Battery cell life diagnosis device and its operation method

Abstract

A battery cell diagnosis device according to an embodiment of the present invention can include a collection unit that collects time-series data of a battery cell according to operating conditions, a conversion unit that converts the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension, an extraction unit that extracts a feature value from the image, and a calculation unit that calculates the life of the battery cell based on the feature value.

Description

Technical Field

[0001] The present invention claims the benefit of priority based on Korean Patent Application No. 10-2022-0067847 filed on June 2, 2022, and all the contents disclosed in the literature of the Korean patent application are incorporated herein by reference in their entirety.

[0002] The embodiments disclosed in this document relate to a battery cell life diagnosis device and an operation method thereof.

Background Art

[0003] In recent years, research and development on secondary batteries have been actively conducted. A secondary battery is a battery that can be charged and discharged, and can include all conventional Ni / Cd batteries, Ni / MH batteries, etc., and recent lithium-ion batteries. Lithium-ion batteries have the advantage of having a much higher energy density compared to conventional Ni / Cd batteries, Ni / MH batteries, etc. In addition, since lithium-ion batteries can be manufactured in a small and lightweight manner, they are used as power sources for mobile devices, and in recent years, their usage range has been extended to power sources for electric vehicles, and they have attracted attention as next-generation energy storage media.

[0004] Batteries show a tendency to deteriorate as they are repeatedly charged and discharged. For example, as batteries are repeatedly charged and discharged, their capacity and resistance deteriorate, and their remaining life can be shortened. In addition, the degree of deterioration and the remaining life of batteries can vary depending on the usage conditions.

[0005] When the remaining life of a battery drops rapidly, safety problems may occur during the use of the battery. Therefore, there is a need for a method to diagnose the life of a battery in advance and prevent risks due to a sharp drop in battery performance.

Summary of the Invention

Problems to be Solved by the Invention

[0006] One object of the embodiments disclosed in this document is to provide an apparatus and an operating method thereof that convert operation characteristic data of a battery cell, which are collected in time series corresponding to operating conditions, into an image so that the life of the battery cell can be effectively calculated.

[0007] One object of the embodiments disclosed in this document is to provide an apparatus and an operating method thereof that diagnose the life of a battery cell based on an image converted from time series data.

[0008] 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 can be clearly understood by those skilled in the art from the following description.

Means for Solving the Problems

[0009] A battery cell diagnosis apparatus according to an embodiment disclosed in this document may include a collection unit that collects time series data of a battery cell according to operating conditions, a conversion unit that converts the time series data into an image corresponding to a spatial trajectory of a predetermined dimension, an extraction unit that extracts a feature value from the image, and a calculation unit that calculates the life of the battery cell based on the feature value.

[0010] In one embodiment, the image may be an image corresponding to a two-dimensional spatial trajectory. In one embodiment, the time series data may be data corresponding to changes in the operating characteristics of the battery cell according to time.

[0011] In one embodiment, the extraction unit can extract the feature value via a first neural network. In one embodiment, the first neural network may be a convolutional neural network including a convolutional layer and a pooling layer.

[0012] In one embodiment, the calculation unit can calculate the life via a second neural network. In one embodiment, the second neural network may be a deep neural network including multiple hidden layers.

[0013] In one embodiment, the operating conditions correspond to the characteristic values, and the calculation unit can generate a prediction function for the operating conditions via a third neural network.

[0014] In one embodiment, the third neural network may be a deep neural network including multiple hidden layers. In one embodiment, the calculation unit can calculate the lifespan based on the prediction function and the characteristic values.

[0015] In one embodiment, the conversion unit can convert the time-series data into the spatial trajectory and use the spatial trajectory to convert the time-series data into the image.

[0016] In one embodiment, the conversion unit represents the distances between points located on the spatial trajectory by a distance matrix and can convert the region corresponding to the distance matrix into the image based on the distance values of the distance matrix.

[0017] The operation method of the battery cell diagnosis device according to one embodiment disclosed in this document can include the steps of collecting time-series data of the battery cell according to operating conditions, converting the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension, extracting characteristic values from the image, and calculating the lifespan of the battery cell based on the characteristic values.

