Battery cell life diagnosis device and operation method thereof
By converting the time-series data of individual battery cells into images and using neural network technology to extract feature values and calculate lifespan, the accuracy problem of individual battery cell lifespan diagnosis is solved, enabling early prediction of battery performance and safe management.
Patent Information
- Application Number
- CN202380035497.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-06-02
- Filing Date
- 2023-04-03
- Publication Date
- 2024-11-26
AI Technical Summary
Existing technologies are insufficient to effectively diagnose the lifespan of individual battery cells, especially the degradation of their capacity and resistance during repeated charging and discharging, leading to increased safety hazards.
By collecting time-series data of individual battery cells, converting it into images, and extracting feature values using convolutional neural networks, the lifespan of individual battery cells is calculated using deep neural networks. The system includes a collection unit, a conversion unit, an extraction unit, and a calculation unit, and uses a recursive graph scheme and multi-layer neural networks for data processing.
It improves the accuracy and efficiency of battery cell life diagnosis, enabling early prediction of battery performance degradation and reducing safety risks.
Smart Images

Figure CN119032282A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims priority to and the benefit of Korean Patent Application No. 10-2022-0067847 filed in the Korean Intellectual Property Office on June 2, 2022, the entire contents of which are incorporated herein by reference. Technical Field
[0003] The embodiments disclosed herein relate to a battery cell life diagnosis device and an operating method thereof. Background Art
[0004] Recently, research and development of secondary batteries have been actively conducted. Secondary batteries as rechargeable / dischargeable batteries may include all conventional nickel (Ni) / cadmium (Cd) batteries, nickel / metal hydride (MH) batteries, etc., and more recently lithium-ion batteries. Lithium-ion batteries have a much higher energy density than conventional Ni / Cd batteries, Ni / MH batteries, etc. In addition, lithium-ion batteries can be made small and light, so that lithium-ion batteries have been used as power sources for mobile devices, and recently their use range has been expanded to power sources for electric vehicles, thereby attracting attention as a next-generation energy storage medium.
[0005] Batteries tend to deteriorate when they are repeatedly charged and discharged. For example, when a battery is repeatedly charged and discharged, its capacity and resistance may deteriorate, and its remaining life may decrease. In addition, the degree of deterioration and remaining life of the battery may vary depending on the conditions of use.
[0006] When the remaining life of a battery decreases rapidly, safety issues may arise in the use of the battery. Therefore, a method for diagnosing the battery life in advance to prevent the risk of sudden battery performance degradation is needed. Summary of the invention
[0007] [Technical issues]
[0008] The embodiments disclosed herein are directed to providing an apparatus for converting operation characteristic data of a battery cell collected in a time series corresponding to an operation condition into an image to effectively calculate the life of the battery cell and an operation method of the apparatus.
[0009] Embodiments disclosed herein are directed to providing an apparatus for diagnosing a life span of a battery cell based on an image converted from time series data and an operating method of the apparatus.
[0010] Technical problems of the embodiments disclosed herein are not limited to the above-mentioned technical problems, and other unmentioned technical problems will be clearly understood by a person of ordinary skill in the art through the following description.
[0011] [Technical solution]
[0012] According to an embodiment disclosed herein, a battery cell diagnostic device includes: a collection unit configured to collect time series data of a battery cell according to operating conditions; a conversion unit configured to convert the time series data into an image corresponding to a spatial trajectory of a predetermined dimension; an extraction unit configured to extract feature values from the image; and a calculation unit configured to calculate the life of the battery cell based on the feature values.
[0013] In an embodiment, the image may be an image corresponding to a trajectory in two-dimensional space.
[0014] In an embodiment, the time series data may be data corresponding to changes in operating characteristics of the battery cell with respect to time.
[0015] In an embodiment, the extracting unit may also be configured to extract feature values through a first neural network.
[0016] In an embodiment, the first neural network may be a convolutional neural network including a convolutional layer and a pooling layer.
[0017] In an embodiment, the calculation unit may be further configured to calculate the lifespan by means of a second neural network.
[0018] In an embodiment, the second neural network may be a deep neural network including a plurality of hidden layers.
[0019] In an embodiment, the operating condition may correspond to a feature value, and the computing unit may be further configured to generate a prediction function for the operating condition through a third neural network.
[0020] In an embodiment, the third neural network may be a deep neural network including a plurality of hidden layers.
