge / discharge voltage, a charge / discharge current, a charge / discharge temperature change, a constant current charging time, a sealing time, and an open circuit voltage of a battery.

8. In paragraph 1, A battery life prediction device further comprising a life prediction unit that predicts the life of the first battery after the first cycle based on tabula data and images of a first battery different from the plurality of batteries through the battery life prediction model.

9. An operation of obtaining tabular data for each of a plurality of batteries, an image of each of the plurality of batteries, and a lifespan of each of the plurality of batteries after a first cycle; and An operating method of a battery life prediction device, comprising: an operation of learning a battery life prediction model through multi-modal learning based on tabula data, images, and life after the first cycle of each of the plurality of batteries; 10. In paragraph 9, The operation of augmenting training data through a generative adversarial network (GAN); further comprising; A method for operating a battery life prediction device, wherein the above generation model is learned based on tabula data and images of each of the plurality of batteries.

11. In paragraph 10, The operation of augmenting training data through the above generative model (GAN) is as follows: The operation of reducing the dimensionality of the augmented tabula data through principal component analysis; and An operating method of a battery life prediction device, comprising: inputting dimensionally reduced data into an image generation model to generate an image corresponding to each augmented tabula data; 12. In paragraph 9, The operation of learning a battery life prediction model through multi-modal learning based on the tabula data, images, and life after the first cycle of each of the above-mentioned plurality of batteries is as follows: An operation of embedding tabula data of each of the plurality of batteries through a first network; An operation of embedding each image of the plurality of batteries through a second network; and An operating method of a battery life prediction device, comprising: an operation of supervised learning the battery life prediction model through a neural network by concatenating the embedded tabula data and the embedded image to output the life after the first cycle; 13. In paragraph 12, A method of operating a battery life prediction device, wherein the first network includes Tabnet and the second network includes a vision transformer.

14. In paragraph 9, The tabular data for each of the plurality of batteries includes information about each of the plurality of batteries when the activation process is completed, A method of operating a battery life prediction device, wherein each image of the plurality of batteries includes an X-ray image of each of the plurality of batteries after the activation process is completed.

15. In paragraph 9, The above tabula data is, An operating method of a battery life prediction device, comprising at least one of a room temperature aging time, a high temperature aging time, a charge / discharge capacity, a charge / discharge voltage, a charge / discharge current, a charge / discharge temperature change, a constant current charging time, a sealing time, and an open circuit voltage of a battery.

16. In paragraph 9, An operating method of a battery life prediction device, further comprising: an operation of predicting the life of the first battery after the first cycle based on tabula data and images of a first battery different from the plurality of batteries through the battery life prediction model;

Images