[0018] In one embodiment, the step of converting the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension can include the step of converting the time-series data into the spatial trajectory and the step of using the spatial trajectory to convert the time-series data into the image.

[0019] In one embodiment, the step of converting the time-series data into the image using the spatial locus may include representing distances between points located on the spatial locus by a distance matrix, and visualizing a region corresponding to the distance matrix based on distance values of the distance matrix.

[0020] In one embodiment, the step of extracting feature values from the image may extract the feature values via a first neural network. In one embodiment, the step of calculating the remaining life may calculate the remaining life via a second neural network, and the second neural network may be a deep neural network including multiple hidden layers.

[0021] In one embodiment, the operating conditions correspond to the feature values, and the step of calculating the remaining life may include generating a prediction function for the operating conditions via a third neural network, concatenating the prediction function and the feature values, and predicting the remaining life based on the prediction function and the feature values.

Advantages of the Invention

[0022] A battery cell life diagnosis device and an operation method thereof according to an embodiment disclosed in this document can calculate the state of a battery cell. A battery cell diagnosis device and an operation method thereof according to an embodiment disclosed in this document can calculate the state of a battery cell by converting time-series data regarding operating characteristics of the battery cell into an image having a preset dimension.

[0023] A battery cell diagnosis device and an operation method thereof according to an embodiment disclosed in this document can calculate the remaining life of a battery cell by analyzing image data via a neural network.

[0024] The battery cell diagnosis device and its operation method according to one embodiment disclosed in this document can improve the life calculation accuracy of a battery cell based on an image corresponding to the operating characteristics of the battery cell and the operating conditions of the battery cell. In addition to this, various effects that can be grasped directly or indirectly are provided by this document.

Brief Description of the Drawings

[0025]

Figure 1

Figure 2

Figure 3a

Figure 3b

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

DETAILED DESCRIPTION OF THE INVENTION

[0026] Hereinafter, embodiments disclosed in this document will be described in detail with reference to exemplary drawings. When assigning reference numerals to the components of each drawing, it should be noted that the same components are assigned the same reference numerals as much as possible when they are shown on other drawings. Also, when explaining the embodiments disclosed in this document, if a specific explanation of a related known configuration or function is determined to impede the understanding of the embodiments disclosed in this document, the detailed explanation thereof will be omitted.

[0027] When explaining the components of the embodiments disclosed in this document, terms such as first, second, A, B, (a), (b), etc. may be used. Such terms are merely for distinguishing the components from other components, and the essence, order, or procedure of the components are not limited by such terms. Also, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by those having ordinary knowledge in the technical field 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 in the context of the related art, and should not be interpreted in an ideal or overly formal sense unless clearly defined in this application.

[0028] FIG. 1 is a block diagram showing a battery cell diagnostic apparatus according to an embodiment disclosed in this document. Referring to FIG. 1, a battery cell diagnostic apparatus 10 according to an embodiment disclosed in this document can include a collection unit 100, a conversion unit 200, an extraction unit 300, and a calculation unit 400.

[0029] The collection unit 100 can acquire the operating conditions of the battery cell and the time-series data of the battery cell corresponding to the operating conditions. SOH (state of health) is an index indicating the performance degradation of a battery cell, and can be expressed as the percentage of the capacity or resistance of a battery cell that has decreased with respect to the initial capacity due to aging. Changes in SOH can occur due to, for example, an increase in resistance due to the polarization phenomenon of the chemical substances contained in the battery cell. When the performance of the battery cell decreases below a predefined critical point, replacement of the battery cell may be required. Generally, if it is at the 80% level of the initial capacity, it can be determined as the end-of-life point of the battery cell.

[0030] The operating conditions may be conditions that are kept constant in order to acquire operating characteristics during a charge-discharge experiment for measuring the performance index of the battery cell. Also, the operating characteristics of the battery cell are result values obtained during the charge-discharge experiment under the operating conditions, and may be values that change over time. The time-series data may be data measured during a predefined time interval of the operating characteristics of the battery cell. Therefore, the time-series data can show the change in the operating characteristics of the battery cell over time.