[0021] In an embodiment, the calculation unit may be further configured to calculate the lifespan based on the prediction function and the characteristic value.
[0022] In an embodiment, the conversion unit may be further configured to convert the time series data into a spatial trajectory, and convert the time series data into an image by using the spatial trajectory.
[0023] In an embodiment, the conversion unit may be further configured to express the distances between points located on the spatial trajectory as a distance matrix, and the region corresponding to the distance matrix may be converted into an image based on the distance values of the distance matrix.
[0024] A battery cell diagnosis method according to an embodiment disclosed herein includes 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, and calculating the life of the battery cell based on the feature values.
[0025] In an embodiment, converting the time series data into an image corresponding to a spatial trajectory of a predetermined dimension may include: converting the time series data into the spatial trajectory, and converting the time series data into the image by using the spatial trajectory.
[0026] In an embodiment, converting the time series data into an image by using the spatial trajectory may include: expressing distances between points located on the spatial trajectory as a distance matrix, and visualizing a region corresponding to the distance matrix based on distance values of the distance matrix.
[0027] In an embodiment, extracting feature values from the image may include extracting feature values through a first neural network.
[0028] In an embodiment, the calculation of the lifespan may include calculating the lifespan by a second neural network, and the second neural network may be a deep neural network including a plurality of hidden layers.
[0029] In an embodiment, the operating condition may correspond to the feature value, and the calculation of the lifetime may include generating a prediction function for the operating condition by a third neural network, connecting the prediction function with the feature value, and predicting the lifetime based on the prediction function and the feature value.
[0030] [Beneficial Effects]
[0031] A battery cell life diagnosis device and an operating method thereof according to embodiments disclosed herein may calculate the status of a battery cell.
[0032] A battery cell diagnosis device and an operating method thereof according to embodiments disclosed herein may calculate the state of a battery cell by converting time series data on operating characteristics of a battery cell into an image having a preset dimension.
[0033] The battery cell diagnosis device and the operating method thereof according to the embodiments disclosed herein may calculate the remaining life of the battery cell by analyzing image data through a neural network.
[0034] The battery cell diagnosis device and the operating method thereof according to the embodiments disclosed herein may improve the accuracy of calculating the lifespan of the battery cell based on images corresponding to the operating characteristics of the battery cell and the operating conditions of the battery cell.
[0035] Furthermore, various effects directly or indirectly recognized from the present disclosure can be provided. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 is a block diagram of a battery cell diagnosis device according to an embodiment disclosed herein.
[0037] Figure 2The time series data processing method of the battery cell diagnosis device according to the embodiment disclosed herein is shown.
[0038] Figure 3a and Figure 3b An image conversion method of a battery cell diagnosis device according to an embodiment disclosed herein is shown.
[0039] Figure 4 A feature value extraction method of a battery cell diagnosis device according to an embodiment disclosed herein is shown.
[0040] Figure 5 The SoH calculation method of the battery cell diagnosis device according to the embodiment disclosed herein is shown.
[0041] Figure 6 A SoH calculation method of a battery cell diagnosis device according to another embodiment disclosed herein is shown.
[0042] Figure 7 is a flow chart of an operating method of a battery cell diagnosis device according to an embodiment disclosed herein.
[0043] Figure 8 is a flow chart of an operating method of a battery cell diagnosis device according to another embodiment disclosed herein.
[0044] Fig. 9 is a graph comparing the lifetime prediction accuracy with the operating method of the diagnostic device according to the embodiment disclosed herein.
[0045] Fig.10 is a block diagram illustrating a hardware configuration of a computing system for executing an operating method of a battery cell diagnosis device according to an embodiment disclosed herein. DETAILED DESCRIPTION
[0046] Hereinafter, the embodiments disclosed in this document will be described in detail with reference to the exemplary drawings. When adding reference numerals to the components of the respective drawings, it should be noted that the same components are given the same reference numerals even if they are indicated in different drawings. In addition, when describing the embodiments disclosed in this document, when it is determined that the detailed description of the related known configuration or function interferes with the understanding of the embodiments disclosed in this document, its detailed description will be omitted.