[0031] The performance index of the battery cell can, by way of example, mean the aforementioned SOH or SOC (state of charge, remaining capacity), etc. The charge-discharge experiment can mean an experiment that maintains the operating conditions and detects the operating characteristics of the battery cell. The operating conditions can mean an experimental environment that affects the output of the battery cell.

[0032] When the operating temperature and charge-discharge current of the battery cell are used as the operating conditions, the operating characteristics of the battery cell may be the charge-discharge voltage. At this time, the time-series data may be data showing the change over time of the charge-discharge voltage of the battery cell.

[0033] Battery cells can deteriorate in characteristics as the number of charge-discharge cycles increases. The collection unit 100 can acquire time-series data of battery cells according to the number of charge-discharge cycles. According to an embodiment, the number of charge-discharge cycles can be an operating condition of the battery cell.

[0034] The operating conditions can be changed according to the time-series data to be acquired. According to other embodiments, the operating characteristics can be the impedance, internal resistance, or capacitance of the battery cell, etc.

[0035] According to one embodiment, the collection unit 100 can include a measuring device that monitors the temperature, charge-discharge current, and charge-discharge voltage of the battery cell so as to obtain time-series data according to the operating conditions. Further, the measuring device can monitor the impedance, internal resistance, and capacitance of the battery cell, etc.

[0036] Also, according to other embodiments, the collection unit 100 can receive from the outside by matching the operating conditions of the battery cell and the time-series data of the battery cell corresponding to the operating conditions.

[0037] The time-series data may be data sampled or normalized at time intervals predetermined by the collection unit 100. For example, when the operating temperature and charge-discharge current of the battery cell are maintained constant, if the collection unit 100 collects raw data of the charge-discharge voltage of the battery cell according to time, a plurality of time-series data can be obtained by dividing the raw data at preset time intervals.

[0038] The conversion unit 200 can convert the time-series data of the battery cell into an image corresponding to a spatial trajectory of a predetermined dimension. The conversion of the time-series data can be performed by the RP (recurrence plot) method. The RP method may be a conversion technique used to convert time-series data into an image.

[0039] The RP method may be a technique that converts time-series data into a spatial trajectory in a predetermined dimension, represents the distances between points located on the spatial trajectory with a distance matrix, and converts the time-series data into an image. At this time, the spatial trajectory may mean a movement trajectory between data included in the time-series data.

[0040] Based on the distance values of the distance matrix, the conversion unit 200 can visualize the time-series data as an image corresponding to a spatial trajectory in a predetermined dimension. According to one embodiment, when the distance value of the distance matrix is equal to or greater than a threshold value, the conversion unit 200 can visualize the time-series data as an image by representing the region corresponding to the distance matrix two-dimensionally.

[0041] According to another embodiment, by representing the color of the region corresponding to the distance matrix two-dimensionally so as to change according to the distance value of the distance matrix, it can be visualized as an image including a red channel, a green channel, and a blue channel. By converting the time-series data into an image by the RP method, the conversion unit 200 can clearly represent features that are difficult to grasp in the time-series data.

[0042] The extraction unit 300 can extract feature values of an image via a first neural network. According to an embodiment, the extraction unit 300 can generate a plurality of feature maps from the image. The extraction unit 300 can process the generated feature maps to extract feature values, and can transmit the extracted feature values to the calculation unit 400.

[0043] The first neural network may be a Convolution Neural Network algorithm. A convolutional neural network is a deep learning algorithm that includes a convolution layer and a pooling layer. The convolution layer can include a filter (convolution kernel) for extracting features of an image and an activation function for calculating the input data passed through the filter into non-linear values. The image passed through the convolution layer can be output as a plurality of feature maps. The pooling layer can remove noise unrelated to features from the feature map.

[0044] Convolutional neural networks can have different architectures depending on the connection method of the convolution layer and the pooling layer. Exemplary convolutional neural networks for extracting feature values from images include AlexNet that uses stacked convolution layers, GoogLeNet that includes an Inception module, ResNet that includes residual blocks, etc.