[0047] To describe the components of the embodiments disclosed herein, terms such as first, second, A, B, (a), (b), etc. may be used. These terms are only used to distinguish one component from another component and do not limit the component to the nature, order, sequence, etc. of the component. The terms used herein, including technical and scientific terms, have the same meaning as those commonly understood by those skilled in the art, as long as these terms are not defined differently. In general, terms defined in commonly used dictionaries should be interpreted as having the same meaning as the contextual meaning of the relevant field, and should not be interpreted as having an idealized or exaggerated meaning unless explicitly defined in this application.
[0048] Figure 1 is a block diagram of a battery cell diagnosis device according to an embodiment disclosed herein.
[0049] Reference Figure 1 , the battery cell diagnosis device 10 according to the embodiment disclosed herein may include a collection unit 100 , a conversion unit 200 , an extraction unit 300 , and a calculation unit 400 .
[0050] The collecting unit 100 may obtain the operating conditions of the battery cells and time series data of the battery cells corresponding to the operating conditions.
[0051] The state of health (SoH), which is an indicator indicating the degradation of the performance of a battery cell, can be expressed as a percentage of the capacity or resistance of the battery cell due to aging relative to the initial capacity. Changes in the SoH may occur due to an increase in resistance caused by the polarization of the chemical substances included in the battery cell. When the performance of the battery cell is degraded to a predetermined threshold point or lower, the battery cell may need to be replaced. Generally, when the initial capacity is at the 80% level, it can be determined as the end of life time of the battery cell.
[0052] The operating conditions may be conditions that are kept constant to obtain operating characteristics during a charge / discharge experiment for measuring a performance index of a battery cell. The operating characteristics of a battery cell may be values obtained during a charge / discharge experiment according to the operating conditions, and may be values that change over time. The time series data may be data obtained by measuring the operating characteristics of a battery cell in a predetermined time period. Therefore, the time series data may indicate changes in the operating characteristics of a battery cell over time.
[0053] The performance index of a battery cell may refer to, for example, the aforementioned SoH or state of charge (SoC, remaining capacity), etc.
[0054] The charge / discharge experiment may mean an experiment for detecting operating characteristics of a battery cell while maintaining an operating condition. The operating condition may mean an experimental environment that affects an output of a battery cell.
[0055] When the operating temperature and the charge / discharge current of the battery cell are the operating conditions, the operating characteristic of the battery cell may be the charge / discharge voltage. In this case, the time series data may be data indicating changes in the charge / discharge voltage of the battery cell over time.
[0056] As the number of repetitions of charge / discharge increases, the characteristics of the battery cell may deteriorate. The collection unit 100 may obtain time series data of the battery cell for each number of repetitions of charge / discharge. According to an embodiment, the number of repetitions of charge / discharge may be an operating condition of the battery cell.
[0057] The operating conditions may be changed according to the time series data to be obtained, and according to another embodiment, the operating characteristics may be the impedance, internal resistance, capacity, etc. of the battery cell.
[0058] According to an embodiment, the collection unit 100 may include a measuring device for monitoring the temperature, charge / discharge current and charge / discharge voltage of the battery cell to obtain time series data corresponding to the operating conditions. The measuring device may also monitor the impedance, internal resistance, capacity, etc. of the battery cell.
[0059] In addition, according to another embodiment, the collection unit 100 may receive matching of the operating conditions of the battery cells and the time series data of the battery cells corresponding to the operating conditions from an external source.
[0060] The time series data may be data sampled or normalized at preset time intervals by the collection unit 100. For example, when the operating temperature and the charge / discharge current of the battery cell are kept constant, once the collection unit 100 collects raw data about the charge and discharge voltage of the battery cell over time, the raw data may be divided at preset time intervals to obtain a plurality of time series data.
[0061] The conversion unit 200 may 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 may be performed using a recurrence graph (RP) scheme. The RP scheme may be a conversion scheme for converting the time series data into an image.
[0062] The RP scheme can convert the time series data into a spatial trajectory of a predetermined dimension and express the distance between points on the spatial trajectory as a distance matrix to convert the time series data into an image. In this case, the spatial trajectory may mean a moving trajectory between data included in the time series data.
[0063] The conversion unit 200 may visualize the time series data as an image corresponding to a spatial trajectory of a predetermined dimension based on the distance values of the distance matrix.
[0064] According to an embodiment, when the distance value of the distance matrix is greater than or equal to a threshold value, the conversion unit 200 may visualize the time series data as an image by expressing a region corresponding to the distance matrix in two dimensions.