[0045] The first neural network used in the extraction unit 300 can utilize a pre-trained model. According to an embodiment, the extraction unit 300 can use a GoogLeNet or ResNet model learned by a visual database (e.g., the ImageNet project) for feature value extraction through transfer learning.

[0046] According to another embodiment, the extraction unit 300 can optimize the first neural network by adjusting the weight of the convolution kernel used in the calculation in the convolution layer of the first neural network or by adjusting the structure of the learned convolutional neural network.

[0047] The first neural network used in the extraction unit 300 can vary according to the amount of images converted from time-series data. In other words, when the amount of images converted from time-series data is small, the neural network can use a conventionally learned model for feature extraction by transfer learning. When the amount of images converted from time-series data is large, the structure and weights of the conventionally learned model can be modified and utilized.

[0048] According to an embodiment, the extraction unit 300 can generate feature values by flattening and concatenating a plurality of generated feature maps. The number of feature maps can vary according to the number of filters included in the first neural network. The feature maps can be referred to as layers.

[0049] The calculation unit 400 can predict or calculate the SOH of the battery cell via a second neural network. According to an embodiment, the second neural network may be a Deep Neural Network algorithm. The calculation unit 400 can predict or calculate the SOH of the battery cell based on the feature values extracted from the image and the operating conditions.

[0050] A deep neural network is a deep learning algorithm that includes multiple hidden layers. By using multiple hidden layers, it can precisely process multiple data with high complexity.

[0051] The second neural network included in the calculation unit 400 can improve the accuracy compared to directly learning the time-series data to calculate the SOH of the battery cell by learning based on the feature values extracted by converting the time-series data into an image.

[0052] When learning without image-converting time-series data, the characteristics of the data cannot be appropriately reflected, the predicted values may be biased, or the tendency of the local region may be overly reflected, and the calculation accuracy of the calculation unit 400 may decrease.

[0053] Exemplarily, in the case of time-series data reflecting the change in the capacity of a battery cell over time, noise may occur in the time-series data due to, for example, a capacity regeneration phenomenon in which the capacity increases irregularly by removing residual reaction products, and the noise may affect the prediction of SOH.

[0054] When the time-series data is converted into an image and SOH is predicted based on the feature values extracted from the image, the image recursively reflects the change in the time-series data, and by facilitating the extraction of features, the prediction accuracy of SOH of the calculation unit 400 can be increased.

[0055] Deep neural networks can have different structures depending on the connection method of the hidden layers. Exemplary deep neural networks include, for example, an RNN (Recurrent Neural Network) in which the hidden layer has a cyclic structure, a gradient vanishing phenomenon, and an LSTM (Long Short-Term Memory) in which a layer for storing previous states is added to improve long-term dependencies.

[0056] Since the feature values input to the calculation unit 400 are those in which the image serving as the basis for feature value extraction is converted from time-series data, they can be analyzed like sequential data by the second neural network. Sequential data may mean data that changes over time.

[0057] According to another embodiment, the calculation unit 400 can generate a prediction function using the operating conditions corresponding to the feature values. The calculation unit 400 can concatenate the feature values and the operating conditions corresponding to the feature values.

[0058] The prediction function is a function that enables the second neural network to reflect the change in SOH due to operating conditions. By concatenating the prediction function and the feature values for learning by the second neural network, the tendency of the SOH change due to operating conditions can be reflected in the second neural network. Therefore, when the second neural network concatenates the prediction function and the feature values for learning, the prediction accuracy of the SOH of the calculation unit 400 can be improved.

[0059] Since the extraction unit 300 extracts feature values from the image obtained by converting the time-series data, any operating condition can correspond to each feature value. The collection unit 100 can transmit the operating condition corresponding to the feature value to the calculation unit 400. The calculation unit 400 can generate a prediction function via the third neural network.

[0060] According to the embodiment, the third neural network may be a deep neural network such as the second neural network. The learning using the second neural network and the third neural network is learning using information (feature values and operating conditions) in different dimensions, and can be referred to as multimodal learning.

[0061] FIG. 2 shows a method for processing time-series data of a battery cell diagnosis device according to an embodiment disclosed in this document. In FIG. 2, the time-series data under an arbitrary operating condition (X) is shown as a first graph (G1).

[0062] The collection unit 100 can collect the charge and discharge voltage (V) according to the time (T) of the battery cell when the preset temperature (t s ), the preset charge and discharge current (i s ), and the preset number of charge and discharge cycles (C s ) are used as the operating condition (X). In other words, the first graph (G1) can show the charge and discharge voltage (V) of the battery cell according to the time (T).

[0063] The collection unit 100 can collect a plurality of time-series data by dividing the charge-discharge voltage (V) according to the time (T) of the battery cell at preset time intervals. In FIG. 2, the case where the collection unit 100 collects the time-series data (D) divided from the first time (T 1 ) to the second time (T 2 ) is exemplarily shown.

[0064] The collection unit 100 can divide and collect the time-series data at any time interval, and according to other embodiments, the time regions where the plurality of time-series data overlap with each other can be included.

[0065] FIGS. 3a and 3b are for explaining an image conversion method of a battery cell diagnosis device according to an embodiment disclosed in this document. The conversion unit 200 can convert the time-series data (D) into an image (I) by the RP method.

[0066] To explain the RP method, an exemplary graph (GX) having a time axis (T) and a voltage axis (V) is shown. Hereinafter, with reference to FIG. 3a, the process of converting the exemplary graph (GX) into an exemplary image (IX) will be described in detail.

[0067] The exemplary graph (GX) can be converted into an exemplary space trajectory (PX) which is a two-dimensional space trajectory. The points (S1, S2, S3) on the exemplary space trajectory can indicate the change in the voltage axis (V) value due to the change in the time axis (T) value of the exemplary graph (GX).

[0068] In other words, S1 may be the one (0, 1) representing in coordinates the change in the voltage axis (V) value that changes from 0 to 1 while the value of the time axis (T) changes from 1 to 2. Similarly, S2 may be the one (1, 2) representing in coordinates the change in the voltage axis (V) value that changes from 1 to 2 while the value of the time axis (T) changes from 2 to 3.

[0069] The spatial trajectory can indicate the movement trajectory of the data included in the time-series data. Therefore, the points located on the spatial trajectory can recursively represent the relationship between adjacent data included in the time-series data.

[0070] The distance matrix may, by way of example, be a matrix composed of points located on the spatial trajectory. By way of example, the distance matrix can have a form such as (S1, S2) or (S1, S3).

[0071] The distance value of the distance matrix can mean the distance between points on the spatial trajectory. According to an embodiment, the time-series data can be visualized as an image by expressing it such that the color of the region corresponding to the distance matrix changes according to the distance value of the distance matrix. For example, the distance value of the distance matrix (S1, S2) can be √2, which is the distance between S1 and S2. Similarly, the distance value of the distance matrix (S1, S3) can be 2, which is the distance between S1 and S3.

[0072] Referring to the exemplary image (IX), the brightness of the corresponding region can change according to the distance value of the distance matrix. Since the RP method represents the distance between points located on the spatial trajectory with a distance matrix, the image represented by the RP method can have a symmetric shape.

[0073] FIG. 3b exemplarily shows the process in which the time-series data (D) in FIG. 2 is converted into image data (I) by the RP method. The image (I) converted by the RP method can clearly represent features compared to the first graph (G1) showing the time-series data.

[0074] FIG. 4 shows a method for extracting characteristic values of a battery cell diagnostic device according to an embodiment disclosed in this document. The extraction unit 300 can receive the image (I) and extract characteristic values from the received image (I) via the first neural network (NN1).

[0075] The first neural network (NN1) can include a convolutional layer (CL) and a pooling layer (PL). The convolutional layer (CL) is a layer that includes a filter and an activation function, and can generate a feature map from an image (I). The pooling layer (PL) can remove noise from the extracted feature map and adjust the size of the feature map.

[0076] For the sake of convenience of explanation, the first neural network (NN1) is shown as including one convolutional layer (CL) and one pooling layer (PL) each. However, the first neural network (NN1) can include a plurality of convolutional layers (CL) and pooling layers (PL), and the convolutional layers (CL) and pooling layers (PL) can have various structures according to the purpose of the first neural network (NN1).