[0065] According to another embodiment, by expressing the color of a region corresponding to a distance matrix in two dimensions to change according to the distance value of the distance matrix, the region may be visualized as an image including a red channel, a green channel, and a blue channel.
[0066] The conversion unit 200 may convert the time series data into an image through the RP scheme to clearly reveal features that are difficult to identify from the time series data.
[0067] The extraction unit 300 may extract feature values of the image through the first neural network. According to an embodiment, the extraction unit 300 may generate a plurality of feature maps from the image. The extraction unit 300 may extract feature values by processing the generated feature maps and send the extracted feature values to the calculation unit 400.
[0068] The first neural network may be a convolutional neural network algorithm. The convolutional neural network may be a deep learning algorithm including a convolution layer and a pooling layer, and the convolution layer may include a filter (convolution kernel) for extracting features of an image and an activation function for calculating input data through the filter as a nonlinear value. The image through the convolution layer may be output as a plurality of feature maps.
[0069] The pooling layer can remove feature-irrelevant noise from the feature map.
[0070] Depending on the connection scheme between the convolution layer and the pooling layer, the convolutional neural network can have different architectures. The convolutional neural network for extracting feature values from an image can use, for example, AlexNet using overlapping convolutional layers, GoogLeNet including an inception module, ResNet including a residual block, etc.
[0071] The first neural network used in the extraction unit 300 may use an existing training model. According to an embodiment, the extraction unit 300 may use a GoogLeNet or ResNet model trained by a visual database (eg, the imageNet project) to perform feature value extraction through transfer learning.
[0072] According to another embodiment, the extraction unit 300 may adjust the weight of the convolution kernel used in the calculation, or adjust the architecture of the trained convolutional neural network in the convolution layer of the first neural network, thereby optimizing the first neural network.
[0073] The first neural network used in the extraction unit 300 may vary with the amount of images converted from the time series data. That is, when the amount of images converted from the time series data is small, the neural network may use an existing trained model to perform feature value extraction through transfer learning, and when the amount of images converted from the time series data is large, the neural network may use an existing trained model by changing the architecture, weights, etc. of the existing trained model.
[0074] According to an embodiment, the extraction unit 300 may generate feature values by flattening and concatenating the generated plurality of feature maps. The number of feature maps may vary with the number of filters included in the first neural network. The feature map may be referred to as a layer.
[0075] The calculation unit 400 may predict or calculate the SoH of the battery cell through the second neural network. According to an embodiment, the second neural network may be a deep neural network algorithm. The calculation unit 400 may predict or calculate the SoH of the battery cell based on the feature values extracted from the image and the operating conditions.
[0076] A deep neural network may be a deep learning algorithm including a plurality of hidden layers, and may finely process a plurality of highly complex data by using the plurality of hidden layers.
[0077] The second neural network included in the calculation unit 400 may learn the time series data based on the feature values extracted by conversion into an image, thereby improving accuracy compared to directly learning the time series data to calculate the SoH of the battery cell.
[0078] When learning time series data and converting it to an image, the characteristics of the data may not be correctly reflected, and the predicted value may be biased, or the trend of the local area may be over-reflected, resulting in reduced calculation accuracy of the calculation unit 400.
[0079] For example, for time series data reflecting the change in capacity of a battery cell over time, noise may occur in the time series data due to capacity regeneration in which the capacity increases irregularly by removing residual reaction products, and the noise may affect SoH prediction.
[0080] When time series data is converted into an image and SoH is predicted based on feature values extracted from the image, the image can cyclically reflect changes in the time series data and easily extract features, thereby improving the SoH prediction accuracy of the calculation unit 400 .
[0081] Depending on how the hidden layers are connected, a deep neural network can have different architectures. For example, a deep neural network can use a recurrent neural network (RNN) with hidden layers (including a recurrent architecture), a long short-term memory (LSTM) with an added layer for memorizing previous states to improve the vanishing gradient phenomenon and long-term dependencies, etc.
[0082] The feature value input to the calculation unit 400 may be obtained by converting an image as a basis for feature value extraction from time series data, and may be analyzed together with sequential data by a second neural network. Sequential data may mean data that changes in time sequence.
[0083] According to another embodiment, the calculation unit 400 may generate a prediction function by using an operating condition corresponding to the feature value. The calculation unit 400 may link the feature value and the operating condition corresponding thereto.