[0077] The number of feature maps generated from the image (I) can vary according to the number of filters included in the convolutional layer (CL). In other words, a plurality of feature maps may be output from one image through the convolutional layer (CL) and the pooling layer (PL), and the feature maps can be referred to as layers (L1, L2, L3, L4, L5~Ln).

[0078] The extraction unit 300 can extract feature values (F) by flattening and concatenating the feature maps generated from the image (I). The generated feature values can correspond to time series data (D).

[0079] FIG. 5 shows a method for calculating the SOH of a battery cell diagnosis device according to an embodiment disclosed in this document. The calculation unit 400 receives the feature value (F), and the second neural network (NN2) can learn based on the received feature value (F). Also, the learned second neural network (NN2) can calculate or predict the SOH based on the feature value (F). The second neural network (NN2) can include a hidden layer (HL), and illustratively, the hidden layer (HL) is shown as having a cyclic structure, but according to an embodiment, a neural network with another structure can be selected. The hidden layer (HL) can learn via a recursive algorithm and predict or calculate the SOH.

[0080] FIG. 6 shows a method for calculating the SOH of a battery cell diagnosis device according to another embodiment disclosed in this document. The calculation unit 400 receives the operating condition (X) from the collection unit 100 and can calculate a prediction function (PF) via the third neural network (NN3). The third neural network (NN3) can illustratively be an RNN, and the calculated prediction function (PF) can correspond to any feature value.

[0081] The calculation unit 400 can concatenate the feature value (F) with the prediction function (PF). Illustratively, the feature value (F) extracted in FIG. 2 can be concatenated with the prediction function (PF) obtained from the operating condition X(t s , i s , C s ) because it is based on the time-series data (D) collected at that time, so that it can be concatenated with the prediction function (PF) obtained from the operating condition (X).

[0082] The second neural network (NN2) included in the calculation unit 400 can learn based on the concatenated feature value and prediction function (F, PF), and can calculate the SOH based on the concatenated feature value and prediction function (F, PF). The second neural network (NN2) can reflect the change in SOH due to the operating condition by using the prediction function (PF) for learning.

[0083] FIG. 7 is a flowchart showing an operation method of a battery cell diagnosis apparatus according to an embodiment disclosed in this document. Referring to FIG. 7, the battery cell diagnosis apparatus can collect time-series data according to the operating conditions of the battery cell (S100). As described above, the operating conditions of the battery cell may mean conditions that are kept constant during the charge and discharge operations of the battery cell. Also, the collected time-series data may be data showing the operating characteristics, which are result values obtained during the charge and discharge experiment according to the operating conditions of the battery cell, over time.

[0084] Exemplarily, the operating conditions may be the charge and discharge current of the battery cell, the temperature during charge and discharge of the battery cell, the charge and discharge voltage of the battery cell, the number of charge and discharge cycles of the battery cell, and the like. The time-series data may be data obtained by sampling the operating characteristics at a predetermined time interval.

[0085] The battery cell diagnosis apparatus can convert the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension (S200). The battery cell diagnosis apparatus can convert the time-series data into an image by the RP method. The RP method may be a method of imaging time-series data by converting the time-series data into a spatial trajectory of a predetermined dimension and representing the distances between points located on the converted spatial trajectory by a distance matrix. According to an embodiment, the predetermined dimension may be two-dimensional.

[0086] The battery cell diagnosis apparatus can extract a feature value from the image (S300). The battery cell diagnosis apparatus can use a first neural network to extract the feature value. Exemplarily, the first neural network may be a convolutional neural network including a convolutional layer and a pooling layer. More specifically, the battery cell diagnosis apparatus can use AlexNet, GoogLeNet, ResNet, or the like to extract the feature value from the image.

[0087] The battery cell diagnosis device can calculate the remaining life of the battery cell based on the extracted characteristic values (S400). The battery cell diagnosis device can train a second neural network based on the characteristic values, and can predict the state of health (SOH) of the battery cell using the trained second neural network. The second neural network may be a deep neural network including a hidden layer. More specifically, the battery cell diagnosis device can use an RNN or an LSTM, etc. to predict the SOH.