[0084] The prediction function is a function that causes the second neural network to reflect changes in SoH with operating conditions, wherein when the second neural network learns the prediction function and the feature value by connecting them to each other, the tendency of the SoH to change with operating conditions can be reflected in the second neural network. Therefore, when the second neural network learns the prediction function and the feature value by connecting them, the SoH prediction accuracy of the calculation unit 400 can be improved.
[0085] The extracting unit 300 may extract feature values from the image converted from the time series data, wherein the random operating conditions may correspond to the feature values. The collecting unit 100 may send the operating conditions corresponding to the feature values to the calculating unit 400. The calculating unit 400 may generate a prediction function through a third neural network.
[0086] According to an embodiment, the third neural network may be a deep neural network like the second neural network.
[0087] Learning using the second neural network and the third neural network may be learning using information of different dimensions (feature values and operating conditions), which may be referred to as multimodal learning.
[0088] Figure 2 The time series data processing method of the battery cell diagnosis device according to the embodiment disclosed herein is shown.
[0089] exist Figure 2 A first graph G1 of time series data regarding a random operating condition X is shown in FIG.
[0090] At the preset temperature t s , preset charge / discharge current i s and preset charge / discharge times C sWhen the operation condition X is met, the collection unit 100 may collect the charge / discharge voltage V of the battery cell with respect to time T. That is, the first curve G1 may indicate the charge / discharge voltage V of the battery cell with respect to time T.
[0091] The collection unit 100 may collect a plurality of time series data by dividing the charge / discharge voltage V of the battery cell with respect to the time T by a preset time interval. Figure 2 FIG. 1 shows that the collecting unit 100 collects data from the first time T 1 To the second time T 2 The case of divided time series data D.
[0092] The collecting unit 100 may collect the time series data by dividing the time series data at random time intervals, and according to another embodiment, a plurality of time series data may include overlapping time domains.
[0093] Figure 3a and Figure 3b An image conversion method of a battery cell diagnosis device according to an embodiment disclosed herein is shown.
[0094] The conversion unit 200 may convert the time series data D into an image I using the RP scheme.
[0095] To describe the RP scheme, an example graph GX having a time axis T and a voltage axis V is shown. Figure 3a The process of converting the example graph GX into the example image IX is described in detail.
[0096] The example graph GX may be converted into an example spatial trajectory PX, which is a two-dimensional spatial trajectory. Points S1 , S2 , and S3 on the example spatial trajectory may indicate changes in values along the voltage axis V relative to changes in values along the time axis T.
[0097] That is, S1 may indicate the change from 0 to 1 along the voltage axis V during the change from 1 to 2 along the time axis T as coordinates (0, 1). Similarly, S2 may indicate the change from 1 to 2 along the voltage axis V during the change from 2 to 3 along the time axis T as coordinates (1, 2).
[0098] The spatial trajectory may indicate a movement trajectory of data included in the time series data. Therefore, a point located on the spatial trajectory may cyclically indicate a relationship between adjacent data included in the time series data.
[0099] The distance matrix may be, for example, a matrix including points located on a spatial trajectory. For example, the distance matrix may have a form such as (S1, S2) or (S1, S3).
[0100] The distance value of the distance matrix may mean the distance between points on the spatial trajectory. According to an embodiment, the time series data may be visualized as an image by expressing the color of the area corresponding to the distance matrix as changing with the distance value of the distance matrix. For example, the distance value of the distance matrix (S1, S2) may be the distance between S1 and S2 Similarly, the distance value of the distance matrix (S1, S3) can be the distance 2 between S1 and S3.
[0101] Referring to example image IX, the contrast of corresponding regions may change with the distance values of the distance matrix.
[0102] The RP scheme may express the distances between points located on a spatial trajectory as a distance matrix, wherein an image expressed by the RP scheme may have a symmetrical shape.
[0103] exist Figure 3b The RP scheme is shown in Figure 2 The processing of converting the time series data D into the image data I may reveal the characteristics of the image I converted by the RP scheme more clearly than the first graph G1 representing the time series data.
[0104] Figure 4 A feature value extraction method of a battery cell diagnosis device according to an embodiment disclosed herein is shown.
[0105] The extraction unit 300 may receive an image I and extract feature values from the image I received through the first neural network NN1.
[0106] The first neural network NN1 may include a convolutional layer CL and a pooling layer PL.
[0107] The convolution layer CL may include filters and activation functions and generate a feature map from the image I. The pooling layer PL may remove noise from the extracted feature map and adjust the size of the feature map.
[0108] Although the first neural network NN1 is shown to include one convolution layer CL and one pooling layer PL, the first neural network NN1 may include a plurality of convolution layers CL and pooling layers PL, which may have various architectures according to the purpose of the first neural network NN1.
[0109] The number of feature maps generated from an image I may vary with the number of filters included in the convolution layer CL. That is, after an image passes through the convolution layer CL and the pooling layer PL, multiple feature maps may be output, and the feature maps may be referred to as layer L. 1 , L 2 , L 3 , L 4 , L 5 ...、L n .
[0110] The extracting unit 300 may extract the feature value F by flattening and concatenating the feature map generated from the image I. The generated feature value may correspond to the time series data D.
[0111] Figure 5 The SoH calculation method of the battery cell diagnosis device according to the embodiment disclosed herein is shown.
[0112] The calculation unit 400 may receive the feature value F, and may train the second neural network NN2 based on the received feature value F. The trained second neural network NN2 may calculate or predict SoH based on the feature value F. The second neural network NN2 may include a hidden layer HL shown as having a circular architecture as an example, but a neural network having another architecture may be selected depending on the embodiment. The hidden layer HL may be trained using a recursive algorithm and predict or calculate SoH.
[0113] Figure 6 A SoH calculation method of a battery cell diagnosis device according to another embodiment disclosed herein is shown.
[0114] The calculation unit 400 may receive the operation condition X from the collection unit 100 and calculate the prediction function PF through the third neural network NN3. The third neural network NN3 may be, for example, an RNN, and the calculated prediction function PF may correspond to a random feature value.
[0115] The calculation unit 400 may connect the feature value F to the prediction function PF. For example, Figure 2 The extracted eigenvalue F can be based on the operating condition X(t s ,i s , C s ) and can be connected to the prediction function PF obtained from the operating condition X.
[0116] The second neural network NN2 included in the calculation unit 400 may be trained based on the connected feature values and the prediction function (F, PF), and calculates SoH based on the connected feature values and the prediction function (F, PF). The second neural network NN2 may reflect the SoH change according to the operating conditions by training using the prediction function PF.
[0117] Figure 7 is a flow chart of an operating method of a battery cell diagnosis device according to an embodiment disclosed herein.
[0118] refer to Figure 7In operation S100, the battery cell diagnostic device may collect time series data according to the operating conditions of the battery cells. As described above, the operating conditions of the battery cells may mean conditions that remain constant during the charge / discharge of the battery cells. The collected time series data may be data showing operating characteristics obtained during a charge / discharge experiment according to the operating conditions with respect to time.
[0119] For example, the operating conditions may include a charge / discharge current of a battery cell, a temperature of the battery cell while charging / discharging, a charge / discharge voltage of the battery cell, a charge / discharge number of times of the battery cell, and the like.
[0120] The time series data may be data obtained by sampling operating characteristics at predetermined time intervals.
[0121] In operation S200, the battery cell diagnostic device may convert the time series data into an image corresponding to a spatial trajectory of a predetermined dimension. The battery cell diagnostic device may convert the time series data into an image according to the RP scheme. The RP scheme may convert the time series data into a spatial trajectory of a predetermined dimension, and express the distance between points on the converted spatial trajectory as a distance matrix to convert the time series data into an image. According to an embodiment, the predetermined dimension may be two-dimensional.
[0122] In operation S300, the battery cell diagnosis device may extract feature values from the image. The battery cell diagnosis device may use a first neural network to extract feature values, and for example, the first neural network may be a convolutional neural network including a convolutional layer and a pooling layer. More specifically, the battery cell diagnosis device may use AlexNet, GoogLeNet, ResNet, etc. to extract feature values from the image.
[0123] In operation S400, the battery cell diagnosis device may calculate the life of the battery cell based on the extracted feature value. The battery cell diagnosis device may train a second neural network based on the feature value and predict the SoH of the battery cell by using the trained second neural network. The second neural network may be a deep neural network including a hidden layer, and more specifically, the battery cell diagnosis device may use RNN, LSTM, etc. to predict SoH.
[0124] Figure 8 is a flow chart of an operating method of a battery cell diagnosis device according to another embodiment disclosed herein.