[0088] FIG. 8 is a flowchart showing a method of operating a battery cell diagnosis device according to another embodiment disclosed in this document. With reference to FIG. 8, a method by which the battery cell diagnosis device predicts the SOH of the battery cell reflecting the prediction function can be described.

[0089] After extracting the characteristic values from the image (S300), the battery cell diagnosis device can predict the SOH by a method different from the method described with reference to FIG. 7 (S400').

[0090] The battery cell diagnosis device can generate a prediction function for the operating conditions based on a third neural network (S410). The prediction function is a value generated by calculating the operating conditions through the third neural network. The battery cell diagnosis device can concatenate the prediction function with the corresponding characteristic values (S420), and can predict the remaining life of the battery cell based on the prediction function and the characteristic values (S430). The battery cell diagnosis device can improve the prediction accuracy of the SOH of the second neural network by concatenating the prediction function and the characteristic values and training the second neural network.

[0091] FIG. 9 is a graph comparing the remaining life prediction accuracies according to the method of operating a diagnosis device according to an embodiment disclosed in this document. The True graph in Fig. 9 plots the SOH values calculated from a plurality of operating conditions and time-series data with a maximum value of 1, and the Predict graph plots the SOH values predicted by the battery cell diagnostic device based on the operating conditions and time-series data with a maximum value of 1. The horizontal axis of the graph may be the number of samples used for prediction.

[0092] Graphs A and B in Fig. 9 are graphs that predict SOH based on a model learned without image conversion of time-series data. However, B is a graph of a model learned by reflecting a prediction function based on operating conditions. It can be seen that the B model has higher prediction accuracy compared to A.

[0093] Graphs C and D in Fig. 9 are graphs that predict SOH based on a model learned by image-converting time-series data. However, D is a graph that predicts SOH by reflecting a prediction function based on operating conditions. It can be confirmed that the prediction accuracy of the models (C, D) learned through image conversion is improved compared to the models (A, B) learned without image conversion. Also, when learning by reflecting the prediction function, the prediction accuracy can be further improved.

[0094] Fig. 10 is a block diagram showing the hardware configuration of a computing system for performing the operation method of the battery cell diagnostic device according to an embodiment disclosed in this document.

[0095] Referring to Fig. 10, a computing system 1000 according to an embodiment disclosed in this document may include an MCU 1010, a memory 1020, an input / output I / F 1030, and a communication I / F 1040.

[0096] The MCU 1010 may be a processor that executes various programs stored in the memory 1020 (for example, a program for collecting the voltage or current of the battery pack, a relay control program included in the battery pack, a remaining life calculation program for the battery cell, a capacity degradation diagnosis program for the battery cell, a resistance degradation determination program for the battery cell, etc.), processes various information including the remaining life information of the battery cell, the capacity degradation information of the battery cell, and the resistance degradation information of the battery cell through such programs, and performs the operation of the battery cell diagnosis device shown in FIG. 1 described above.

[0097] The memory 1020 can store various programs related to the collection and diagnosis of the battery's log information. Further, the memory 1020 can store various information such as the current of the battery, the voltage, the charge / discharge condition information, the voltage information of the battery cell in the charge / discharge cycle section within the set number of times, and the dQ / dV information of the battery cell in the charge / discharge cycle section within the set number of times.

[0098] Such a memory 1020 may be provided in plural as needed. The memory 1020 may be a volatile memory or a non-volatile memory. As the volatile memory, the memory 1020 can use RAM, DRAM, SRAM, etc. As the non-volatile memory, the memory 1020 can use ROM, PROM, EAROM, EPROM, EEPROM, flash memory, etc. The examples of the memory 1020 listed above are merely illustrative and are not limited to these examples.

[0099] The input / output I / F 1030 can provide an interface that connects between input devices (not shown) such as a keyboard, a mouse, a touch panel, etc., output devices such as a display (not shown), and the MCU 1010 so as to enable data transmission and reception.

[0100] The communication I / F 1040 is configured to be able to transmit and receive various data with a server, and may be various devices that can support wired or wireless communication. For example, the battery cell diagnostic device can transmit and receive a relay control program included in the battery pack, currents of various battery packs, current, dQ / dV values, charge / discharge condition information, remaining life, capacity degradation, or resistance degradation information from an externally provided server via the communication I / F 1040.