[0125] refer to Figure 8 , a method for a battery cell diagnosis device to predict the SoH of a battery cell by reflecting a prediction function will be described.
[0126] The battery cell diagnosis device may extract a feature value from the image in operation S300, and then use the feature value to compare with the reference image. Figure 7 The method described in operation S400' is different from the method for predicting SoH.
[0127] In operation S410 , the battery cell diagnosis device may generate a prediction function for the operating condition based on the third neural network.
[0128] The prediction function is a value generated by calculating the operating condition using the third neural network, and the battery cell diagnostic device may connect the prediction function with the corresponding characteristic value in operation S420, and predict the life of the battery cell based on the prediction function and the characteristic value in operation S430. The battery cell diagnostic device may learn the second neural network by connecting the prediction function with the characteristic value, thereby improving the accuracy of SoH prediction of the second neural network.
[0129] Fig. 9 is a graph for comparing the accuracy of life prediction with the method of operation of the diagnostic device according to the embodiments disclosed herein.
[0130] Fig. 9 The real graph depicts the SoH value calculated according to the plurality of operating conditions and the time series data as a maximum value of 1, and the predicted graph depicts the SoH value predicted by the battery cell diagnosis device based on the operating conditions and the image series data as a maximum value of 1. The horizontal axis of the graph may indicate the number of samples used for prediction.
[0131] Fig. 9 A and B in FIG. 1 are graphs for predicting SoH based on models trained without converting time series data into images. However, B is a graph of a model trained by reflecting a prediction function based on operating conditions. The model of B may have higher prediction accuracy than the model of A.
[0132] Fig. 9 C and D in FIG. 1 are graphs for predicting SoH based on a model trained by converting time series data into images. However, D is a graph for predicting SoH by reflecting a prediction function based on operating conditions. It can be seen that the prediction accuracy of models C and D trained by image conversion is improved when compared with models A and B not trained by image conversion. In addition, when training is performed by reflecting a prediction function, the prediction accuracy can be further improved.
[0133] Fig.10 is a block diagram illustrating a hardware configuration of a computing system for executing an operating method of a battery cell diagnosis device according to an embodiment disclosed herein.
[0134] refer to Fig.10, the computing system 1000 according to the embodiments disclosed herein may include an MCU 1010 , a memory 1020 , an input / output I / F 1030 , and a communication I / F 1040 .
[0135] The MCU 1010 may be a processor that executes various programs stored in the memory 1020 (e.g., a battery pack voltage or current collection program, a relay control program included in the battery pack, a battery cell remaining life calculation program, a battery cell capacity degradation diagnosis program, a battery cell resistance degradation determination program, etc.), processes various information including battery cell remaining life information, battery cell capacity degradation information, and battery cell resistance degradation information through these programs, and performs Figure 1 The above functions of the battery cell diagnostic device are shown.
[0136] The memory 1020 may store various programs related to battery log information collection and diagnosis, etc. The memory 1020 may store various information, including battery current, voltage, and charge / discharge condition information, voltage information of battery cells in a preset number of charge / discharge cycles, dQ / dV information of battery cells in a preset number of charge / discharge cycles, etc.
[0137] Depending on the need, a plurality of memories 1020 may be provided. The memory 1020 may be a volatile memory or a nonvolatile memory. For the memory 1020 as a volatile memory, a random access memory (RAM), a dynamic RAM (DRAM), a static RAM (SRAM), etc. may be used. For the memory 1020 as a nonvolatile memory, a read-only memory (ROM), a programmable ROM (PROM), an electrically variable ROM (EAROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), a flash memory, etc. may be used. The examples listed above of the memory 1020 are only examples and are not limited thereto.
[0138] The input / output I / F 1030 may provide an interface for transmitting and receiving data by connecting an input device (not shown) such as a keyboard, a mouse, a touch pad, etc. and an output device (not shown) such as a display, etc. to the MCU 1010 .
[0139] The communication I / F 1040, which is a component capable of transmitting and receiving various data to and from a server, may be various devices capable of supporting wired or wireless communication. For example, the battery cell diagnostic device may transmit and receive information such as current, current, value dQ / dV, charge / discharge condition information, remaining life, capacity degradation, or resistance degradation information of various battery packs separately provided from an external server through the communication I / F 1040, including a relay control program in the battery pack.
[0140] In this way, the computer program according to the embodiment disclosed herein may be recorded in the memory 1020 and processed by the MCU 1010, thereby being implemented as an execution Figure 1 Modules with the functions shown.
[0141] The above description merely illustrates the technical idea of the present disclosure, and a person skilled in the art to which the embodiments disclosed herein pertain will be able to make various modifications and variations without departing from the essential features of the embodiments of the present disclosure.
[0142] Therefore, the embodiments disclosed herein are intended to describe rather than limit the technical spirit of the embodiments disclosed herein, and the scope of the technical spirit of the present disclosure is not limited to these embodiments disclosed herein. The protection scope of the technical spirit disclosed herein should be interpreted by the attached claims, and all technical spirits within the same scope should be understood to be included in the scope of the present disclosure.
Claims
1. A battery cell diagnostic device, include: a collecting unit configured to collect time series data of the battery cells according to operating conditions; a conversion unit configured to convert the time series data into an image corresponding to a spatial trajectory of a predetermined dimension; an extraction unit, wherein the extraction unit is configured to extract a feature value from the image; as well as A calculation unit is configured to calculate the life of the battery cell based on the characteristic value.
2. The battery cell diagnostic device according to claim 1, in, The image is an image corresponding to the two-dimensional space trajectory.
3. The battery cell diagnostic device according to claim 1, in, The time series data is data corresponding to changes in operation characteristics of the battery cell with respect to time.
4. The battery cell diagnostic device according to claim 1, in, The extracting unit is further configured to extract the feature value through a first neural network.
5. The battery cell diagnostic device according to claim 4, in, The first neural network is a convolutional neural network including a convolutional layer and a pooling layer.
6. The battery cell diagnostic device according to claim 1, in, The calculation unit is further configured to calculate the lifespan through a second neural network.
7. The battery cell diagnosis device according to claim 6, in, The second neural network is a deep neural network including multiple hidden layers.
8. The battery cell diagnostic device according to claim 1, in, The operating condition corresponds to the characteristic value, and The computing unit is further configured to generate a prediction function for the operating condition through a third neural network.
9. The battery cell diagnosis device according to claim 8, in, The third neural network is a deep neural network including multiple hidden layers.
10. The battery cell diagnosis device according to claim 8, in, The calculation unit is further configured to calculate the lifetime based on the prediction function and the characteristic value.
11. The battery cell diagnosis device according to claim 1, in, The conversion unit is further configured to convert the time series data into the spatial trajectory, and convert the time series data into the image by using the spatial trajectory.
12. The battery cell diagnosis device according to claim 11, in, The conversion unit is further configured to express the distances between points on the spatial trajectory as a distance matrix, and A region corresponding to the distance matrix is converted into the image based on the distance values of the distance matrix.
13. A battery cell diagnosis method, include: Collect time series data of battery cells 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; as well as The life of the battery cell is calculated based on the characteristic value.
14. The battery cell diagnosis method according to claim 13, in, The conversion of the time series data into the image corresponding to the spatial trajectory of the predetermined dimension comprises: converting the time series data into a spatial trajectory; and The time series data is converted into the image by using the spatial trajectory.
15. The battery cell diagnosis method according to claim 14, in, The conversion of the time series data to the image by using the spatial trajectory comprises: expressing the distances between points on the spatial trajectory as a distance matrix; and Regions corresponding to the distance matrix are visualized based on distance values of the distance matrix.
16. The battery cell diagnosis method according to claim 13, in, Extracting the feature value from the image includes extracting the feature value through a first neural network.
17. The battery cell diagnosis method according to claim 13, in, The calculating of the lifespan includes calculating the lifespan by a second neural network, and The second neural network is a deep neural network including multiple hidden layers.
18. The battery cell diagnosis method according to claim 13, in, The operating condition corresponds to the characteristic value, and The calculation of the lifespan includes: generating a prediction function for the operating condition by a third neural network; connecting the prediction function to the feature value; and The lifetime is predicted based on the prediction function and the characteristic value.
Citation Information
Cited By
Battery life prediction method based on imaging single-cycle sampling amount and related device
CN119805235A
Battery life prediction method based on imaged single-cycle sampling quantity and related device
CN119805235B
Battery remaining service life prediction method and device, computer equipment, computer readable storage medium and computer program product
CN120428105A