[0101] As described above, the computer program according to one embodiment disclosed in this document may be recorded in the memory 1020 and processed by the MCU 1010, and may be realized as a module that performs each function shown in FIG. 1, for example.

[0102] The above description merely exemplarily explains the technical idea disclosed in this document. Those having ordinary knowledge in the technical field to which the embodiments disclosed in this document belong can make various modifications and variations without departing from the essential characteristics of the embodiments disclosed in this document.

[0103] Therefore, the embodiments disclosed in this document are not for limiting the technical idea disclosed in this document but for explaining it, and the scope of the technical idea disclosed in this document is not limited by such embodiments. The protection scope of the technical idea disclosed in this document should be interpreted according to the scope of the following claims, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this document.

Claims

1. A collecting unit that collects time-series data of a battery cell according to operating conditions, A conversion unit that converts the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension, An extraction unit that extracts feature values from the image, A calculation unit that calculates the life of the battery cell based on the feature values, A battery cell diagnosis device comprising:

2. The battery cell diagnosis device according to claim 1, wherein the image is an image corresponding to a two-dimensional spatial trajectory.

3. The battery cell diagnosis device according to claim 1 or 2, wherein the time-series data is data corresponding to changes in the operating characteristics of the battery cell over time.

4. The battery cell diagnosis device according to claim 1 or 2, wherein the extraction unit extracts the feature values via a first neural network.

5. The battery cell diagnosis device according to claim 4, wherein the first neural network is a convolutional neural network including a convolutional layer and a pooling layer.

6. The battery cell diagnosis device according to claim 1 or 2, wherein the calculation unit calculates the life via a second neural network.

7. The battery cell diagnosis device according to claim 6, wherein the second neural network is a deep neural network including multiple hidden layers.

8. The operating conditions correspond to the feature values, The battery cell diagnosis device according to claim 1 or 2, wherein the calculation unit generates a prediction function for the operating conditions via a third neural network.

9. The battery cell diagnosis device according to claim 8, wherein the third neural network is a deep neural network including multiple hidden layers.

10. The battery cell diagnosis device according to claim 8, wherein the calculation unit calculates the life based on the prediction function and the feature values.

11. The battery cell diagnosis device according to claim 1 or 2, wherein the conversion unit converts the time-series data into the spatial trajectory and uses the spatial trajectory to convert the time-series data into the image.

12. The conversion unit represents the distances between points located on the spatial trajectory as a distance matrix, The battery cell diagnosis device according to claim 11, wherein the region corresponding to the distance matrix is converted into the image based on the distance values of the distance matrix.

13. A step of collecting time-series data of a battery cell according to operating conditions, Converting the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension; Extracting feature values from the image; Calculating the lifespan of the battery cell based on the feature values; A battery cell diagnosis method including the above steps.

14. The step of converting the time-series data into an image corresponding to a spatial trajectory of a predetermined dimension includes: Converting the time-series data into a spatial trajectory; Converting the time-series data into the image using the spatial trajectory. The battery cell diagnosis method according to claim 13.

15. The step of converting the time-series data into the image using the spatial trajectory includes: Representing the distances between points located on the spatial trajectory by a distance matrix; Visualizing the region corresponding to the distance matrix based on the distance values of the distance matrix. The battery cell diagnosis method according to claim 14.

16. The step of extracting feature values from the image includes: Extracting the feature values via a first neural network. The battery cell diagnosis method according to any one of claims 13 to 15.

17. The step of calculating the lifespan includes: Calculating the lifespan via a second neural network; The second neural network is a deep neural network including multiple hidden layers. The battery cell diagnosis method according to any one of claims 13 to 15.

18. The operating conditions correspond to the feature values; The step of calculating the lifespan includes: Generating a prediction function for the operating conditions via a third neural network; Concatenating the prediction function and the feature values; Predicting the lifespan based on the prediction function and the feature values. The battery cell diagnosis method according to any one of claims 13 to 15.

Images